Task-Boundary Mode
How task-instance boundaries are drawn from the event stream. Applies to every Task SoP, Step SoP, and Variants view.
Intake
📝 Task SoP
Mined from 640 instances
Begin Case Processing
How 7 analysts execute this task across 171 cases — auto-generated Standard Operating Procedure with adherence scoring.
82.2
Avg Adherence Score (0-100)
Median 82.5 · Range 58.5–90.0
20.7 min
Avg Task Duration
0.1 min median
1.3
Avg Sub-Steps per Instance
21.4% have multiple sub-steps
78.6%
Reference Path Coverage
Instances following mined SoP exactly
Reference Baseline
78.6% of instances follow this exactly
Begin Case Processing — Mined SoP
The most common sub-step sequence observed across 863 task instances. Use the dropdown on the right to compare any deviation to this baseline.
1
View Intake Processing
0.0s
Phobos
97%
Variant Explorer
Reference
Begin Case Processing Reference Path
678 instances · 7 analysts
This is the mined reference path — 678 instances across 7 analysts follow this exact sequence. Treat as the proposed Standard Operating Procedure for Begin Case Processing.
Genuine Complexity
Complex-Case Execution Path
103 instances · 7 analysts
· Δ adherence +2.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 84.9 vs reference 82.2 (delta +2.7). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Genuine Complexity
Complex-Case Execution Path
29 instances · 5 analysts
· Δ adherence +7.8
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 90.0 vs reference 82.2 (delta +7.8). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
7 instances · 2 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Genuine Complexity
Complex-Case Execution Path
5 instances · 3 analysts
· Δ adherence -0.2
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View Worklist
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 82.0 vs reference 82.2 (delta -0.2). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
3 instances · 2 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Cases
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
2 instances · 2 analysts
· Δ adherence -19.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Microsoft Teams
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Microsoft Teams
2 instances · 2 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Microsoft Teams
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Microsoft Teams during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Acrobat
2 instances · 1 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Acrobat
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Acrobat during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Phobos
2 instances · 2 analysts
· Δ adherence -7.2
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View MRN Investigation
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Phobos during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -19.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Select Validation Results
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -19.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Notepad
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View Case
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Application Mix
Which applications analysts touch while executing this task, measured by event count and active dwell time.
| Application | Events | Event Share | Dwell Share |
|---|---|---|---|
| Phobos | 1,010 | 91.7% | 51.6% |
| Veeva Safety | 53 | 4.8% | 2.5% |
| RIM Vault | 20 | 1.8% | 1.3% |
| Microsoft Teams | 6 | 0.5% | 0.0% |
| Microsoft Excel | 4 | 0.4% | 0.0% |
| Acrobat | 3 | 0.3% | 0.0% |
| Quality Vault | 2 | 0.2% | 44.6% |
| Microsoft Outlook | 2 | 0.2% | 0.0% |
| Notepad | 1 | 0.1% | 0.0% |
| collaboration.merck.com | 1 | 0.1% | 0.0% |
Top Cross-App Transitions (within task instance)
Consecutive events that cross application boundaries within a task instance — the signal where swivel-chair patterns live.
| From | To | Transitions |
|---|---|---|
| Phobos | Veeva Safety | 36 |
| Phobos | Microsoft Teams | 3 |
| Microsoft Excel | Veeva Safety | 3 |
| Veeva Safety | Microsoft Teams | 3 |
| Phobos | Microsoft Excel | 3 |
| Veeva Safety | Phobos | 3 |
| Phobos | Acrobat | 2 |
| Phobos | RIM Vault | 2 |
| Microsoft Teams | Veeva Safety | 1 |
| Phobos | Quality Vault | 1 |
On-Case Application Journey (±15 min around anchor)
Apps the same analyst touched on the same case within ±15 minutes of the task anchor. Captures work that spans task boundaries — the Veeva/Acrobat/Outlook/Word context surrounding the anchor click.
Case counts per app are not mutually exclusive — a case that touched Veeva AND Phobos in this window is counted in both rows. Overlap reflects real cross-app work, not double-counted cases.
Phobos DOM 2,216 events 171 cases 58.7 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
469×
inputld
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
308×
pyzeClick
<button>
"Proceed"
269×
pyzeClick
<button>
"Save Comments"
229×
pyzeClick
<button>
"Push to Case Processing"
133×
pyzeClick
<select>
"Select Compound · - sitagliptin phosphate Tablet (G"
129×
pyzeClick
<button>
"Advance"
97×
pyzeClick
<button>
"Send to process"
97×
Veeva Safety DOM 1,257 events 154 cases 10.9 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Adverse Events"
73×
pyzeClick
<span>
"Complete"
70×
pyzeClick
<a>
"Show more"
59×
pyzeClick
<span>
"Products"
42×
pyzeClick
<li>
"Details"
38×
pyzeClick
<span>
"Narrative"
35×
pyzeClick
<li>
"Narrative"
33×
pyzeClick
<span>
"Case Assessment Results"
33×
RIM Vault DOM 64 events 6 cases 0.4 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Continue"
11×
pyzeClick
<div>
"This report requires you to complete the report v"
6×
pyzeClick
<span>
"Agreement Documents"
4×
pyzeClick
<a>
"Kelun (SKB) MK-2870, MK-2870 (with MK-3475) & MK-"
4×
pyzeClick
<a>
"Agreements by Clinical Study"
2×
pyzeClick
<a>
"Daiichi Sankyo MK-3475 PV Agreement"
2×
pyzeClick
<a>
"Reports"
2×
pyzeClick
<a>
"MK-2870"
2×
collaboration.merck.com App 50 events 18 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
437 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Teams App 47 events 35 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
532 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Acrobat App 36 events 13 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
278 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Excel Document 34 events 20 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Microsoft Outlook Message 34 events 19 cases
Message-level capture
Task Mining captures email subjects. Case numbers frequently appear in the subject line.
usc-excel.officeapps.live.com Document 23 events 5 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Microsoft Word Document 19 events 11 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Adherence Vectors
Each task instance is scored on four behavioral vectors. Lower values are more adherent to the mined SoP.
