Overview
Row
Who are the heavy GitHub Copilot users, and where are they?
Synthetic demo · Real export schema 10 May 2026 to 04 Jul 2026 · 360 developers · 6 teams · 8 complete Person Query weeks
37 developers (13.8% of the 268 with complete GitHub observation) are heavy GitHub Copilot users. Use this lens, then the four pillars, to explore how their working conditions compare.
Synthetic examples, not productivity scores or intervention effects. The heavy-user share uses all developers with complete GitHub observation as its denominator, not just those with recorded use. The percentile threshold is relative to recorded users; including ties can make that group’s heavy-user share larger than the nominal cut. Microsoft 365 complete-window rates are unavailable without independent collection evidence; missing observation is not non-use.
Row
How the heavy GitHub-use group is defined, and where it concentrates
Definition. Heavy GitHub use: volume at or above the
80 percentile among GitHub-observed developers with recorded use. Volume
combines accepted code completions and user-initiated chat requests over
the 8-week baseline; these are different interactions, not equivalent
units of value. All ties at the threshold are included, so the heavy
group can exceed 20% of recorded users (all recorded users when every
volume is equal). The headline denominator is the complete
GitHub-observed population, not just developers with recorded use. The
268 GitHub-observed developers split into three groups: Heavy
GitHub use (37), Other recorded use (147) and
No recorded use (84). Developers with unresolved GitHub
observation are excluded, never recoded as non-users. These labels are
deliberately distinct from the canonical
vivainsights::identify_usage_segments() names.
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool.
Team and role concentration withheld: a cell, complement or intersection recoverable from linked breakdowns falls below the privacy floor. The overall intensity comparison is assessed separately.
Heavy-group composition by team and role
Unavailable: independent coverage missing or privacy threshold not met.
Unavailable: independent coverage missing or privacy threshold not met.
Comparisons on the pillar pages are descriptive and unadjusted. Groups may differ in team and role mix; any observed differences may reflect that composition rather than an effect of the tool.
Row
Where should we look more closely at developer working conditions?
Baseline percentiles and how to read the working week
| Metric | Unit | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Collaboration hours | Hours / person / week | 11.7 | 15.0 | 18.1 |
| Meetings | Hours / person / week | 5.7 | 8.2 | 10.6 |
| Hours / person / week | 1.8 | 2.2 | 2.9 | |
| Chat | Hours / person / week | 1.3 | 1.6 | 2.1 |
| Available-to-focus hours | Hours / person / week | 28.0 | 30.5 | 33.1 |
| Uninterrupted time | Hours / person / week | 8.8 | 12.1 | 14.7 |
| Interrupted time | Hours / person / week | 16.9 | 18.6 | 20.6 |
| After-hours collaboration | Hours / person / week | 0.9 | 1.6 | 3.2 |
| Internal network size | People (distinct) | 32.5 | 42.2 | 58.5 |
| External network size | People (distinct) | 8.2 | 10.2 | 12.3 |
Show remaining 3 rows (13 total)
| Metric | Unit | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Strong ties | People (distinct) | 9.6 | 12.7 | 17.8 |
| Diverse ties | People (distinct) | 7.1 | 9.2 | 13.4 |
| Network outside organisation | People (distinct) | 4.8 | 6.5 | 8.9 |
Reading the working week. Collaboration hours include meeting hours, scheduled calls, email and chat activity, so they must not be added to meeting hours. Available-to-focus hours are the hours remaining during working hours after excluding meetings and scheduled Teams calls. Uninterrupted hours and interrupted hours partition available-to-focus hours. The working-hours basis is configurable per tenant through metric rules, so the 40-hour basis in this synthetic file is not universal. Available-to-focus hours are calendar availability; they do not measure coding time or psychological flow.
The roster contains 480 synthetic people. The
IsDeveloper attribute selects developers, not their product
activity. Product feeds cover weekdays in 06 Apr 2026 to 03 Jul 2026, a
shorter window than the Person Query history.
See source and observation
contracts.
Collaboration
Row
How does collaboration load differ by GitHub intensity?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 268 GitHub-observed developers · Person-level eight-week averages
Collaboration load compared across Heavy GitHub use (37), Other recorded use (147) and No recorded use (84).
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool. Total collaboration overlaps its components: compare the panels, do not add them. These are descriptive, unadjusted comparisons, not delivery-quality measures or an effect of tool use.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Collaboration hours
Meetings
Chat
Scheduled calls
Unscheduled calls
10 May 2026 to 04 Jul 2026 | 268 distinct developers | 8 complete Person Query weeks | hours/person/week. Dates use UTC convention.
Exact GitHub-intensity group sizes and percentiles
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Heavy GitHub use | Collaboration hours | 37 | 12.1 | 15.4 | 17.7 |
| Other recorded use | Collaboration hours | 147 | 11.5 | 14.9 | 17.9 |
| No recorded use | Collaboration hours | 84 | 12.0 | 15.0 | 19.6 |
| Heavy GitHub use | Meetings | 37 | 5.9 | 8.4 | 9.9 |
| Other recorded use | Meetings | 147 | 6.0 | 8.4 | 10.3 |
| No recorded use | Meetings | 84 | 5.6 | 7.4 | 10.6 |
| Heavy GitHub use | 37 | 1.9 | 2.2 | 2.6 | |
| Other recorded use | 147 | 1.8 | 2.2 | 2.9 | |
| No recorded use | 84 | 1.9 | 2.3 | 2.8 | |
| Heavy GitHub use | Chat | 37 | 1.4 | 1.6 | 1.9 |
Show remaining 8 rows (18 total)
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Other recorded use | Chat | 147 | 1.3 | 1.6 | 2.1 |
| No recorded use | Chat | 84 | 1.4 | 1.6 | 2.2 |
| Heavy GitHub use | Scheduled calls | 37 | 1.1 | 1.3 | 1.4 |
| Other recorded use | Scheduled calls | 147 | 1.0 | 1.3 | 1.5 |
| No recorded use | Scheduled calls | 84 | 1.1 | 1.3 | 1.6 |
| Heavy GitHub use | Unscheduled calls | 37 | 1.1 | 1.2 | 1.3 |
| Other recorded use | Unscheduled calls | 147 | 1.0 | 1.1 | 1.3 |
| No recorded use | Unscheduled calls | 84 | 1.0 | 1.1 | 1.3 |
Row
How is collaboration load distributed across teams?
