How dev agencies should track time
How dev agencies should track time
Dev agency time tracking exists for two reasons: accurate client billing and honest utilization measurement. Pick one tool that ties hours to clients and projects (Harvest or Toggl Track for small shops, Productive.io or Tempo for agencies on Jira), enter time weekly in 30-minute blocks, and never use the data to police engineers. Treat timesheets as a billing instrument, not a productivity microscope, and the numbers stay honest.
That distinction is the whole post. Most agencies get it wrong, ship surveillance instead of measurement, and then wonder why their senior engineers are interviewing elsewhere by Q3.
What dev agency time tracking is actually for
Two valid uses. One invalid one. Get this clear before you pick a tool.
Valid use 1: client billing. If you bill hourly or by retainer with hour caps, you need defensible records. The client will ask "what did you work on" and "how many hours" and you need an answer with timestamps and project codes attached. This is a legal and financial requirement, not optional.
Valid use 2: utilization measurement. You need to know what percentage of your team's available hours are billable. Below 60% across the team and you have a pipeline problem. Above 85% and you have a burnout problem. You can't run an agency without this number, and you can only get it from logged time.
Invalid use: engineer productivity surveillance. The moment you start using timesheets to evaluate individual performance ("Sarah only logged 32 billable hours this week, she's slacking"), three things happen. First, the data becomes fiction: engineers round up, log fake time, or stop pushing back on scope creep because they need the hours. Second, your best people leave. Third, you lose the ability to spot real problems because the signal is buried in defensive logging.
If you can't say out loud at an all-hands what the timesheet data is used for and have engineers nod, you're using it wrong. Honest framing produces honest data. Surveillance framing produces noise.
The 8 tools dev agencies actually use in 2026
The market hasn't really changed in five years. Here's the actual landscape, with real per-seat 2026 pricing.
| Tool | Pricing (per seat/mo) | Best for | Watch out for |
|---|---|---|---|
| Harvest | $11 (Pro) | Small services agencies, freelance to ~20 people | No real PM features; pair with Asana |
| Toggl Track | $9 (Starter) | Teams that hate timesheets | Reporting weaker than Harvest |
| Clockify | $3.99-$5.49 | Cost-conscious agencies, contractors | Paid features lag rivals |
| Productive.io | $9-$24 | 10-50 person agencies wanting one system | Steep learning curve |
| Everhour | $8.50 (Team) | Teams already on Asana, Trello, ClickUp, or GitHub | Min 5 users on Team plan |
| Tempo Timesheets | ~$3.80 at 20 users | Jira-first dev shops | Jira-only ecosystem |
| Float | ~$7.50 (Resource) | Capacity planning + light tracking | Not a primary timesheet tool |
| TMetric | ~$5 | Budget option with solid invoicing | UX less polished |
A few honest notes on each:
Harvest is the incumbent for a reason. The invoicing flow is mature, the integrations work, the iOS app actually opens fast. If you're under 20 people and bill hourly, this is the safe choice. The Pro plan at $11/seat is the one that includes invoice scheduling and Slack reminders.
Toggl Track wins on UX. Engineers will actually use it because the timer is one click and the desktop app stays out of the way. Reporting is thinner than Harvest, so you'll need to live in CSV exports if you want anything custom.
Clockify is the only tool here with a free tier that supports unlimited users. If you're a 6-person agency and don't yet need invoicing, the free plan covers tracking. The Standard tier at $5.49/seat unlocks invoicing and is still cheaper than every rival.
Productive.io is a full agency operating system: time tracking, project management, resource planning, invoicing, profitability dashboards. Essential is $9/seat annual, Professional is $24/seat. If you're 10+ people and tired of stitching together Harvest + Asana + QuickBooks, this collapses the stack into one login. The downside is real: there's a two-week learning curve, and any tool that does this much will have rough edges somewhere.
Everhour matters if your team already lives in Asana, Trello, ClickUp, Basecamp, or GitHub Issues. The timer renders inside those tools, so engineers don't context-switch to log. Team plan is $8.50/seat with a 5-user minimum.