0.03
Swivel Rate
events in non-primary app / total
0.96
Zero-Edit Visit Rate
inspection-only clicks
1.01
Avg Revisit Depth
max repeats of any activity
0.00
Cross-App Time Share
time outside primary app
Meta Score Distribution
Instance count in each 10-point adherence bucket. Right-skewed = healthy execution; long left tail = concentrated pain points.
0
10
20
30
40
50
60
70
80
90
Top Concern Vector
Zero-Edit Visit Rate — current value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Vector Contribution to Adherence Loss
Each vector's weighted contribution to lost adherence. The top row is your highest-leverage fix.
Zero-Edit Visit Rate
value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Revisit Depth
value 1.01
High revisit depth = mid-flow rework. Trace which sub-step is being repeated; usually a data-readiness gap upstream.
Swivel Rate
value 0.03
Reduce cross-app hops by surfacing the secondary-app data inline (API integration or pre-fetched context panel).
Cross-App Time Share
value 0.00
Time outside primary app dominates this score. Identify which secondary app(s) and whether their function can be in-lined.
Lowest-Adherence Instances
The 5 task instances that scored lowest. These are the concrete cases worth investigating to understand the worst patterns.
| Case | Analyst | Score | Duration | Sub-Steps | Swivel | X-App | Revisit |
|---|---|---|---|---|---|---|---|
| 2355231 | Analyst 13 | 58.5 | 43.4m | 5 | 0.80 | 0.00 | 1 |
| 2346321 | Analyst 13 | 60.0 | 1.3m | 4 | 0.75 | 0.00 | 1 |
| 2347506 | Analyst 12 | 62.5 | 3.2m | 6 | 0.67 | 0.00 | 1 |
| 2350291 | Analyst 02 | 62.5 | 1.3m | 3 | 0.67 | 0.00 | 1 |
| 2320304 | Analyst 09 | 62.5 | 0.1m | 3 | 0.67 | 0.00 | 1 |
Users on This Task
7 analysts executed this task during the pilot. Scatter plots productivity (instances per active day) against adherence (meta score). Top-right = healthy high-volume analysts; bottom-right = firefighters; top-left = careful low-volume; bottom-left = struggling.
Adherence (meta score) →
Productivity (instances / active day) →
05
03
01
12
09
13
02
User-by-User Breakdown
Click any analyst to see their dominant sequence vs the reference SoP.
| Analyst | Instances | Inst / Day | Adherence | Weakest Vector | Swivel | X-App | Min | |
|---|---|---|---|---|---|---|---|---|
| ▸ | Analyst 05 | 84 | 14.0 | 82.5 | Zero-Edit Visits | 0.01 | 0.00 | 145 |
|
Reference SoP
1
View Intake Processing
Analyst 05's Dominant Sequence
— 69 of 84 instances
1
View Intake Processing
This analyst exhibits 5 distinct sequences across their 84 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 03 | 218 | 18.2 | 82.4 | Zero-Edit Visits | 0.05 | 0.00 | 8,051 |
|
Reference SoP
1
View Intake Processing
Analyst 03's Dominant Sequence
— 159 of 218 instances
1
View Intake Processing
This analyst exhibits 16 distinct sequences across their 218 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 01 | 131 | 14.6 | 82.4 | Zero-Edit Visits | 0.02 | 0.00 | 8,945 |
|
Reference SoP
1
View Intake Processing
Analyst 01's Dominant Sequence
— 98 of 131 instances
1
View Intake Processing
This analyst exhibits 8 distinct sequences across their 131 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 12 | 129 | 18.4 | 82.3 | Zero-Edit Visits | 0.01 | 0.00 | 247 |
|
Reference SoP
1
View Intake Processing
Analyst 12's Dominant Sequence
— 113 of 129 instances
1
View Intake Processing
This analyst exhibits 6 distinct sequences across their 129 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 09 | 94 | 18.8 | 82.1 | Zero-Edit Visits | 0.02 | 0.00 | 124 |
|
Reference SoP
1
View Intake Processing
Analyst 09's Dominant Sequence
— 77 of 94 instances
1
View Intake Processing
This analyst exhibits 7 distinct sequences across their 94 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 13 | 138 | 15.3 | 81.9 | Zero-Edit Visits | 0.04 | 0.00 | 289 |
|
Reference SoP
1
View Intake Processing
Analyst 13's Dominant Sequence
— 106 of 138 instances
1
View Intake Processing
This analyst exhibits 13 distinct sequences across their 138 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 02 | 69 | 9.9 | 81.1 | Zero-Edit Visits | 0.05 | 0.00 | 60 |
|
Reference SoP
1
View Intake Processing
Analyst 02's Dominant Sequence
— 56 of 69 instances
1
View Intake Processing
This analyst exhibits 8 distinct sequences across their 69 instances; the dominant one is shown.
|
||||||||
82.2
Avg Adherence Score (0-100)
Median 82.5 · Range 58.5–90.0
20.6 min
Avg Task Duration
0.1 min median
1.3
Avg Sub-Steps per Instance
21.6% have multiple sub-steps
78.4%
Reference Path Coverage
Instances following mined SoP exactly
Reference Baseline
78.4% of instances follow this exactly
Begin Case Processing — Mined SoP
The most common sub-step sequence observed across 863 task instances. Use the dropdown on the right to compare any deviation to this baseline.
1
View Intake Processing
0.0s
Phobos
97%
Variant Explorer
Reference
Begin Case Processing Reference Path
677 instances · 7 analysts
This is the mined reference path — 677 instances across 7 analysts follow this exact sequence. Treat as the proposed Standard Operating Procedure for Begin Case Processing.