Platform has the highest median collaboration load at 22.2 hours/person/week. The team view remains because coordination load is a team-level property that the GitHub split does not replace.
Total collaboration overlaps its components: compare the panels, do not add them. These are working conditions, not delivery-quality measures.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Collaboration hours
Meetings
Chat
Scheduled calls
Unscheduled calls
10 May 2026 to 04 Jul 2026 | 360 distinct developers | 8 complete Person Query weeks | hours/person/week. Dates use UTC convention.
Exact team medians
| Rank | Team | Developers (count) | Collaboration hours | Meeting hours | Email hours | Chat hours | Scheduled call hours | Unscheduled call hours |
|---|---|---|---|---|---|---|---|---|
| 1 | Platform | 60 | 22.2 | 12.0 | 3.7 | 2.7 | 1.9 | 1.2 |
| 2 | Payments | 60 | 17.1 | 8.9 | 2.9 | 2.1 | 1.5 | 1.1 |
| 3 | Data Engineering | 60 | 15.5 | 8.7 | 2.3 | 1.6 | 1.3 | 1.1 |
| 4 | Identity | 60 | 14.5 | 7.6 | 2.1 | 1.5 | 1.2 | 1.0 |
| 5 | Developer Experience | 60 | 13.1 | 6.8 | 1.9 | 1.5 | 1.2 | 1.2 |
| 6 | Mobile | 60 | 9.5 | 4.5 | 1.4 | 1.1 | 0.8 | 1.1 |
Team and role composition context
Roster composition provides context for coordination needs. Groups are ranked by the share of Application engineering.
| Group | Category | Developers (count) | Share (%) |
|---|---|---|---|
| Data Engineering | Application engineering | 30 | 50.0% |
| Data Engineering | Engineering management | 12 | 20.0% |
| Data Engineering | Infrastructure engineering | 18 | 30.0% |
| Developer Experience | Application engineering | 26 | 43.3% |
| Developer Experience | Engineering management | 12 | 20.0% |
| Developer Experience | Infrastructure engineering | 22 | 36.7% |
| Identity | Application engineering | 28 | 46.7% |
| Identity | Engineering management | 13 | 21.7% |
| Identity | Infrastructure engineering | 19 | 31.7% |
| Mobile | Application engineering | 32 | 53.3% |
Show remaining 8 rows (18 total)
| Group | Category | Developers (count) | Share (%) |
|---|---|---|---|
| Mobile | Engineering management | 12 | 20.0% |
| Mobile | Infrastructure engineering | 16 | 26.7% |
| Payments | Application engineering | 27 | 45.0% |
| Payments | Engineering management | 13 | 21.7% |
| Payments | Infrastructure engineering | 20 | 33.3% |
| Platform | Application engineering | 24 | 40.0% |
| Platform | Engineering management | 12 | 20.0% |
| Platform | Infrastructure engineering | 24 | 40.0% |
| Attribute | Category | Developers (count) | Share of developers (%) |
|---|---|---|---|
| Role | Application engineering | 167 | 46.4% |
| Role | Engineering management | 74 | 20.6% |
| Role | Infrastructure engineering | 119 | 33.1% |
| Seniority | Early career | 96 | 26.7% |
| Seniority | Experienced | 162 | 45.0% |
| Seniority | Senior / lead | 102 | 28.3% |
| Tenure | 2 to 5 years | 155 | 43.1% |
| Tenure | Over 5 years | 94 | 26.1% |
| Tenure | Under 2 years | 111 | 30.8% |
Meeting characteristics: recurring, conflicting and short notice
| Team | Meeting characteristic | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Payments | Conflicting meetings | 1.4 | 1.7 | 2.1 |
| Platform | Conflicting meetings | 1.3 | 1.5 | 1.7 |
| Data Engineering | Conflicting meetings | 0.5 | 0.9 | 1.1 |
| Identity | Conflicting meetings | 0.5 | 0.7 | 1.0 |
| Developer Experience | Conflicting meetings | 0.4 | 0.6 | 0.8 |
| Mobile | Conflicting meetings | 0.2 | 0.3 | 0.5 |
| Platform | Recurring meetings | 6.8 | 7.8 | 9.2 |
| Payments | Recurring meetings | 4.0 | 5.3 | 6.2 |
| Data Engineering | Recurring meetings | 3.5 | 4.8 | 6.0 |
| Identity | Recurring meetings | 2.8 | 4.1 | 5.4 |
Show remaining 8 rows (18 total)
| Team | Meeting characteristic | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Developer Experience | Recurring meetings | 2.6 | 3.4 | 4.9 |
| Mobile | Recurring meetings | 1.4 | 2.2 | 3.0 |
| Payments | Short-notice meetings | 1.7 | 2.2 | 2.6 |
| Platform | Short-notice meetings | 1.5 | 1.8 | 2.2 |
| Identity | Short-notice meetings | 0.7 | 1.0 | 1.2 |
| Data Engineering | Short-notice meetings | 0.8 | 0.9 | 1.3 |
| Developer Experience | Short-notice meetings | 0.6 | 0.8 | 1.1 |
| Mobile | Short-notice meetings | 0.3 | 0.5 | 0.6 |
Recurring meeting hours cover meetings set to recur. Conflicting hours count only the overlapping portion of calendar meetings. Short-notice meetings were scheduled six hours or less before their start. These are overlapping characteristics, not an additive stack.
Focus
Row
How does focus time differ by GitHub intensity?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 268 GitHub-observed developers · 8 complete Person Query weeks each
Calendar availability compared across Heavy GitHub use (37), Other recorded use (147) and No recorded use (84).
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool. Calendar availability does not establish coding time or psychological flow. These are descriptive, unadjusted comparisons, not an effect of tool use.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Available-to-focus hours
Uninterrupted time
Interrupted time
10 May 2026 to 04 Jul 2026 | 268 distinct developers | 8 complete Person Query weeks | hours/person/week. Dates use UTC convention.
