Activity Classification

Know what your team actually does, not what Jira says

Your current tools count commits and PRs. They can't tell the difference between deep coding work, firefighting, review marathons, and meeting-driven days. Gitrevio derives a work-mode distribution from your team's engineering activity, breaking the week into meaningful activity types. Some modes (debugging, mentoring, context switching) depend on a connected calendar and AI-tagging.

GitRevio Overview Activity Reports AI Chat Settings AA ACTIVITY Sarah Chen Week of Apr 14 Team Summary Trend Analysis Comparisons SIGNALS Commits PR Activity Calendar Slack Windows Activity Classification — Sarah Chen Week of April 14 · work-mode distribution · illustrative Deep coding 0 % 15.2h this week Code review 22 % 8.8h this week Debugging 12 % 4.8h this week Work modes tracked 7 derived from activity Activity Breakdown — Week of April 14 Deep coding 38% Code review 22% Debugging 12% Meetings 11% Context switching 9% Mentoring 5% Admin / process 3% Team Comparison — Week of April 14 Backend (6 engineers) Deep coding 45% Code review 22% Debugging 14% Frontend (4 engineers) Deep coding 28% Code review 15% Debugging 31% ! Frontend debugging is 2.2x backend rate

How It Classifies

How it classifies a week of work

# Activity classification — Sarah Chen, week of April 14

Deep coding     ████████████░░░░  38%  (15.2h)

Code review     ████████░░░░░░░░  22%  ( 8.8h)

Debugging       ████░░░░░░░░░░░░  12%  ( 4.8h)

Meetings        ████░░░░░░░░░░░░  11%  ( 4.4h)

Context switch  ███░░░░░░░░░░░░░   9%  ( 3.6h)

Mentoring       ██░░░░░░░░░░░░░░   5%  ( 2.0h)

Admin/process   ░░░░░░░░░░░░░░░   3%  ( 1.2h)

# Work-mode distribution (illustrative)

# Derived from engineering activity across your integrations

The classifier ingests signals from every tool your engineers touch. Commit patterns, PR timelines, file types touched, Jira transitions, calendar events, review comment patterns, Slack activity windows — hundreds of events per day, per person.

It doesn't just look at what happened — it understands the shape of the work. A developer who commits to three files in one repo over four hours is doing deep coding. The same developer touching five repos in one hour is context switching.

No self-reporting. No time tracking tools. No interruptions. Your engineers keep working. Gitrevio watches the signals they already produce.

What It Sees

What the classifier sees

Each activity type is defined by a distinct pattern of signals, not a single event. The model weighs dozens of features to distinguish between categories with overlapping indicators.

Deep coding

Sustained commits to few files, low task switching, test writes, long uninterrupted sessions

Code review

PR comment timing, review depth, files viewed vs commented, suggestion-to-approval ratio

Debugging

Hot-fix branches, log file changes, rapid commit-revert cycles, stack trace references in comments

Mentoring

Review comments on junior PRs, pair programming patterns, explanatory comment length, repeated reviewer pairing

Meetings

Calendar blocks, gaps in commit activity, Slack silence during meeting windows, post-meeting commit bursts

Context switching

3+ repos in one hour, task-switching velocity, incomplete commits, branch hopping across projects

Admin / process

Jira grooming, sprint planning artifacts, config file updates, CI pipeline maintenance, access requests

Documentation

Markdown and README edits, wiki contributions, ADR creation, inline comment density spikes

Firefighting

Off-hours commits, incident channel activity, rollback patterns, rapid deploy cycles to production

Team-level patterns tell a bigger story

Individual activity breakdowns are useful. Team-level comparisons are where the real insights emerge. When your backend team spends 45% of their time in deep coding but your frontend team only hits 28%, that's not a talent gap — it's a process signal.

The frontend team's 31% debugging rate tells you they're spending more time fixing than building. Maybe it's a testing gap. Maybe it's a flaky CI pipeline. Maybe it's tech debt nobody wants to touch. Now you have the data to find out.

Track these patterns over time to see if your interventions actually work. Moved standup to async? Watch the meeting percentage drop. Invested in test coverage? Watch debugging hours shrink.

# Team comparison — week of April 14

Backend (6 engineers)

Deep coding     ████████████░░░░  45%

Code review     ██████░░░░░░░░░░  22%

Debugging       ████░░░░░░░░░░░░  14%

Meetings        ███░░░░░░░░░░░░░   8%

Other           ███░░░░░░░░░░░░░  11%

Frontend (4 engineers)

Deep coding     ███████░░░░░░░░░  28%

Code review     ████░░░░░░░░░░░░  15%

Debugging       ████████░░░░░░░░  31%

! Frontend debugging is 2.2x backend rate

What This Unlocks

What this unlocks

01

01

Identify meeting-heavy teams

Find teams where meetings consume more than 20% of the week — and track whether async experiments actually reduce that number.

02

02

Find context-switching hotspots

Spot engineers who are spread across too many repos and projects. Context switching kills deep work — now you can measure it.

03

03

Understand velocity differences

When two teams have different velocity, activity classification tells you why. One team codes 45% of the time; the other only 28%.

04

04

Measure mentoring investment

See how much senior engineer time actually goes to mentoring. Track whether new hires get the review attention they need.

05

05

Detect burnout patterns

Declining deep coding percentage, increasing context switching, growing off-hours work — early warning signals you can act on before it's too late.

06

06

Validate process changes

Moved retros to async? Added a code review SLA? See the before-and-after in activity data, not just subjective survey responses.

Ready to See Your Engineering work clearly?

Get started free