Genuine Complexity
Complex-Case Execution Path
84 instances · 7 analysts
· Δ adherence +3.1
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 85.3 vs reference 82.2 (delta +3.1). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Genuine Complexity
Complex-Case Execution Path
19 instances · 7 analysts
· Δ adherence +1.1
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
Enter Inbox ID
reference at this step: View Intake Processing
2
View Intake Processing
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 83.3 vs reference 82.2 (delta +1.1). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Genuine Complexity
Complex-Case Execution Path
19 instances · 4 analysts
· Δ adherence +7.8
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 90.0 vs reference 82.2 (delta +7.8). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Genuine Complexity
Complex-Case Execution Path
10 instances · 3 analysts
· Δ adherence +7.8
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
Select Compound
reference at this step: View Intake Processing
2
View Intake Processing
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 90.0 vs reference 82.2 (delta +7.8). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
7 instances · 2 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Genuine Complexity
Complex-Case Execution Path
5 instances · 3 analysts
· Δ adherence -0.2
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View Worklist
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 82.0 vs reference 82.2 (delta -0.2). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
3 instances · 2 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Cases
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
2 instances · 2 analysts
· Δ adherence -19.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Microsoft Teams
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Acrobat
2 instances · 1 analysts
· Δ adherence -14.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Acrobat
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Acrobat during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Phobos
2 instances · 2 analysts
· Δ adherence -7.2
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View MRN Investigation
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Phobos during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -19.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Select Validation Results
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -19.7
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Notepad
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Application Mix
Which applications analysts touch while executing this task, measured by event count and active dwell time.
| Application | Events | Event Share | Dwell Share |
|---|---|---|---|
| Phobos | 1,010 | 91.5% | 51.6% |
| Veeva Safety | 54 | 4.9% | 2.5% |
| RIM Vault | 20 | 1.8% | 1.3% |
| Microsoft Teams | 6 | 0.5% | 0.0% |
| Microsoft Excel | 4 | 0.4% | 0.0% |
| Acrobat | 3 | 0.3% | 0.0% |
| Quality Vault | 2 | 0.2% | 44.6% |
| Microsoft Outlook | 2 | 0.2% | 0.0% |
| apps.powerapps.com | 1 | 0.1% | 0.0% |
| collaboration.merck.com | 1 | 0.1% | 0.0% |
Top Cross-App Transitions (within task instance)
Consecutive events that cross application boundaries within a task instance — the signal where swivel-chair patterns live.
| From | To | Transitions |
|---|---|---|
| Phobos | Veeva Safety | 37 |
| Veeva Safety | Microsoft Teams | 4 |
| Veeva Safety | Phobos | 3 |
| Microsoft Excel | Veeva Safety | 3 |
| Phobos | Microsoft Excel | 3 |
| Phobos | Microsoft Teams | 2 |
| Phobos | Acrobat | 2 |
| Phobos | RIM Vault | 2 |
| Phobos | Microsoft Outlook | 1 |
| Microsoft Teams | Veeva Safety | 1 |
On-Case Application Journey (±15 min around anchor)
Apps the same analyst touched on the same case within ±15 minutes of the task anchor. Captures work that spans task boundaries — the Veeva/Acrobat/Outlook/Word context surrounding the anchor click.
Case counts per app are not mutually exclusive — a case that touched Veeva AND Phobos in this window is counted in both rows. Overlap reflects real cross-app work, not double-counted cases.
Phobos DOM 2,216 events 171 cases 58.7 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
469×
inputld
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
308×
pyzeClick
<button>
"Proceed"
269×
pyzeClick
<button>
"Save Comments"
229×
pyzeClick
<button>
"Push to Case Processing"
133×
pyzeClick
<select>
"Select Compound · - sitagliptin phosphate Tablet (G"
129×
pyzeClick
<button>
"Advance"
97×
pyzeClick
<button>
"Send to process"
97×
Veeva Safety DOM 1,257 events 154 cases 10.9 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Adverse Events"
73×
pyzeClick
<span>
"Complete"
70×
pyzeClick
<a>
"Show more"
59×
pyzeClick
<span>
"Products"
42×
pyzeClick
<li>
"Details"
38×
pyzeClick
<span>
"Narrative"
35×
pyzeClick
<li>
"Narrative"
33×
pyzeClick
<span>
"Case Assessment Results"
33×
RIM Vault DOM 64 events 6 cases 0.4 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Continue"
11×
pyzeClick
<div>
"This report requires you to complete the report v"
6×
pyzeClick
<span>
"Agreement Documents"
4×
pyzeClick
<a>
"Kelun (SKB) MK-2870, MK-2870 (with MK-3475) & MK-"
4×
pyzeClick
<a>
"Agreements by Clinical Study"
2×
pyzeClick
<a>
"Daiichi Sankyo MK-3475 PV Agreement"
2×
pyzeClick
<a>
"Reports"
2×
pyzeClick
<a>
"MK-2870"
2×
collaboration.merck.com App 50 events 18 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
437 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Teams App 47 events 35 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
532 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Acrobat App 36 events 13 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
278 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Outlook Message 34 events 19 cases
Message-level capture
Task Mining captures email subjects. Case numbers frequently appear in the subject line.
Microsoft Excel Document 34 events 20 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
usc-excel.officeapps.live.com Document 23 events 5 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Microsoft Word Document 19 events 11 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Adherence Vectors
Each task instance is scored on four behavioral vectors. Lower values are more adherent to the mined SoP.
0.03
Swivel Rate
events in non-primary app / total
0.96
Zero-Edit Visit Rate
inspection-only clicks
1.01
Avg Revisit Depth
max repeats of any activity
0.00
Cross-App Time Share
time outside primary app
Meta Score Distribution
Instance count in each 10-point adherence bucket. Right-skewed = healthy execution; long left tail = concentrated pain points.
0
10
20
30
40
50
60
70
80
90
Top Concern Vector
Zero-Edit Visit Rate — current value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Vector Contribution to Adherence Loss
Each vector's weighted contribution to lost adherence. The top row is your highest-leverage fix.