Exact GitHub-intensity group percentiles
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Heavy GitHub use | Available-to-focus hours | 37 | 28.5 | 30.3 | 32.8 |
| Other recorded use | Available-to-focus hours | 147 | 28.1 | 30.3 | 32.9 |
| No recorded use | Available-to-focus hours | 84 | 27.8 | 31.2 | 33.1 |
| Heavy GitHub use | Uninterrupted time | 37 | 9.9 | 12.4 | 13.9 |
| Other recorded use | Uninterrupted time | 147 | 8.5 | 11.6 | 14.9 |
| No recorded use | Uninterrupted time | 84 | 8.4 | 12.5 | 14.9 |
| Heavy GitHub use | Interrupted time | 37 | 17.6 | 19.0 | 19.9 |
| Other recorded use | Interrupted time | 147 | 16.6 | 18.5 | 20.7 |
| No recorded use | Interrupted time | 84 | 17.2 | 18.8 | 20.3 |
Row
How much calendar time is available for focus?
Mobile has the most uninterrupted time at 17.5 hours/person/week. The team view remains because working-hour rules and calendar load are team-level properties.
Calendar availability does not establish coding time or psychological flow. Working-hour rules vary by tenant; the synthetic 40-hour basis is not universal.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Available-to-focus hours
Uninterrupted time
Interrupted time
10 May 2026 to 04 Jul 2026 | 360 distinct developers | 8 complete Person Query weeks | hours/person/week. Dates use UTC convention.
Focus definitions and exact team percentiles
Definitions, paraphrased from Microsoft Learn. Available-to-focus hours are hours remaining during working hours after excluding meetings and scheduled Teams calls for focused work. Uninterrupted hours are one-hour-or-longer blocks of uninterrupted time. Interrupted hours are available-to-focus time interrupted by email, Teams chat, unscheduled calls or Teams channel activity. Open 1-hour block is a calendar count without scheduled meetings during the workday. The working-hours basis is set by tenant metric rules; the 40-hour basis in this synthetic data is not universal.
| Team | Metric | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Data Engineering | Available-to-focus hours | 28.5 | 30.1 | 32.7 |
| Developer Experience | Available-to-focus hours | 29.7 | 31.9 | 33.7 |
| Identity | Available-to-focus hours | 29.2 | 31.2 | 33.2 |
| Mobile | Available-to-focus hours | 32.9 | 34.5 | 36.2 |
| Payments | Available-to-focus hours | 28.5 | 29.6 | 31.7 |
| Platform | Available-to-focus hours | 24.4 | 26.0 | 27.8 |
| Data Engineering | Uninterrupted time | 9.9 | 11.9 | 13.7 |
| Developer Experience | Uninterrupted time | 12.7 | 14.2 | 16.0 |
| Identity | Uninterrupted time | 10.2 | 12.2 | 13.8 |
| Mobile | Uninterrupted time | 16.2 | 17.5 | 18.6 |
Show remaining 8 rows (18 total)
| Team | Metric | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Payments | Uninterrupted time | 8.5 | 10.3 | 11.6 |
| Platform | Uninterrupted time | 3.5 | 4.8 | 6.8 |
| Data Engineering | Interrupted time | 16.5 | 18.1 | 19.6 |
| Developer Experience | Interrupted time | 15.9 | 17.7 | 19.1 |
| Identity | Interrupted time | 17.7 | 19.2 | 20.8 |
| Mobile | Interrupted time | 15.7 | 17.3 | 18.3 |
| Payments | Interrupted time | 18.0 | 19.4 | 21.0 |
| Platform | Interrupted time | 19.3 | 20.9 | 21.9 |
After-hours
Row
How does after-hours collaboration differ by GitHub intensity?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 268 GitHub-observed developers · 8 observed weeks each
After-hours collaboration compared across Heavy GitHub use (37), Other recorded use (147) and No recorded use (84).
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool. Recorded collaboration outside configured working hours is not total work time or a wellbeing diagnosis. These are descriptive, unadjusted comparisons, not an effect of tool use.
Dot: median. Line: middle 50% of person-level eight-week averages. No health bands are assigned.
Exact GitHub-intensity group percentiles
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Heavy GitHub use | After-hours collaboration | 37 | 0.8 | 1.5 | 2.5 |
| Other recorded use | After-hours collaboration | 147 | 0.9 | 1.6 | 3.1 |
| No recorded use | After-hours collaboration | 84 | 0.9 | 1.7 | 3.3 |
Row
How is collaboration outside working hours distributed and repeated?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 360 developers · 8 observed weeks each
Payments has the highest median after-hours collaboration at 4.0 hours/person/week, compared with 1.6 hours across the population.
Recorded collaboration outside configured working hours is not total work time or a wellbeing diagnosis.
Dot: median. Line: middle 50% of person-level eight-week averages. No health bands are assigned.
Definition and interpretation limits
After-hours collaboration is time in meetings, email, Teams chats, calls and channels with at least one other person outside working hours, with overlapping activity deduplicated. Flexible schedules, on-call duties and time zones may matter, but none is recorded here. Definition paraphrased from Microsoft Learn.
Row
Persistence over the observed weeks
Convention: 3 or more after-hours collaboration hours in a week. No burnout diagnosis is implied.
Network
Row
How does network breadth differ by GitHub intensity?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 268 GitHub-observed developers · Trailing-window levels in distinct people
Collaboration-network breadth compared across Heavy GitHub use (37), Other recorded use (147) and No recorded use (84).
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool. Network measures count distinct people over a trailing window, not hours; they are standing levels and are never summed across weeks. Differences are unadjusted and descriptive; neither direction implies an effect of tool use.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Internal network size
External network size
Strong ties
Diverse ties
Network outside organisation
10 May 2026 to 04 Jul 2026 | 268 distinct developers | 8 complete Person Query weeks | distinct people per person. Trailing-window measures are never summed across weeks. Dates use UTC convention.
