The problem with time tracking data is not that agencies do not look at it. It is that the most available numbers are the ones most likely to point somewhere misleading, and they do it while looking authoritative.
Here are four that come as standard on every dashboard in this category, including ours.
1. Total hours logged
What people read it as: how hard the team is working.
What it measures: how diligently the team fills in a timesheet.
These correlate weakly. A team whose logged hours rise 8% may be working more, or may have got better at logging — and the second is more common, particularly in the months after a rollout, when the number rises steadily for reasons that have nothing to do with effort.
It runs in the other direction too. When people believe hours are being read as effort, weekly totals converge on contracted hours: nobody logs the 34-hour week, and nobody logs the 52-hour one either. The metric goes quiet exactly when it had something to tell you. See why timesheets get reconstructed on Friday.
Read alongside: variance. A team whose weekly totals vary is logging honestly. One where every week lands on 37.5 is filling in a form.
2. Utilization rate
What people read it as: productivity, and higher is better.
What it measures: the proportion of available time sold — which is a resourcing signal, not a performance one.
Two failures, both common. Averaging conceals distribution: a comfortable 72% agency average routinely contains someone at 95% and someone at 48%, and the average describes neither. And high is not good — above about 85% sustained, you have no capacity for new business, training or absence, and you are running down your people to buy a margin smaller than the cost of replacing one of them.
Read alongside: the per-person distribution, and effective hourly rate. High utilization on underpriced work is an efficient way to lose money, and utilization alone will never show it. More on the calculation and its traps in how to calculate utilization.
3. Billable percentage
What people read it as: how much of our time earns money.
What it measures: how your team interprets the word "billable".
If travel, internal meetings about a client, and rework are categorized differently by different people — and without a written policy they will be — this number is partly a measure of interpretation. It also moves when you change the definitions, which makes year-on-year comparison meaningless unless you kept the rules stable.
The deeper issue is that billable and billed are different. Hours written off, absorbed into an exhausted fixed fee, or lost to a retainer cap were billable by intent and worth nothing. An agency with healthy billable percentages and disappointing revenue is usually looking at this gap.
Read alongside: realisation — the proportion of billable hours that were actually invoiced. The gap between the two is where the money went.
4. Average project margin
What people read it as: how profitable our work is.
What it measures: the middle of a distribution that is usually bimodal, and therefore describes almost none of your actual projects.
An agency averaging 45% might have most work at 55% and two projects at −20%. The average suggests a broad efficiency problem. The reality is two specific projects, with a specific and probably identical cause, and the average has actively hidden it.
Margin is also flattering by default, because the costs most often left out — non-billable time on the account, unlogged account management, rework — all fall on the same side. See how to tell whether a project made money.
Read alongside: the sorted list. Sort by margin, look at the bottom five, and ask what they have in common. Ten minutes there beats an hour with the average.
The pattern
Every one of these fails the same way: it is an average that conceals a distribution, or a proxy that gets optimised once people know it is watched.
Two habits fix most of it.
Look at distributions before averages. The mean is the least informative number in any of these datasets. The outliers are where the decisions are.
Never make one of these a target for an individual. Goodhart's law is not a theory in this domain, it is a description of what happens in the second month. Set a billable-hours target and you will hit it — by logging generously and abandoning the unbillable work that keeps the agency running.
These numbers are diagnostics. They are good at telling you where to look and bad at telling you what to do, and the agencies that get value from them treat them as the beginning of a question rather than the answer to one.