Zero-Edit Visit Rate
value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Revisit Depth
value 1.01
High revisit depth = mid-flow rework. Trace which sub-step is being repeated; usually a data-readiness gap upstream.
Swivel Rate
value 0.03
Reduce cross-app hops by surfacing the secondary-app data inline (API integration or pre-fetched context panel).
Cross-App Time Share
value 0.00
Time outside primary app dominates this score. Identify which secondary app(s) and whether their function can be in-lined.
Lowest-Adherence Instances
The 5 task instances that scored lowest. These are the concrete cases worth investigating to understand the worst patterns.
| Case | Analyst | Score | Duration | Sub-Steps | Swivel | X-App | Revisit |
|---|---|---|---|---|---|---|---|
| 2355231 | Analyst 13 | 58.5 | 43.4m | 5 | 0.80 | 0.00 | 1 |
| 2346321 | Analyst 13 | 60.0 | 1.3m | 4 | 0.75 | 0.00 | 1 |
| 2352741 | Analyst 01 | 62.5 | 1.4m | 3 | 0.67 | 0.00 | 1 |
| 2279519 | Analyst 13 | 62.5 | 0.1m | 3 | 0.67 | 0.00 | 1 |
| 2350291 | Analyst 02 | 62.5 | 1.3m | 3 | 0.67 | 0.00 | 1 |
Users on This Task
7 analysts executed this task during the pilot. Scatter plots productivity (instances per active day) against adherence (meta score). Top-right = healthy high-volume analysts; bottom-right = firefighters; top-left = careful low-volume; bottom-left = struggling.
Adherence (meta score) →
Productivity (instances / active day) →
05
03
01
12
09
13
02
User-by-User Breakdown
Click any analyst to see their dominant sequence vs the reference SoP.
| Analyst | Instances | Inst / Day | Adherence | Weakest Vector | Swivel | X-App | Min | |
|---|---|---|---|---|---|---|---|---|
| ▸ | Analyst 05 | 84 | 14.0 | 82.5 | Zero-Edit Visits | 0.01 | 0.00 | 145 |
|
Reference SoP
1
View Intake Processing
Analyst 05's Dominant Sequence
— 69 of 84 instances
1
View Intake Processing
This analyst exhibits 6 distinct sequences across their 84 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 03 | 218 | 18.2 | 82.4 | Zero-Edit Visits | 0.05 | 0.00 | 8,047 |
|
Reference SoP
1
View Intake Processing
Analyst 03's Dominant Sequence
— 159 of 218 instances
1
View Intake Processing
This analyst exhibits 18 distinct sequences across their 218 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 01 | 131 | 14.6 | 82.4 | Zero-Edit Visits | 0.02 | 0.00 | 8,945 |
|
Reference SoP
1
View Intake Processing
Analyst 01's Dominant Sequence
— 98 of 131 instances
1
View Intake Processing
This analyst exhibits 9 distinct sequences across their 131 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 12 | 129 | 18.4 | 82.3 | Zero-Edit Visits | 0.01 | 0.00 | 247 |
|
Reference SoP
1
View Intake Processing
Analyst 12's Dominant Sequence
— 113 of 129 instances
1
View Intake Processing
This analyst exhibits 7 distinct sequences across their 129 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 09 | 94 | 18.8 | 82.1 | Zero-Edit Visits | 0.02 | 0.00 | 124 |
|
Reference SoP
1
View Intake Processing
Analyst 09's Dominant Sequence
— 77 of 94 instances
1
View Intake Processing
This analyst exhibits 8 distinct sequences across their 94 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 13 | 138 | 15.3 | 81.8 | Zero-Edit Visits | 0.04 | 0.00 | 217 |
|
Reference SoP
1
View Intake Processing
Analyst 13's Dominant Sequence
— 105 of 138 instances
1
View Intake Processing
This analyst exhibits 15 distinct sequences across their 138 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 02 | 69 | 9.9 | 81.1 | Zero-Edit Visits | 0.05 | 0.00 | 60 |
|
Reference SoP
1
View Intake Processing
Analyst 02's Dominant Sequence
— 56 of 69 instances
1
View Intake Processing
This analyst exhibits 8 distinct sequences across their 69 instances; the dominant one is shown.
|
||||||||
81.0
Avg Adherence Score (0-100)
Median 82.5 · Range 58.5–90.0
27.8 min
Avg Task Duration
0.1 min median
1.7
Avg Sub-Steps per Instance
37.1% have multiple sub-steps
72.6%
Reference Path Coverage
Instances following mined SoP exactly
Reference Baseline
72.6% of instances follow this exactly
Begin Case Processing — Mined SoP
The most common sub-step sequence observed across 642 task instances. Use the dropdown on the right to compare any deviation to this baseline.
1
View Intake Processing
0.0s
Phobos
99%
Variant Explorer
Reference
Begin Case Processing Reference Path
466 instances · 7 analysts
This is the mined reference path — 466 instances across 7 analysts follow this exact sequence. Treat as the proposed Standard Operating Procedure for Begin Case Processing.
Reviewer Sweep
Reviewer Verification Pass
75 instances · 7 analysts
· Δ adherence -0.1
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
Reviewer Sweep — a viewing-heavy pattern with 3.3 avg sub-steps and zero edits captured. Consistent with a downstream reviewer re-opening the task to verify earlier work rather than progressing new data. 7 analysts across 59 cases.
Genuine Complexity
Complex-Case Execution Path
19 instances · 6 analysts
· Δ adherence -0.1
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 80.9 vs reference 81.0 (delta -0.1). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Reviewer Sweep
Reviewer Verification Pass
15 instances · 3 analysts
· Δ adherence +3.2
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
Reviewer Sweep — a viewing-heavy pattern with 3.2 avg sub-steps and zero edits captured. Consistent with a downstream reviewer re-opening the task to verify earlier work rather than progressing new data. 3 analysts across 12 cases.
