Guides
Translation agency KPIs owners actually use
Eight monthly measures for a translation agency, each with a definition and a trigger, including decline rate and bench depth as the early warnings.
What to take away
- Track eight numbers monthly, each with a written definition, a data source, and an action it triggers.
- Write the definition before the first month. A measure whose formula changes halfway through is a measure you cannot compare.
- Two of the eight are about people you do not employ. Those are the early warnings.
- If a measure has produced no action in six months, delete it.
The eight
1. New work won, by source
Definition: projects accepted this month, grouped by where the inquiry came from: referral, search, intermediary, existing client, tender.
The eight monthly measures
- New work won, by source
- Quote conversion, by buyer type
- Contribution per internal hour, by line
- Days from delivery to cash, by client
- On-time delivery, with cause
- Decline rate (early warning)
- Bench depth (early warning)
- Spoken and written work counted separately
Action it triggers: a source producing nothing for three months either gets attention or gets dropped from your effort.
Illustrative trigger: zero wins from a source for three months, or a fall to half its own three-month average.
2. Quote conversion, by buyer type
Definition: quotes accepted divided by quotes issued, split by content owner, procurement, intermediary, and individual.
Action: a falling rate in one type is a positioning or scope problem in that type, not a general price problem. Do not respond to it with an across-the-board discount.
Illustrative trigger: a fall of two percentage points or more from that type's three-month average.
3. Contribution per internal hour, by line
Definition: revenue less external cost, divided by your own recorded hours, calculated per service line.
Action: two consecutive months below target sends the line's process to review, starting with where the hours actually go.
Illustrative trigger: two months below target, or a fall of ten percent from the line's three-month average.
4. Days from delivery to cash, by client
Definition: delivery date to payment received, averaged per client.
Action: a client drifting past agreed terms twice gets a conversation about terms, not a third reminder.
Illustrative trigger: average days more than five days past agreed terms, or a rise of three days from the client's prior quarter.
5. On-time delivery, with cause
Definition: deliveries meeting the committed date, with every miss classified as ours, the client's, or external.
Action: any miss with no recorded cause is treated as a process failure in its own right, because an unexplained miss cannot be prevented.
Illustrative trigger: any month below 95 percent, or two points below your own three-month average.
6. Reviewer changes per project, split by type
Definition: changes made by the client's reviewer, counted as corrections or preferences.
Action: corrections rising two periods running means briefing or linguist selection. Preferences failing to fall means terminology is not being captured.
Illustrative trigger: corrections up in two consecutive months, or preferences above one per project.
7. Assignment decline rate, by pair
Definition: assignments declined by linguists divided by assignments offered, per pair.
Action: a rising rate means your rate, your deadlines, or your briefs have fallen behind what that pair's market expects. All three are fixable and none is visible anywhere else.
Illustrative trigger: a rise of two percentage points in a month, or three months above the pair's prior average.
8. Bench depth, by active cell
Definition: the number of tested, available linguists per pair and subject you currently sell.
Action: any active cell at one name gets recruitment started this month, before a client is affected.
Illustrative trigger: any active cell at one name, or any cell at two names with a client deadline inside 30 days.
Why the last two matter most
Six of the eight describe what already happened. Numbers seven and eight describe what is about to happen.
Lagging vs leading measures
Lagging (six)
- What they show
- Already happened
- Examples
- Revenue, on-time delivery
- Risk if ignored
- Hidden until account lost
- Fix mechanics
- Process review
Leading (two)
- What they show
- About to happen
- Examples
- Decline rate, bench depth
- Risk if ignored
- Missed deadline waiting
- Fix mechanics
- Recruit, test, tier bench
A pair with rising declines and depth at one person is a missed deadline waiting for a calendar clash. It stays hidden in revenue, on-time delivery, and client satisfaction until it costs you an account.
The fix mechanics are in how the bench is recruited, tested, and tiered, and what it costs to fix is in the labor cost stack.
Rules for the sheet
| Rule | Reason |
|---|---|
| One page, one row per measure. | A longer sheet gets skimmed or abandoned. |
| Every row has a definition, a source, an owner, and a trigger. | A number with no source or decision point cannot be acted on. |
| Update on the same date each month. | A moving date breaks month-to-month comparison. |
| Keep the raw counts beside the percentages. | Small bases can make one project look like a trend. |
| Retire any measure that produces no action in six months. | The sheet stays short and every row earns its space. |
Rules for the KPI sheet
- One page, eight rows, this month and last three
- Definition and data source recorded once
- Every row carries its trigger
- Spoken and written work counted separately
- No industry comparison
The operations lead owns the sheet and updates it on the same date each month. Each row has a named owner who can act on its trigger. Review the sheet quarterly and retire any measure that has produced no action in six months. A measure that survives six months without a decision is removed at that review.
Where the data has to come from
All eight need project-level records kept as the work happens: internal hours, subcontractor invoices, query and change logs, and payment dates. Reconstructing them at month end produces numbers nobody trusts. Keeping them properly is also the general expectation described in the IRS guidance on records a business should keep.
Wherever those records live, they hold client-confidential material. The same access decisions apply as to the content itself.
The NIST small business quick-start guides and the CISA material for small and medium organizations are reasonable starting points for setting a baseline. What a particular client requires is a question for their compliance function, and what applies to you is one for a qualified adviser.
Where this connects
The measurement model and the tool stack that produces it are set out in the software and measurement guide. Numbers five and six come out of the process in onboarding and delivery. Numbers one and two are the output of the channels described in the marketing and growth guide.
Common questions
Eight seems like a lot for a small agency.
Six of them come from records you already keep. The two that need new discipline are internal hours and reviewer change classification, and both take minutes per project.
Should I share these with my team?
Share the ones they influence, with the trigger attached. A coordinator who knows what on-time delivery triggers will record causes properly, and that is most of the value.
What if a number looks bad because of one project?
Note it beside the row and move on. Excluding an outlier is fine when it is labeled and permanent exclusion is how a sheet becomes fiction.
How do I set a target with no benchmark?
Take your trailing three-month average and add one deliberate step. If on-time delivery averaged 95 percent across those months, set the next target at 96 percent. That target comes from your own record and can be explained to the team.