Tempo Timesheets is the answer if you're a Jira shop and don't want anyone learning a second tool. Per-seat pricing scales down hard at volume: 20 users on Tempo Timesheets only is roughly $76/month, or under $4/seat. For 50+ engineers running on Jira, no rival comes close on price.
Float is a capacity planning tool that added time tracking. If your problem is "we don't know who's free next week," Float solves that better than any tool here. If your problem is "we need defensible billing records," pair it with Harvest or use Productive instead.
TMetric is the budget pick when Clockify feels too bare and Harvest feels too pricey. Solid invoicing, billable/non-billable splits, and a project budget feature. UX is functional, not delightful.
Decision matrix by agency size
Tool choice tracks team size more than industry. Here's the rule of thumb most agencies converge on by year three.
Solo or 1-2 people: Toggl Track free or Harvest free (1 user, 2 projects). Don't pay for what you can't yet justify.
2-10 people: Harvest, Toggl Track Premium, or Clockify Standard. All three handle multi-project billing, all three integrate with QuickBooks or Xero. Pick on UX preference; the engineers using it daily should test all three for a week before you commit.
10-30 people: Productive.io or Everhour. At this size, the cost of running tracking, PM, invoicing, and resource planning as separate tools exceeds the cost of one unified system. Productive wins if you want one platform; Everhour wins if your team already loves Asana, ClickUp, or Linear and won't migrate.
30+ people: Productive.io Professional, Tempo (if Jira), or Float plus Harvest as a combo. At this size you also need a head of operations who actually owns the tooling, because no SaaS platform survives growth without an internal champion.
You can survive scaling Harvest to 50 engineers, but you'll spend Friday afternoons every quarter cleaning up project codes. Better to migrate at 20.
Five rules for entering time without losing engineers
Tool choice matters less than entry discipline. These five rules are the difference between a clean dataset and a fiction.
1. Track in 30-minute blocks, not 1-minute precision. No engineer remembers what they did at 2:13pm versus 2:47pm. Asking them to log it at minute resolution generates either fake precision or burnout. Round to the nearest 30 minutes. Your billing accuracy improves because the data is honest.
2. Enter weekly on Friday, not daily. Daily timesheet entry sounds disciplined and is actually a productivity tax: you context-switch out of deep work to log time, then have to context-switch back. Weekly entry on Friday afternoon, with auto-sync filling in the bulk, takes 15 minutes and produces equivalent data. Daily entry produces 30 minutes of lost focus every day, five days a week.
3. Auto-sync from Linear, Jira, GitHub, and Slack where possible. Modern tools (Everhour, Tempo, Productive) can pull issue activity, commit timestamps, and meeting calendars directly. Engineers then approve or edit, instead of remembering and typing. The fewer keystrokes per logged hour, the more honest the data.
4. Expose the utilization dashboard to engineers. If you ask people to log time, they should see the same numbers you see. Hide the dashboard and you're sending the message that the data is "for management." Show it openly: team utilization, project profitability, retainer hours used. The same transparency you'd want on your own performance.
5. Never use timesheets for performance reviews. Pin this on the wall. The second timesheets become an evaluation input, the data dies. Use them for billing and capacity. Use code review, shipped features, and peer feedback for performance. These are different signals; don't conflate them.
These five rules are why agencies that "have great time tracking discipline" usually also have low engineer churn. Causation runs both ways.
What healthy utilization actually looks like
Once you have clean data, here are the benchmarks you're trying to hit, by role.
| Role | Healthy utilization | Below this means | Above this means |
|---|---|---|---|
| Developers | 60-75% | Pipeline problem | Burnout queued up |
| Designers | 70-85% | Design pipeline gap | Design quality dropping |
| PMs/leads | 40-60% billable | Under-deployed | Sales/internal work suffering |
| Founders | 20-40% billable | Healthy split | You're a contractor, not an owner |
A few things to notice. First, no role should be at 100%; the math doesn't work because everyone has email, sprint planning, internal meetings, and PTO. Second, leads and PMs are intentionally lower because their job includes business development and team coaching that doesn't bill. Third, agency utilization rates vary widely by service mix, but these ranges hold across most dev shops we've talked to.