Genuine Complexity
Complex-Case Execution Path
11 instances · 3 analysts
· Δ adherence +5.8
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 86.8 vs reference 81.0 (delta +5.8). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
7 instances · 2 analysts
· Δ adherence -13.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Genuine Complexity
Complex-Case Execution Path
5 instances · 3 analysts
· Δ adherence +1.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View Worklist
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 82.0 vs reference 81.0 (delta +1.0). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Backfill Loop
Mid-Flow Data Backfill
2 instances · 2 analysts
· Δ adherence -0.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
4
Enter Inbox ID
extra step (reference ends at step 1)
5
View Intake Processing
extra step (reference ends at step 1)
Backfill Loop — instances revisit the same activity 3.0× on average, consistent with mid-flow detours to fetch or enter data that wasn't ready at the right step. Ordering issue, not an app issue — 2 analysts across 2 cases. Suggests upstream data readiness or UX sequencing is at fault.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
2 instances · 2 analysts
· Δ adherence -13.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Cases
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Phobos
2 instances · 2 analysts
· Δ adherence -6.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View MRN Investigation
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Phobos during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -13.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
4
Navigate to Inbox
extra step (reference ends at step 1)
5
Select Validation Results
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 40% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -18.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Notepad
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -13.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
4
Navigate to Inbox
extra step (reference ends at step 1)
5
Microsoft Teams
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 40% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Application Mix
Which applications analysts touch while executing this task, measured by event count and active dwell time.
| Application | Events | Event Share | Dwell Share |
|---|---|---|---|
| Phobos | 1,010 | 91.7% | 51.6% |
| Veeva Safety | 53 | 4.8% | 2.5% |
| RIM Vault | 20 | 1.8% | 1.3% |
| Microsoft Teams | 6 | 0.5% | 0.0% |
| Microsoft Excel | 4 | 0.4% | 0.0% |
| Acrobat | 3 | 0.3% | 0.0% |
| Quality Vault | 2 | 0.2% | 44.6% |
| Microsoft Outlook | 2 | 0.2% | 0.0% |
| collaboration.merck.com | 1 | 0.1% | 0.0% |
| Notepad | 1 | 0.1% | 0.0% |
Top Cross-App Transitions (within task instance)
Consecutive events that cross application boundaries within a task instance — the signal where swivel-chair patterns live.
| From | To | Transitions |
|---|---|---|
| Phobos | Veeva Safety | 37 |
| Veeva Safety | Microsoft Teams | 4 |
| Veeva Safety | Phobos | 4 |
| Phobos | Microsoft Excel | 3 |
| Microsoft Excel | Veeva Safety | 3 |
| Phobos | RIM Vault | 2 |
| Phobos | Microsoft Teams | 2 |
| Phobos | Acrobat | 2 |
| Veeva Safety | collaboration.merck.com | 1 |
| Microsoft Outlook | Veeva Safety | 1 |
On-Case Application Journey (±15 min around anchor)
Apps the same analyst touched on the same case within ±15 minutes of the task anchor. Captures work that spans task boundaries — the Veeva/Acrobat/Outlook/Word context surrounding the anchor click.
Case counts per app are not mutually exclusive — a case that touched Veeva AND Phobos in this window is counted in both rows. Overlap reflects real cross-app work, not double-counted cases.
Phobos DOM 2,216 events 171 cases 58.7 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
469×
inputld
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
308×
pyzeClick
<button>
"Proceed"
269×
pyzeClick
<button>
"Save Comments"
229×
pyzeClick
<button>
"Push to Case Processing"
133×
pyzeClick
<select>
"Select Compound · - sitagliptin phosphate Tablet (G"
129×
pyzeClick
<button>
"Advance"
97×
pyzeClick
<button>
"Send to process"
97×
Veeva Safety DOM 1,257 events 154 cases 10.9 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Adverse Events"
73×
pyzeClick
<span>
"Complete"
70×
pyzeClick
<a>
"Show more"
59×
pyzeClick
<span>
"Products"
42×
pyzeClick
<li>
"Details"
38×
pyzeClick
<span>
"Narrative"
35×
pyzeClick
<li>
"Narrative"
33×
pyzeClick
<span>
"Case Assessment Results"
33×
RIM Vault DOM 64 events 6 cases 0.4 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Continue"
11×
pyzeClick
<div>
"This report requires you to complete the report v"
6×
pyzeClick
<span>
"Agreement Documents"
4×
pyzeClick
<a>
"Kelun (SKB) MK-2870, MK-2870 (with MK-3475) & MK-"
4×
pyzeClick
<div>
"Products"
2×
pyzeClick
<a>
"Reports"
2×
pyzeClick
<a>
"Consortium Agreement NCI MK-3475"
2×
pyzeClick
<a>
"Kelun"
2×
collaboration.merck.com App 50 events 18 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
437 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Teams App 47 events 35 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
532 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Acrobat App 36 events 13 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
278 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Excel Document 34 events 20 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Microsoft Outlook Message 34 events 19 cases
Message-level capture
Task Mining captures email subjects. Case numbers frequently appear in the subject line.
usc-excel.officeapps.live.com Document 23 events 5 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Microsoft Word Document 19 events 11 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Adherence Vectors
Each task instance is scored on four behavioral vectors. Lower values are more adherent to the mined SoP.
0.04
Swivel Rate
events in non-primary app / total
0.96
Zero-Edit Visit Rate
inspection-only clicks
1.36
Avg Revisit Depth
max repeats of any activity
0.00
Cross-App Time Share
time outside primary app
Meta Score Distribution
Instance count in each 10-point adherence bucket. Right-skewed = healthy execution; long left tail = concentrated pain points.
0
10
20
30
40
50
60
70
80
90
Top Concern Vector
Zero-Edit Visit Rate — current value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Vector Contribution to Adherence Loss
Each vector's weighted contribution to lost adherence. The top row is your highest-leverage fix.
Zero-Edit Visit Rate
value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Revisit Depth
value 1.36
High revisit depth = mid-flow rework. Trace which sub-step is being repeated; usually a data-readiness gap upstream.