Exact network percentiles by GitHub-intensity group
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Heavy GitHub use | Internal network size | 37 | 36.1 | 43.6 | 60.2 |
| Other recorded use | Internal network size | 147 | 33.8 | 43.2 | 59.3 |
| No recorded use | Internal network size | 84 | 31.2 | 43.6 | 56.7 |
| Heavy GitHub use | External network size | 37 | 7.9 | 9.1 | 11.5 |
| Other recorded use | External network size | 147 | 8.3 | 10.4 | 12.3 |
| No recorded use | External network size | 84 | 8.2 | 10.2 | 12.8 |
| Heavy GitHub use | Strong ties | 37 | 10.6 | 12.8 | 16.5 |
| Other recorded use | Strong ties | 147 | 10.0 | 13.4 | 18.8 |
| No recorded use | Strong ties | 84 | 9.3 | 12.9 | 17.4 |
| Heavy GitHub use | Diverse ties | 37 | 6.9 | 9.1 | 12.8 |
Show remaining 5 rows (15 total)
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Other recorded use | Diverse ties | 147 | 7.5 | 9.5 | 13.6 |
| No recorded use | Diverse ties | 84 | 7.0 | 9.4 | 12.7 |
| Heavy GitHub use | Network outside organisation | 37 | 5.0 | 6.6 | 8.6 |
| Other recorded use | Network outside organisation | 147 | 5.0 | 6.6 | 9.4 |
| No recorded use | Network outside organisation | 84 | 4.8 | 6.4 | 8.6 |
Row
How broad is each team’s collaboration network?
Data Engineering has the largest median internal network at 46.1 people, against a population median of 42.2. The team view remains because network breadth is shaped by role and team coordination requirements.
Network measures count distinct people, not hours. Viva Insights derives them over a trailing window, so they describe a standing level of connection rather than activity inside a single week. A larger network is not inherently better; it reflects role and coordination requirements.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Internal network size
External network size
Strong ties
Diverse ties
Network outside organisation
10 May 2026 to 04 Jul 2026 | 360 distinct developers | 8 complete Person Query weeks | distinct people per person. Network measures use a trailing window and are never summed across weeks. Dates use UTC convention.
Population network baseline and exact team percentiles
| Metric | 25th percentile | Median | 75th percentile |
|---|---|---|---|
| Internal network size | 32.5 | 42.2 | 58.5 |
| External network size | 8.2 | 10.2 | 12.3 |
| Strong ties | 9.6 | 12.7 | 17.8 |
| Diverse ties | 7.1 | 9.2 | 13.4 |
| Network outside organisation | 4.8 | 6.5 | 8.9 |
| Team | Metric | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Data Engineering | Internal network size | 33.0 | 46.1 | 59.8 |
| Developer Experience | Internal network size | 35.6 | 43.7 | 63.2 |
| Identity | Internal network size | 30.2 | 39.9 | 58.4 |
| Mobile | Internal network size | 32.1 | 41.6 | 57.1 |
| Payments | Internal network size | 31.0 | 39.8 | 55.4 |
| Platform | Internal network size | 36.7 | 44.2 | 54.8 |
| Data Engineering | External network size | 7.8 | 9.7 | 11.8 |
| Developer Experience | External network size | 8.0 | 9.9 | 12.5 |
| Identity | External network size | 8.1 | 10.2 | 12.3 |
| Mobile | External network size | 8.1 | 10.4 | 12.1 |
Show remaining 20 rows (30 total)
| Team | Metric | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|
| Payments | External network size | 8.5 | 10.8 | 12.5 |
| Platform | External network size | 8.7 | 10.8 | 12.8 |
| Data Engineering | Strong ties | 9.2 | 12.8 | 18.1 |
| Developer Experience | Strong ties | 9.8 | 13.6 | 18.4 |
| Identity | Strong ties | 9.3 | 12.0 | 17.7 |
| Mobile | Strong ties | 9.8 | 12.2 | 18.4 |
| Payments | Strong ties | 9.2 | 11.9 | 16.7 |
| Platform | Strong ties | 10.7 | 13.0 | 16.7 |
| Data Engineering | Diverse ties | 7.0 | 9.8 | 14.7 |
| Developer Experience | Diverse ties | 7.2 | 10.0 | 14.8 |
| Identity | Diverse ties | 6.7 | 8.9 | 13.6 |
| Mobile | Diverse ties | 7.4 | 9.1 | 13.6 |
| Payments | Diverse ties | 6.7 | 9.1 | 12.5 |
| Platform | Diverse ties | 8.2 | 9.3 | 11.2 |
| Data Engineering | Network outside organisation | 4.6 | 6.4 | 9.3 |
| Developer Experience | Network outside organisation | 5.0 | 6.8 | 9.4 |
| Identity | Network outside organisation | 4.8 | 6.3 | 8.2 |
| Mobile | Network outside organisation | 4.6 | 6.1 | 9.3 |
| Payments | Network outside organisation | 4.7 | 6.4 | 8.2 |
| Platform | Network outside organisation | 5.4 | 7.3 | 8.7 |
Internal network size counts distinct internal people collaborated with. Strong ties and diverse ties describe the depth and spread of those connections. Network outside organisation counts connections beyond the person’s own organisational unit, and external network size counts connections outside the company. These overlap and must not be added.
GitHub breakdowns
Row
Which GitHub workflows, models and languages appear?
Synthetic demo · 10 May 2026 to 04 Jul 2026 weekdays · Daily GitHub breakdown exports
The largest reportable feature is code_completion at 30.8% of supplied feature usage.
Shared allocation and completion-model attribution are synthetic illustrations, not verified real-export semantics. Shares are calculated separately within each supplied breakdown.
Feature
Model
Language
What the synthetic allocation does and does not establish
The generator constructs all four breakdown totals from accepted
completions plus user-initiated chats and checks its own internal
consistency. This is not an authoritative real-export metric definition;
the reader does not reject inputs whose totals differ. Model
attribution, including completions, is illustrative. An authoritative
metric-definition source is needed before enforcing these equalities on
real data. Share is derived here, not an export column. All
linked margins and cross-tabs are withheld if a privacy gate fails.