Track these at the team or role level, never the individual. The moment Sarah-the-engineer sees her individual utilization above the group average and gets praised for it, you've started a race to the bottom on quality.
Where Cadence fits if you're an agency reselling engineers
Two patterns we see at agencies running on Cadence.
Pattern 1: white-label resale. Agencies book engineers from Cadence at the platform rate (junior $500/week, mid $1,000/week, senior $1,500/week, lead $2,000/week) and bill the end client at agency markup ($5,000-$8,000/week is typical). The engineer tracks their own time inside Cadence; the agency just bills the client at the marked-up weekly rate, no hourly reconciliation required. Every engineer on Cadence is AI-native by default, vetted on Cursor, Claude Code, and Copilot fluency before they unlock bookings, so the engineer you book ramps fast. Median time to first commit across the 12,800-engineer pool is 27 hours.
Pattern 2: partner referral. Agencies who don't want to resell can join the partner program and earn 10% recurring on every founder they refer. Useful when a prospect comes in below your minimum project size and you want to send them somewhere reputable instead of saying no.
If you run a dev agency and want a structural revenue lane that doesn't require hiring, the partner program is a 10-minute application and pays out monthly.
Three time-tracking mistakes that bleed margin
The mistakes are usually mechanical, not philosophical.
Mistake 1: tracking only billable hours. You can't compute utilization without total hours worked. If engineers only log billable time, your denominator is broken and every utilization number is wrong. Require all hours logged, with billable/non-billable as a flag.
Mistake 2: letting engineers self-categorize without project codes. By end of quarter you'll have 200 line items with names like "client work," "meeting," and "stuff for Mike." Define the project codes upfront, lock the dropdown, and review monthly. This is the single most important operations rule at most agencies.
Mistake 3: quarterly invoicing review. Review weekly. Errors compound: a wrong project code in week 2 becomes a wrong invoice in week 4 becomes a billing dispute in week 12 that costs you the client. Friday afternoon, 30 minutes, the operations lead sweeps the timesheets. Done.
These are the operational basics that separate agencies that hit 50% gross margin from agencies that don't. They're boring. They're also the difference between profitable and not.
The honest summary
Pick one tool that fits your team size. Set the entry rules so engineers don't hate Friday afternoons. Use the data for billing and utilization, not for performance management. Review weekly. Tune the pricing model underneath the tracking so the data actually drives margin decisions, not just compliance reports.
If you do this right, time tracking becomes invisible. Engineers log 15 minutes on Friday, the operations lead reviews on Monday, invoices go out, utilization shows up on a dashboard everyone can see. Nobody talks about it. That's the goal: a system so quiet it does its job without anyone having to fight it.
If you're also rebuilding your proposal process or thinking about how to find clients at the same time, do tracking first. You can't price work or measure proposal win rates without clean utilization data underneath.
If you run a dev agency and want a no-overhead revenue lane: the Cadence partner program pays 10% recurring on every founder you refer, monthly, for as long as they stay active. White-label resale is the other route: book engineers at $1,000-$2,000/week, bill the client at your normal markup.
FAQ
Should developers really track every hour?
No. Track in 30-minute blocks against client projects, not minute-by-minute. Precision past 30 minutes is invented data and erodes trust without improving billing accuracy.
Daily or weekly time entry for dev agencies?
Weekly, on Friday afternoon. Daily entry creates a context-switching tax and produces no better data once you're using auto-sync from Linear, Jira, or GitHub. Friday entry takes 15 minutes; daily entry costs 30 minutes of lost focus every day.
What's a healthy utilization rate for developers at an agency?
60-75% billable across the dev team. Below 60% signals a pipeline problem. Above 85% means burnout is queued up and quality is about to drop.
Is automatic time tracking (Rize, Timely) worth it?
For solo freelancers, often yes. For agencies, usually no: auto-categorization fails on multi-client context-switching, and engineers feel monitored. Use auto-sync from Linear, Jira, or GitHub instead, which logs project activity without screen recording.
Can I use timesheet data for performance reviews?
Don't. The moment you do, the data becomes fiction. Use timesheets for billing and capacity planning. Use code review, shipped features, and peer feedback for performance. They're different signals.