Swivel Rate
value 0.04
Reduce cross-app hops by surfacing the secondary-app data inline (API integration or pre-fetched context panel).
Cross-App Time Share
value 0.00
Time outside primary app dominates this score. Identify which secondary app(s) and whether their function can be in-lined.
Lowest-Adherence Instances
The 5 task instances that scored lowest. These are the concrete cases worth investigating to understand the worst patterns.
| Case | Analyst | Score | Duration | Sub-Steps | Swivel | X-App | Revisit |
|---|---|---|---|---|---|---|---|
| 2355231 | Analyst 13 | 58.5 | 43.4m | 5 | 0.80 | 0.00 | 1 |
| 2346321 | Analyst 13 | 60.0 | 1.3m | 4 | 0.75 | 0.00 | 1 |
| 2350275 | Analyst 05 | 61.1 | 6.3m | 4 | 0.75 | 0.01 | 2 |
| 2350291 | Analyst 02 | 62.5 | 1.3m | 3 | 0.67 | 0.00 | 1 |
| 2279519 | Analyst 13 | 62.5 | 0.1m | 3 | 0.67 | 0.00 | 1 |
Users on This Task
7 analysts executed this task during the pilot. Scatter plots productivity (instances per active day) against adherence (meta score). Top-right = healthy high-volume analysts; bottom-right = firefighters; top-left = careful low-volume; bottom-left = struggling.
Adherence (meta score) →
Productivity (instances / active day) →
03
09
05
12
01
13
02
User-by-User Breakdown
Click any analyst to see their dominant sequence vs the reference SoP.
| Analyst | Instances | Inst / Day | Adherence | Weakest Vector | Swivel | X-App | Min | |
|---|---|---|---|---|---|---|---|---|
| ▸ | Analyst 03 | 170 | 14.2 | 81.3 | Zero-Edit Visits | 0.05 | 0.00 | 8,052 |
|
Reference SoP
1
View Intake Processing
Analyst 03's Dominant Sequence
— 113 of 170 instances
1
View Intake Processing
This analyst exhibits 19 distinct sequences across their 170 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 09 | 74 | 14.8 | 81.3 | Zero-Edit Visits | 0.03 | 0.00 | 124 |
|
Reference SoP
1
View Intake Processing
Analyst 09's Dominant Sequence
— 57 of 74 instances
1
View Intake Processing
This analyst exhibits 8 distinct sequences across their 74 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 05 | 62 | 10.3 | 81.3 | Zero-Edit Visits | 0.01 | 0.00 | 145 |
|
Reference SoP
1
View Intake Processing
Analyst 05's Dominant Sequence
— 48 of 62 instances
1
View Intake Processing
This analyst exhibits 6 distinct sequences across their 62 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 12 | 96 | 13.7 | 81.2 | Zero-Edit Visits | 0.02 | 0.00 | 248 |
|
Reference SoP
1
View Intake Processing
Analyst 12's Dominant Sequence
— 80 of 96 instances
1
View Intake Processing
This analyst exhibits 6 distinct sequences across their 96 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 01 | 91 | 10.1 | 81.1 | Zero-Edit Visits | 0.03 | 0.00 | 8,947 |
|
Reference SoP
1
View Intake Processing
Analyst 01's Dominant Sequence
— 58 of 91 instances
1
View Intake Processing
This analyst exhibits 10 distinct sequences across their 91 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 13 | 97 | 10.8 | 80.5 | Zero-Edit Visits | 0.05 | 0.00 | 291 |
|
Reference SoP
1
View Intake Processing
Analyst 13's Dominant Sequence
— 69 of 97 instances
1
View Intake Processing
This analyst exhibits 17 distinct sequences across their 97 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 02 | 52 | 7.4 | 80.2 | Zero-Edit Visits | 0.05 | 0.00 | 61 |
|
Reference SoP
1
View Intake Processing
Analyst 02's Dominant Sequence
— 41 of 52 instances
1
View Intake Processing
This analyst exhibits 9 distinct sequences across their 52 instances; the dominant one is shown.
|
||||||||
81.0
Avg Adherence Score (0-100)
Median 82.5 · Range 52.3–90.0
27.8 min
Avg Task Duration
0.1 min median
1.7
Avg Sub-Steps per Instance
37.2% have multiple sub-steps
72.3%
Reference Path Coverage
Instances following mined SoP exactly
Reference Baseline
72.3% of instances follow this exactly
Begin Case Processing — Mined SoP
The most common sub-step sequence observed across 640 task instances. Use the dropdown on the right to compare any deviation to this baseline.
1
View Intake Processing
0.0s
Phobos
99%
Variant Explorer
Reference
Begin Case Processing Reference Path
463 instances · 7 analysts
This is the mined reference path — 463 instances across 7 analysts follow this exact sequence. Treat as the proposed Standard Operating Procedure for Begin Case Processing.
Reviewer Sweep
Reviewer Verification Pass
75 instances · 7 analysts
· Δ adherence -0.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
Reviewer Sweep — a viewing-heavy pattern with 3.3 avg sub-steps and zero edits captured. Consistent with a downstream reviewer re-opening the task to verify earlier work rather than progressing new data. 7 analysts across 59 cases.
Genuine Complexity
Complex-Case Execution Path
19 instances · 6 analysts
· Δ adherence -0.1
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 80.9 vs reference 81.0 (delta -0.1). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Reviewer Sweep
Reviewer Verification Pass
15 instances · 3 analysts
· Δ adherence +3.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
Reviewer Sweep — a viewing-heavy pattern with 3.3 avg sub-steps and zero edits captured. Consistent with a downstream reviewer re-opening the task to verify earlier work rather than progressing new data. 3 analysts across 12 cases.