Row
Cross-tabs: model × feature and language × model
| Model | Feature | People (count) | Usage count | Share within model (%) |
|---|---|---|---|---|
| claude-sonnet-5 | code_completion | 184 | 4,576 | 31.5% |
| gpt-5.4-mini | code_completion | 184 | 4,039 | 30.5% |
| gpt-5.4 | code_completion | 184 | 4,008 | 30.3% |
| claude-sonnet-5 | chat_panel_ask_mode | 184 | 3,026 | 20.9% |
| gpt-5.4 | chat_panel_ask_mode | 184 | 2,770 | 21.0% |
| gpt-5.4-mini | chat_panel_ask_mode | 184 | 2,652 | 20.0% |
| claude-sonnet-5 | chat_inline | 184 | 2,505 | 17.3% |
| gpt-5.4-mini | chat_inline | 184 | 2,273 | 17.2% |
| gpt-5.4 | chat_inline | 184 | 2,257 | 17.1% |
| claude-sonnet-5 | agent_edit | 63 | 1,045 | 7.2% |
Show remaining 5 rows (15 total)
| Model | Feature | People (count) | Usage count | Share within model (%) |
|---|---|---|---|---|
| gpt-5.4-mini | agent_edit | 63 | 967 | 7.3% |
| gpt-5.4-mini | chat_panel_agent_mode | 90 | 956 | 7.2% |
| claude-sonnet-5 | chat_panel_agent_mode | 90 | 950 | 6.5% |
| gpt-5.4 | chat_panel_agent_mode | 90 | 939 | 7.1% |
| gpt-5.4 | agent_edit | 63 | 890 | 6.7% |
| Language | Model | People (count) | Usage count | Share within language (%) |
|---|---|---|---|---|
| unknown | claude-sonnet-5 | 184 | 2,469 | 34.0% |
| unknown | gpt-5.4-mini | 184 | 2,408 | 33.1% |
| unknown | gpt-5.4 | 184 | 2,392 | 32.9% |
| typescript | claude-sonnet-5 | 66 | 2,092 | 37.5% |
| typescript | gpt-5.4 | 66 | 1,776 | 31.8% |
| typescript | gpt-5.4-mini | 66 | 1,718 | 30.8% |
| go | claude-sonnet-5 | 88 | 1,536 | 34.3% |
| go | gpt-5.4-mini | 88 | 1,484 | 33.2% |
| go | gpt-5.4 | 88 | 1,456 | 32.5% |
| python | claude-sonnet-5 | 59 | 1,448 | 35.3% |
Show remaining 5 rows (15 total)
| Language | Model | People (count) | Usage count | Share within language (%) |
|---|---|---|---|---|
| javascript | claude-sonnet-5 | 34 | 1,413 | 35.9% |
| python | gpt-5.4 | 59 | 1,333 | 32.5% |
| python | gpt-5.4-mini | 59 | 1,317 | 32.1% |
| javascript | gpt-5.4-mini | 34 | 1,282 | 32.5% |
| javascript | gpt-5.4 | 34 | 1,244 | 31.6% |
Language × feature detail
| Language | Feature | People (count) | Usage count | Share within language (%) |
|---|---|---|---|---|
| typescript | code_completion | 66 | 1,912 | 34.2% |
| unknown | code_completion | 184 | 1,832 | 25.2% |
| javascript | code_completion | 34 | 1,499 | 38.1% |
| unknown | chat_panel_ask_mode | 184 | 1,486 | 20.4% |
| unknown | chat_inline | 184 | 1,360 | 18.7% |
| go | code_completion | 88 | 1,331 | 29.7% |
| python | code_completion | 59 | 1,306 | 31.9% |
| typescript | chat_panel_ask_mode | 66 | 1,141 | 20.4% |
| sql | code_completion | 61 | 1,075 | 32.4% |
| markdown | code_completion | 34 | 1,014 | 36.0% |
Show remaining 10 rows (20 total)
| Language | Feature | People (count) | Usage count | Share within language (%) |
|---|---|---|---|---|
| go | chat_panel_ask_mode | 88 | 958 | 21.4% |
| typescript | chat_inline | 66 | 893 | 16.0% |
| java | code_completion | 63 | 872 | 29.5% |
| python | chat_panel_ask_mode | 59 | 849 | 20.7% |
| go | chat_inline | 88 | 835 | 18.7% |
| javascript | chat_panel_ask_mode | 34 | 814 | 20.7% |
| yaml | code_completion | 59 | 688 | 26.8% |
| typescript | agent_edit | 34 | 668 | 12.0% |
| python | chat_inline | 59 | 662 | 16.2% |
| sql | chat_panel_ask_mode | 61 | 654 | 19.7% |
Manager next step. Use these breakdowns to ask which workflows each feature supports. Do not treat agent-type features or model choice as maturity, quality, productivity or time-saved measures.
GitHub coverage
Row
What GitHub observation and product coverage do these exports support?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 360-developer roster
184 developers have recorded GitHub use among 268 with complete activity observation in the baseline. The heavy-user definition on the Overview is built on this population.
Observed means complete activity-row coverage, not licensing. Microsoft 365 complete-window rates and two-product comparisons remain unavailable without independent completeness evidence.
No recorded use requires complete observation. Unresolved observation is not inactivity.
Why joint-product comparisons are unavailable
Microsoft 365 eligibility uses that week’s
Total_Copilot_enabled_days; static metadata cannot override
a zero-enabled week. Eligibility and collection completeness are
different requirements. The product calendar, static licence metadata
and positive consumption rows do not certify complete collection.
Only and neither require established eligibility and observation. Any sub-10 cell withholds the entire partition. Groups are ranked by the share of GitHub observed; M365 not enabled all weeks.
Product denominators and coverage derivation
GitHub activity includes explicit zero rows for inactive weekdays;
absent rows leave activity observation unknown. GitHub credits are
sparse billable-day records, so credit coverage is resolved separately
and never invalidates an observed activity week. Missing Microsoft 365
values stay NA unless independent person-week coverage and
positive enabled days justify a zero.