Genuine Complexity
Complex-Case Execution Path
11 instances · 3 analysts
· Δ adherence +5.8
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 86.8 vs reference 81.0 (delta +5.8). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
7 instances · 2 analysts
· Δ adherence -13.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Linked Remediation Opportunities
Genuine Complexity
Complex-Case Execution Path
5 instances · 3 analysts
· Δ adherence +1.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View Worklist
extra step (reference ends at step 1)
Genuine Complexity — this sequence differs from the reference path (2 sub-steps vs 1 in reference) but shows no strong pain-point signals: adherence score 82.0 vs reference 81.0 (delta +1.0). Likely a legitimate case-complexity path — not an automation candidate, but worth cataloging.
Backfill Loop
Mid-Flow Data Backfill
2 instances · 2 analysts
· Δ adherence -0.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Select Compound
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
4
Enter Inbox ID
extra step (reference ends at step 1)
5
View Intake Processing
extra step (reference ends at step 1)
Backfill Loop — instances revisit the same activity 3.0× on average, consistent with mid-flow detours to fetch or enter data that wasn't ready at the right step. Ordering issue, not an app issue — 2 analysts across 2 cases. Suggests upstream data readiness or UX sequencing is at fault.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
2 instances · 2 analysts
· Δ adherence -13.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Cases
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Linked Remediation Opportunities
Cross-App Swivel
Cross-App Lookup to Phobos
2 instances · 2 analysts
· Δ adherence -6.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
View MRN Investigation
extra step (reference ends at step 1)
Cross-App Swivel — 2 analysts pivot into Phobos during this task (0% of task time in non-primary apps, 50% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -13.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
4
Navigate to Inbox
extra step (reference ends at step 1)
5
Select Validation Results
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 40% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Linked Remediation Opportunities
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -18.5
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Navigate to Inbox
extra step (reference ends at step 1)
3
Notepad
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 67% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Linked Remediation Opportunities
Cross-App Swivel
Cross-App Lookup to Veeva Safety
1 instances · 1 analysts
· Δ adherence -13.0
Observed Sequence (vs Reference SoP on left)
differs from reference
matches reference
1
View Intake Processing
2
Enter Inbox ID
extra step (reference ends at step 1)
3
View Intake Processing
extra step (reference ends at step 1)
4
Navigate to Inbox
extra step (reference ends at step 1)
5
Microsoft Teams
extra step (reference ends at step 1)
Cross-App Swivel — 1 analysts pivot into Veeva Safety during this task (0% of task time in non-primary apps, 40% of events outside the primary system). Pattern is consistent with data lookup or cross-referencing rather than a system defect — lighter-weight than a Silent Workaround.
Linked Remediation Opportunities
Application Mix
Which applications analysts touch while executing this task, measured by event count and active dwell time.
| Application | Events | Event Share | Dwell Share |
|---|---|---|---|
| Phobos | 1,010 | 91.5% | 51.6% |
| Veeva Safety | 54 | 4.9% | 2.5% |
| RIM Vault | 20 | 1.8% | 1.3% |
| Microsoft Teams | 6 | 0.5% | 0.0% |
| Microsoft Excel | 4 | 0.4% | 0.0% |
| Acrobat | 3 | 0.3% | 0.0% |
| Quality Vault | 2 | 0.2% | 44.6% |
| Microsoft Outlook | 2 | 0.2% | 0.0% |
| Notepad | 1 | 0.1% | 0.0% |
| collaboration.merck.com | 1 | 0.1% | 0.0% |
Top Cross-App Transitions (within task instance)
Consecutive events that cross application boundaries within a task instance — the signal where swivel-chair patterns live.
| From | To | Transitions |
|---|---|---|
| Phobos | Veeva Safety | 37 |
| Veeva Safety | Phobos | 4 |
| Veeva Safety | Microsoft Teams | 4 |
| Phobos | Microsoft Excel | 3 |
| Microsoft Excel | Veeva Safety | 3 |
| Phobos | Microsoft Teams | 2 |
| Phobos | Acrobat | 2 |
| Phobos | RIM Vault | 2 |
| Phobos | Microsoft Outlook | 1 |
| Quality Vault | Phobos | 1 |
On-Case Application Journey (±15 min around anchor)
Apps the same analyst touched on the same case within ±15 minutes of the task anchor. Captures work that spans task boundaries — the Veeva/Acrobat/Outlook/Word context surrounding the anchor click.
Case counts per app are not mutually exclusive — a case that touched Veeva AND Phobos in this window is counted in both rows. Overlap reflects real cross-app work, not double-counted cases.
Phobos DOM 2,216 events 171 cases 58.7 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
469×
inputld
<select>
"Select Attachment description · Legal · Legal Pre-Sca"
308×
pyzeClick
<button>
"Proceed"
269×
pyzeClick
<button>
"Save Comments"
229×
pyzeClick
<button>
"Push to Case Processing"
133×
pyzeClick
<select>
"Select Compound · - sitagliptin phosphate Tablet (G"
129×
pyzeClick
<button>
"Advance"
97×
pyzeClick
<button>
"Send to process"
97×
Veeva Safety DOM 1,257 events 154 cases 10.9 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Adverse Events"
73×
pyzeClick
<span>
"Complete"
70×
pyzeClick
<a>
"Show more"
59×
pyzeClick
<span>
"Products"
42×
pyzeClick
<li>
"Details"
38×
pyzeClick
<span>
"Narrative"
35×
pyzeClick
<li>
"Narrative"
33×
pyzeClick
<span>
"Case Assessment Results"
33×
RIM Vault DOM 64 events 6 cases 0.4 min
DOM-level capture
Pyze JS captures every element interaction — button labels, field IDs, section panels.
pyzeClick
<span>
"Continue"
11×
pyzeClick
<div>
"This report requires you to complete the report v"
6×
pyzeClick
<span>
"Agreement Documents"
4×
pyzeClick
<a>
"Kelun (SKB) MK-2870, MK-2870 (with MK-3475) & MK-"
4×
pyzeClick
<a>
"Agreements by Clinical Study"
2×
pyzeClick
<a>
"Daiichi Sankyo MK-3475 PV Agreement"
2×
pyzeClick
<a>
"Reports"
2×
pyzeClick
<a>
"MK-2870"
2×
collaboration.merck.com App 50 events 18 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
437 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Teams App 47 events 35 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
532 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Acrobat App 36 events 13 cases
App-level capture
Only application name and dwell time available. Direct instrumentation would unlock deeper detail.