| Published tool-use status | Developers (count) | Share of roster (%) |
|---|---|---|
| GitHub observed; M365 not enabled all weeks | 31 | 8.6% |
| Observation unresolved | 329 | 91.4% |
| Product | Status | People |
|---|---|---|
| GitHub Copilot | Eligibility or observation unknown | 92 |
| GitHub Copilot | Eligible and valid | 268 |
| Microsoft 365 Copilot | Eligible; completeness unresolved or incomplete | 317 |
| Microsoft 365 Copilot | Not eligible for the full window | 43 |
| GitHub activity coverage | GitHub credit coverage | Developers (count) |
|---|---|---|
| Resolved | Resolved | 268 |
| Unresolved | Unresolved | 92 |
| Question | Signal used |
|---|---|
| Person Query roster and weekly completeness | PersonQuery.csv has one row per PersonId x week in this sample. |
| Microsoft 365 Copilot eligibility | Person Query Total_Copilot_enabled_days in the relevant week; static metadata never overrides zero enabled days. |
| GitHub Copilot observation | An activity row establishes observation for that week, not licence or provisioning status. Absence is unknown. |
| GitHub weekly activity coverage | Five explicit weekday activity rows, including measured zeros; a weekday-only reporting convention, not an ingestion certificate. No absent GitHub values are zero-filled. |
| GitHub weekly credit coverage | The credit export is sparse and carries a row only for billable usage, so coverage is resolved when every observed billable day has a credit row. A week with no billable day resolves to a measured zero. Credit coverage never invalidates observed activity. |
| Microsoft 365 weekly observation | Unknown by default. Independent completeness evidence is required; licence metadata and a calendar are insufficient. |
| Recorded product use | Positive observations are not proof of complete collection. M365 zero filling requires independently evidenced completeness and positive weekly enabled days. |
Row
Product-specific frequency, intensity and services
| Product measure | Microsoft 365 Copilot | GitHub Copilot |
|---|---|---|
| Valid developers (count) | 0 | 268 |
| Recorded active (count) | 0 | 184 |
| Recorded active (%) | Undefined — no valid denominator | 68.7% |
| Active days / developer / week (mean) | N/A — unavailable or withheld | 2.3 |
| Credit-resolved developers (count) | 0 | 268 |
| Credits / developer / week (mean) | N/A — unavailable or withheld | 51.8 |
| Session or request unit | M365 credits | GitHub credits |
| Top 10% active credit share (%) | N/A — unavailable or withheld | 38.6% |
| Top group (count) | N/A — unavailable or withheld | 19 |
Microsoft 365 population rates are unavailable here, not zero. GitHub AI credits and Microsoft 365 Copilot credits are different units and are never pooled, totalled or cross-shared; completion suggestions and chat requests are different units again.
Observed Microsoft 365 service records and concentration definition
| Service | People (count) | Sessions | Credits | Credit share (%) |
|---|---|---|---|---|
| Cowork | 256 | 17,565 | 1,005,887.5 | 34.0% |
| WorkIQ | 256 | 17,326 | 992,502.5 | 33.5% |
| Analyst | 121 | 8,399 | 524,136.1 | 17.7% |
| Researcher | 131 | 7,973 | 436,248.1 | 14.7% |
The concentration row is the share of GitHub credits or Microsoft 365 credits attributable to the highest-volume 10% of active developers for that product. Identities and individual rankings are never displayed.
Working patterns
Row
How do working conditions differ across GitHub-intensity groups?
Synthetic demo · 10 May 2026 to 04 Jul 2026 · 268 developers with complete GitHub activity coverage
Compare working conditions across Heavy GitHub use (37), Other recorded use (147) and No recorded use (84).
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool. Unadjusted, descriptive comparisons only. Use and working conditions share the same window, so temporal direction and intervention effects cannot be established.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Collaboration hours
Meetings
Available-to-focus hours
Uninterrupted time
After-hours collaboration
10 May 2026 to 04 Jul 2026 | 268 distinct developers | 8 complete Person Query weeks | hours/person/week. Dates use UTC convention.
Exact reportable group sizes and percentile values
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| No recorded use | After-hours collaboration | 84 | 0.9 | 1.7 | 3.3 |
| Other recorded use | After-hours collaboration | 147 | 0.9 | 1.6 | 3.1 |
| Heavy GitHub use | After-hours collaboration | 37 | 0.8 | 1.5 | 2.5 |
| No recorded use | Available-to-focus hours | 84 | 27.8 | 31.2 | 33.1 |
| Other recorded use | Available-to-focus hours | 147 | 28.1 | 30.3 | 32.9 |
| Heavy GitHub use | Available-to-focus hours | 37 | 28.5 | 30.3 | 32.8 |
| Heavy GitHub use | Collaboration hours | 37 | 12.1 | 15.4 | 17.7 |
| No recorded use | Collaboration hours | 84 | 12.0 | 15.0 | 19.6 |
| Other recorded use | Collaboration hours | 147 | 11.5 | 14.9 | 17.9 |
| Other recorded use | Meetings | 147 | 6.0 | 8.4 | 10.3 |
Show remaining 5 rows (15 total)
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Heavy GitHub use | Meetings | 37 | 5.9 | 8.4 | 9.9 |
| No recorded use | Meetings | 84 | 5.6 | 7.4 | 10.6 |
| No recorded use | Uninterrupted time | 84 | 8.4 | 12.5 | 14.9 |
| Heavy GitHub use | Uninterrupted time | 37 | 9.9 | 12.4 | 13.9 |
| Other recorded use | Uninterrupted time | 147 | 8.5 | 11.6 | 14.9 |
Comparison population and limits
The comparison includes developers valid for the Person Query and GitHub activity throughout the baseline. Unresolved GitHub observation remains in the full roster but is excluded here, not recoded as no use. This is not statistical matching. A two-product comparison remains unavailable because Microsoft 365 completeness is not established. Independent completeness evidence and the privacy floor are prerequisites for extending it.
Row
Network breadth by GitHub intensity
Team and role composition of the heavy-use group is withheld because linked breakdowns could reveal a small group. Unadjusted comparisons may reflect team or role composition rather than an effect of the tool. Whether GitHub intensity coincides with broader or narrower collaboration networks. Network measures count distinct people over a trailing window, so they are standing levels rather than weekly activity. Differences are unadjusted and descriptive; neither direction implies an effect of tool use.
Dot: median. Line: middle 50% of person-level averages. Each panel has its own horizontal scale; group order is shared.
Internal network size
External network size
Strong ties
Diverse ties
Network outside organisation
10 May 2026 to 04 Jul 2026 | 268 distinct developers | 8 complete Person Query weeks | distinct people per person. Trailing-window measures are never summed across weeks. Dates use UTC convention.
Exact network percentiles by reportable group
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Heavy GitHub use | Internal network size | 37 | 36.1 | 43.6 | 60.2 |
| Other recorded use | Internal network size | 147 | 33.8 | 43.2 | 59.3 |
| No recorded use | Internal network size | 84 | 31.2 | 43.6 | 56.7 |
| Heavy GitHub use | External network size | 37 | 7.9 | 9.1 | 11.5 |
| Other recorded use | External network size | 147 | 8.3 | 10.4 | 12.3 |
| No recorded use | External network size | 84 | 8.2 | 10.2 | 12.8 |
| Heavy GitHub use | Strong ties | 37 | 10.6 | 12.8 | 16.5 |
| Other recorded use | Strong ties | 147 | 10.0 | 13.4 | 18.8 |
| No recorded use | Strong ties | 84 | 9.3 | 12.9 | 17.4 |
| Heavy GitHub use | Diverse ties | 37 | 6.9 | 9.1 | 12.8 |
Show remaining 5 rows (15 total)
| Group | Metric | Developers (count) | 25th percentile | Median | 75th percentile |
|---|---|---|---|---|---|
| Other recorded use | Diverse ties | 147 | 7.5 | 9.5 | 13.6 |
| No recorded use | Diverse ties | 84 | 7.0 | 9.4 | 12.7 |
| Heavy GitHub use | Network outside organisation | 37 | 5.0 | 6.6 | 8.6 |
| Other recorded use | Network outside organisation | 147 | 5.0 | 6.6 | 9.4 |
| No recorded use | Network outside organisation | 84 | 4.8 | 6.4 | 8.6 |
Row
Role and team context for the comparison
Group sizes and roles can shape the comparison. Any group with a role cell below 10 is omitted from this breakdown. Groups are ranked by the share of Application engineering.
Exact role shares and team composition
| Group | Category | Developers (count) | Share (%) |
|---|---|---|---|
| GitHub use recorded | Application engineering | 85 | 46.2% |
| GitHub use recorded | Engineering management | 42 | 22.8% |
| GitHub use recorded | Infrastructure engineering | 57 | 31.0% |
| No GitHub use recorded | Application engineering | 39 | 46.4% |
| No GitHub use recorded | Engineering management | 15 | 17.9% |
| No GitHub use recorded | Infrastructure engineering | 30 | 35.7% |
| Group | Category | Developers (count) | Share (%) |
|---|---|---|---|
| GitHub use recorded | Data Engineering | 30 | 16.3% |
| GitHub use recorded | Developer Experience | 34 | 18.5% |
| GitHub use recorded | Identity | 28 | 15.2% |
| GitHub use recorded | Mobile | 32 | 17.4% |
| GitHub use recorded | Payments | 31 | 16.8% |
| GitHub use recorded | Platform | 29 | 15.8% |
| No GitHub use recorded | Data Engineering | 14 | 16.7% |
| No GitHub use recorded | Developer Experience | 14 | 16.7% |
| No GitHub use recorded | Identity | 14 | 16.7% |
| No GitHub use recorded | Mobile | 14 | 16.7% |
Show remaining 2 rows (12 total)
| Group | Category | Developers (count) | Share (%) |
|---|---|---|---|
| No GitHub use recorded | Payments | 14 | 16.7% |
| No GitHub use recorded | Platform | 14 | 16.7% |
Interpretation. Team, role, seniority, task selection and tenure can influence working patterns and tool use. Heavy GitHub use may differ across these groups. These charts are intentionally limited to agreed measures. No broad metric scan, adjustment model, causal claim or preferred tool-use group is presented.
Row
Weekly trends and valid denominators
At least one positive activity day that week. Missing coverage is not inactivity; Microsoft 365 rates are unavailable by default without independent completeness evidence.
Coverage and numerators for every weekly window
| Week beginning (UTC) | PQ developers (count) | GH valid (count) | GH active (count) | GH active (%) | M365 valid (count) | M365 active (count) | M365 active (%) | Both valid (count) |
|---|---|---|---|---|---|---|---|---|
| 2026-01-04 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-01-11 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-01-18 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-01-25 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-02-01 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-02-08 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-02-15 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-02-22 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-03-01 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-03-08 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
Show remaining 16 rows (26 total)
| Week beginning (UTC) | PQ developers (count) | GH valid (count) | GH active (count) | GH active (%) | M365 valid (count) | M365 active (count) | M365 active (%) | Both valid (count) |
|---|---|---|---|---|---|---|---|---|
| 2026-03-15 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-03-22 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-03-29 | 360 | 0 | 0 | Undefined — no valid denominator | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-04-05 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-04-12 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-04-19 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-04-26 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-05-03 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-05-10 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-05-17 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-05-24 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-05-31 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-06-07 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-06-14 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-06-21 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
| 2026-06-28 | 360 | 268 | 184 | 68.7% | 0 | 0 | Undefined — no valid denominator | 0 |
Row
What evidence would support an intervention?
Select a workflow question, then combine these signals with anonymous developer feedback and delivery outcomes. Survey, review-delay, deployment and service-quality outcomes are not in these exports. See the SPACE, DevEx and DORA evidence guide in Methods.
Decision supported now. Select a workflow question and close its measurement gaps. An intervention effect, productivity gain or wellbeing conclusion requires additional evidence. No operational action is justified by synthetic values.
Methods
Row
Definitions, sources and reporting contracts
Synthetic demo · Real export schema
Scope and units. Pages use the last eight complete Person Query weeks, 10 May 2026 to 04 Jul 2026, and product weekdays that fall in the final 13-week product window. The full Person Query source window is 04 Jan 2026 to 04 Jul 2026, with Sunday-start weeks. Dates are ISO dates interpreted using a UTC convention. This is an illustrative date contract, not evidence about developers’ time zones. Available-to-focus hours are based on working hours configured through tenant metric rules; the 40-hour basis in this synthetic file is not universal.
Aggregation. Working-condition values are weekly Person Query metrics. Baseline charts first average each person’s eight weekly values and then compute group percentiles. Each developer has equal weight. Product-active rates divide active eligible, observed developers by all eligible, observed developers for that product and window. Daily Microsoft 365 service rows are aggregated before person-day and person-week analysis.