278 granular events captured, but no DOM, document, or message-level metadata is available for this application. Instrumenting this app directly with Pyze would unlock element-level detail.
Microsoft Outlook Message 34 events 19 cases
Message-level capture
Task Mining captures email subjects. Case numbers frequently appear in the subject line.
Microsoft Excel Document 34 events 20 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
usc-excel.officeapps.live.com Document 23 events 5 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Microsoft Word Document 19 events 11 cases
Document-level capture
Task Mining captures the specific file open during work — file names often encode case context.
Adherence Vectors
Each task instance is scored on four behavioral vectors. Lower values are more adherent to the mined SoP.
0.04
Swivel Rate
events in non-primary app / total
0.96
Zero-Edit Visit Rate
inspection-only clicks
1.36
Avg Revisit Depth
max repeats of any activity
0.00
Cross-App Time Share
time outside primary app
Meta Score Distribution
Instance count in each 10-point adherence bucket. Right-skewed = healthy execution; long left tail = concentrated pain points.
0
10
20
30
40
50
60
70
80
90
Top Concern Vector
Zero-Edit Visit Rate — current value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Vector Contribution to Adherence Loss
Each vector's weighted contribution to lost adherence. The top row is your highest-leverage fix.
Zero-Edit Visit Rate
value 0.96
Investigate why analysts inspect-and-leave without editing — usually missing upstream data or unclear UI affordances.
Revisit Depth
value 1.36
High revisit depth = mid-flow rework. Trace which sub-step is being repeated; usually a data-readiness gap upstream.
Swivel Rate
value 0.04
Reduce cross-app hops by surfacing the secondary-app data inline (API integration or pre-fetched context panel).
Cross-App Time Share
value 0.00
Time outside primary app dominates this score. Identify which secondary app(s) and whether their function can be in-lined.
Lowest-Adherence Instances
The 5 task instances that scored lowest. These are the concrete cases worth investigating to understand the worst patterns.
| Case | Analyst | Score | Duration | Sub-Steps | Swivel | X-App | Revisit |
|---|---|---|---|---|---|---|---|
| 2350275 | Analyst 05 | 52.3 | 8.2m | 6 | 0.83 | 0.01 | 4 |
| 2355231 | Analyst 13 | 58.5 | 43.4m | 5 | 0.80 | 0.00 | 1 |
| 2346321 | Analyst 13 | 60.0 | 1.3m | 4 | 0.75 | 0.00 | 1 |
| 2347506 | Analyst 12 | 62.5 | 3.2m | 6 | 0.67 | 0.00 | 1 |
| 2320304 | Analyst 09 | 62.5 | 0.1m | 3 | 0.67 | 0.00 | 1 |
Users on This Task
7 analysts executed this task during the pilot. Scatter plots productivity (instances per active day) against adherence (meta score). Top-right = healthy high-volume analysts; bottom-right = firefighters; top-left = careful low-volume; bottom-left = struggling.
Adherence (meta score) →
Productivity (instances / active day) →
09
03
12
05
01
13
02
User-by-User Breakdown
Click any analyst to see their dominant sequence vs the reference SoP.
| Analyst | Instances | Inst / Day | Adherence | Weakest Vector | Swivel | X-App | Min | |
|---|---|---|---|---|---|---|---|---|
| ▸ | Analyst 09 | 74 | 14.8 | 81.3 | Zero-Edit Visits | 0.03 | 0.00 | 124 |
|
Reference SoP
1
View Intake Processing
Analyst 09's Dominant Sequence
— 57 of 74 instances
1
View Intake Processing
This analyst exhibits 8 distinct sequences across their 74 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 03 | 169 | 14.1 | 81.2 | Zero-Edit Visits | 0.06 | 0.00 | 8,048 |
|
Reference SoP
1
View Intake Processing
Analyst 03's Dominant Sequence
— 112 of 169 instances
1
View Intake Processing
This analyst exhibits 19 distinct sequences across their 169 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 12 | 96 | 13.7 | 81.2 | Zero-Edit Visits | 0.02 | 0.00 | 248 |
|
Reference SoP
1
View Intake Processing
Analyst 12's Dominant Sequence
— 80 of 96 instances
1
View Intake Processing
This analyst exhibits 6 distinct sequences across their 96 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 05 | 61 | 10.2 | 81.2 | Zero-Edit Visits | 0.01 | 0.00 | 146 |
|
Reference SoP
1
View Intake Processing
Analyst 05's Dominant Sequence
— 47 of 61 instances
1
View Intake Processing
This analyst exhibits 6 distinct sequences across their 61 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 01 | 91 | 10.1 | 81.1 | Zero-Edit Visits | 0.03 | 0.00 | 8,947 |
|
Reference SoP
1
View Intake Processing
Analyst 01's Dominant Sequence
— 58 of 91 instances
1
View Intake Processing
This analyst exhibits 10 distinct sequences across their 91 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 13 | 97 | 10.8 | 80.3 | Zero-Edit Visits | 0.05 | 0.00 | 219 |
|
Reference SoP
1
View Intake Processing
Analyst 13's Dominant Sequence
— 68 of 97 instances
1
View Intake Processing
This analyst exhibits 18 distinct sequences across their 97 instances; the dominant one is shown.
|
||||||||
| ▸ | Analyst 02 | 52 | 7.4 | 80.2 | Zero-Edit Visits | 0.05 | 0.00 | 61 |
|
Reference SoP
1
View Intake Processing
Analyst 02's Dominant Sequence
— 41 of 52 instances
1
View Intake Processing
This analyst exhibits 9 distinct sequences across their 52 instances; the dominant one is shown.
|
||||||||