Source definitions. Metric descriptions are
paraphrased from the Microsoft
Viva Insights metrics reference. The source manifest is
_data/README.md; the report reads the committed CSVs listed
there.
| Relative file under _data | Grain | Contract |
|---|---|---|
| person-query/PersonQuery.csv | Person x week | Weekly panel spanning 26 weeks. Person Query anchors the roster. |
| consumption-query/PeopleMetaData.csv | Person | Consumption metadata keyed by PeopleHistoricalId; static licence context only, not period eligibility. |
| consumption-query/PersonM365CreditsMetrics.csv | Person x service x day | Microsoft 365 Copilot credits and session counts; service rows are aggregated before daily or weekly use. |
| consumption-query/PersonGitHubCreditsMetrics.csv | Person x day | GitHub AI credits by person and day. |
| github-query/PersonGitHubActivityMetrics.csv | Person x day | Synthetic GitHub activity includes explicit inactive weekday zeros; absent rows leave provisioning unknown. |
| github-query/GitHubActivityBreakdownByFeatureMetrics.csv | Person x day x feature | Feature usage count. Synthetic fixture allocation is not a verified real metric equality. |
| github-query/GitHubActivityBreakdownByLanguageFeatureMetrics.csv | Person x day x language x feature | Language x feature usage; shared allocation is synthetic only. |
| github-query/GitHubActivityBreakdownByLanguageModelMetrics.csv | Person x day x language x model | Language x model usage; model attribution, including completions, is illustrative. |
| github-query/GitHubActivityBreakdownByModelFeatureMetrics.csv | Person x day x model x feature | Model x feature usage; model attribution, including completions, is illustrative. |
Joins and missingness. Person Query anchors the
roster and weekly panel. Consumption metadata is keyed by
PeopleHistoricalId, which is an opaque key read from the
activity-file crosswalk and never reconstructed from
PersonId; the Person Query is used for HR attributes
because it already carries PersonId. Daily feeds aggregate
to unique person-week keys before left joins. One-to-one and many-to-one
join assertions prevent row amplification. 0 is an observed
or justified zero. NA is unresolved measurement and is
never silently replaced outside valid observation. GitHub activity
coverage and GitHub credit coverage are separate flags, each gating only
the measures it governs.
Row
Privacy, reproducibility and validation
Privacy floor. At least 10 distinct people are
required for a published group. Percentile visuals use
vivainsights::create_boxplot(..., mingroup = 10, return = "table").
Linked HR margins and cross-tabs share one release gate: every cell of
every published margin must clear the floor, so no published margin can
be differenced against another to expose a sub-floor group. Joint-use
and eligibility partitions are withheld in full when unsafe. Linked
GitHub margins and cross-tabs are withheld together, and their release
additionally requires that no person’s pattern of cell membership is
shared by fewer than 10 people. Service releases check positive
contributors and overlapping membership-pattern complements within the
developer roster. Product rates check both numerator and complement;
unavailable values are labelled, not plotted as zero. Missing category
labels are shown as “Unknown / not supplied”. No individual points,
identifiers or individual leaderboards appear in the HTML.
Simulation boundary. The CSV values are synthetic,
but the report no longer invents CSV schemas or writes files. It does
not use a standalone eligibility or coverage export, fabricated
Microsoft 365 action categories, model or language share files, or an
agent-use indicator that is absent from the real GitHub activity export.
It parses the real Agent adoption field and derives all
shares locally.
Coverage boundary. There is no default
fixture-completeness exception or filename-based trust.
derive_m365_coverage(person_weeks, evidence = NULL) returns
unknown completeness unless a caller explicitly supplies independently
obtained, person-week-specific evidence with its source. That is an
in-memory analysis input, not a fabricated export column or file. A
calendar, static licence status, observed positive rows, or missing rows
cannot supply that evidence. A future fixture-only opt-in would need
validated provenance tied to the exact files; assumptions must not
survive replacement or row deletion.
| Check | Result |
|---|---|
| Population | 480 people; 360 developers |
| Teams | Team metrics require the privacy floor; complementary product cells are withheld together. |
| Person Query developer-weeks | 9,360 |
| Baseline developer-weeks | 2,880 |
| Focus-time identities | Available-to-focus hours = working-hours basis minus meetings and scheduled calls; uninterrupted + interrupted = available-to-focus |
| Source keys and joins | Unique keys; no join amplification. PeopleHistoricalId is read from the activity crosswalk, never reconstructed from PersonId. |
| Non-negative metrics and acceptance bounds | Passed |
| GitHub synthetic allocation invariant | Checked only by the synthetic generator; not enforced as a real-export contract. Model attribution is illustrative, including completions. |
| Eligibility and observation derivation | Weekly enabled days govern M365 eligibility; independent M365 completeness unavailable by default. No coverage file invented. |
| Activity and credit coverage separated | Activity and credit coverage are assessed separately. Sparse credit rows never void observed activity; publishable counts appear in the coverage section. |
Show remaining 2 rows (12 total)
| Check | Result |
|---|---|
| Joint reconciliation | Internal joint counts reconcile to the roster; public partitions are withheld if any positive cell is below the privacy floor. |
| Privacy floor | At least 10 distinct people in every published group |
Runtime versions and package choice
| Component | Version |
|---|---|
| R | 4.6.1 |
| dplyr | 1.2.1 |
| tidyr | 1.3.2 |
| ggplot2 | 4.0.3 |
| vivainsights | 0.7.2 |
| rmarkdown | 2.31 |
| flexdashboard | 0.6.3 |
The vivainsights function index supports the person-averaged boxplot summary used for percentile calculations. Custom joint coverage aggregation is necessary because single-product usage segments do not establish comparable observation for two products.
Row
Research framework and future extensions
| Framework | What this report can contribute | Evidence still needed |
|---|---|---|
| SPACE | Activity, collaboration and calendar-derived working conditions across multiple dimensions | Satisfaction and wellbeing surveys, performance outcomes, direct evidence of efficiency and flow |
| DevEx | Context for feedback loops, cognitive load and flow discussions | Anonymous developer feedback, build/test waits, review waits, task context and perceived ability to focus |
| DORA | A framework for evaluating delivery change, without an outcome claim | Service boundaries, production deployments, change lead times, recovery events, failed changes and deployment rework |
- The SPACE of Developer Productivity, ACM Queue, 2021. Productivity is multidimensional and cannot be reduced to one activity measure.
- DevEx: What Actually Drives Productivity, ACM Queue, 2023. Connect feedback loops, cognitive load and flow with developer feedback and system measures.
- DORA’s software delivery performance metrics. Current guidance describes five metrics across throughput and instability, best applied to one application or service at a time.
- Viva Insights metrics reference. Primary reference for the paraphrased collaboration and focus definitions.