Measurement
Part of Accounting services measurement: a practical guide
Choosing accounting service metrics that actually lead to a decision
Measure accounting services in England with defined volume, timeliness, quality, effort, experience and outcome metrics that lead to decisions.
Measure an accounting service as a chain, not a single percentage. Volume, timeliness, quality, client effort, experience and business outcome answer different questions.
What to take away
- Measure an accounting service as a chain covering volume, timeliness, quality, client effort, experience and business outcome.
- A compact metric set with clear decision rules beats a crowded dashboard nobody acts upon.
- Write a definition card for every metric covering purpose, numerator, denominator, owner and action threshold.
- Retire any metric that no longer guides a decision instead of keeping a green board.
Begin with the service objective
State what the service must achieve for a defined customer and period. Examples include producing reviewed records for decisions, paying employees correctly, or delivering an authorised submission with recoverable evidence. Avoid an objective such as "provide excellent accounts", which cannot be tested.
The Government Digital Service's service-standard guidance is written for government services. Its advice to define success, select suitable metrics and combine performance data with user research remains a helpful design principle, not a binding rule for private accounting practices.
Select a balanced metric set
For a small practice, five to seven metrics is a typical working range, chosen from the groups that matter for the service.
Balanced metric groups
Metric group
- Demand
- New work by risk
- Flow
- Work in progress ageing
- Timeliness
- Completed by cut-off
- Quality
- Files corrected after review
- Effort
- Staff and client time
- Experience
- Query resolved no repeat
- Outcome
- Decisions or obligations completed
- Resilience
- Recovery exercises met objective
Example measure
- Demand
- Flow
- Timeliness
- Quality
- Effort
- Experience
- Outcome
- Resilience
| Group | Example |
|---|---|
| Demand | New work received by service and risk category |
| Flow | Work in progress and ageing at each control point |
| Timeliness | Proportion completed by the defined internal cut-off |
| Quality | Files corrected after review or submission, by consequence |
| Effort | Staff and client time for a specified output |
| Experience | Query resolved without repeat contact |
| Outcome | Decisions or obligations completed using the output |
| Resilience | Recovery exercises meeting their stated objective |
Do not reward fast closure alone. Pair response time with usefulness, and completion with post-completion correction. Show missed statutory deadlines separately rather than letting a large volume of routine work dilute them.
Write a definition card
For every metric, record name, purpose, numerator, denominator, population, exclusions, time zone, source, frequency, owner, quality check and action threshold. Include examples of records that do and do not count.
Metric definition card fields
- Name and purpose
- Numerator and denominator
- Population and exclusions
- Time zone and source
- Frequency and owner
- Quality check and action threshold
- Examples that count and do not count
For a private company registered in England, Companies House normally requires later accounts 9 months after the accounting reference date. Use that deadline in the worked card below.
Field / Worked example: on-time accounts completion
- Purpose
- Show whether accounts reach Companies House before the filing deadline
- Population
- England-registered companies with a Companies House accounts filing date in the reporting month
- Numerator
- Accounts filed with Companies House at least 5 working days before the filing deadline
- Denominator
- All England-registered companies in the population with that filing deadline
- Exclusions
- Companies with a Companies House extension or a pause agreed in writing
- Time zone
- UK local time
- Source
- Accounts production tracker and Companies House filing receipt
- Frequency
- Monthly
- Owner
- Accounts production lead
- Quality check
- Reconcile the tracker count to filing receipts before sign-off
- Threshold
- Green at 95 per cent or above; amber 90 to 94.9 per cent; red below 90 per cent
- Action
- Red triggers a same-week review of the three latest late files and a named recovery plan
If "on time" means before an internal date, do not present it as statutory timeliness. If paused client time is removed, state the pause rules and display total elapsed time as a companion measure.
Preserve evidence and privacy
Link aggregated results to controlled source records so a reviewer can reproduce a sample. Log corrections to the metric itself. Freeze monthly data after review and show later restatements.
Collect only the personal information needed for the decision. Restrict client and staff detail in dashboards, apply retention, and assess small groups that could identify individuals. The ICO's data minimisation guidance explains the UK GDPR principle of keeping personal data adequate, relevant and limited to what is necessary.
Segment without distorting
Break results down by service, customer type, risk, channel, team and record condition where sample size supports it. Averages can hide a difficult queue or an accessibility problem. Display counts beside percentages and avoid ranking tiny groups.
Connect measures to action
Set a named response before reporting begins. Rising unresolved queries may trigger a root-cause sample; repeated review corrections may change training or templates; late client records may require different cut-offs and communication.
Worked scenario: on-time accounts completion falls to 87 per cent in May. The accounts production lead runs the late-file root-cause sample. The sample shows two files waited on client bank records, so the client manager moves those two clients to the 20th-of-month records chase. The panel logs the action and reviews the metric in June.
Create a small metric-change panel involving service, quality and data owners. Require it to approve new calculations, thresholds and retired measures. Before a definition changes, estimate how the old and new versions affect the trend. Run both briefly when feasible, mark the break and preserve the prior logic.
Audit for gaming as well as arithmetic. Interview users about tasks closed and reopened, pauses, reclassification and unrecorded work. A measure can calculate exactly while describing the service poorly.
Review definitions when software, scope, staff or deadlines change. Retire a metric that no longer guides a decision. The aim is not a green board. It is timely evidence showing where accounting work is reliable, where it is failing and who will respond.
Before you act
- State what the service must achieve for a defined customer and period.
- Run a realistic number of metrics, typically five to seven for a small practice.
- Record name, purpose, numerator, denominator, owner and action threshold for every metric.
- Link aggregated results to controlled source records so a reviewer can reproduce a sample.
- Set a named response before reporting begins.
- Review definitions when software, scope, staff or deadlines change.
Common questions
How should an accounting service begin when choosing metrics?
Begin with the service objective. State what the service must achieve for a defined customer and period, such as producing reviewed records for decisions or paying employees correctly. Avoid an objective such as provide excellent accounts, which cannot be tested.
Why is a balanced metric set better than one percentage?
Volume, timeliness, quality, client effort, experience and business outcome answer different questions, so one percentage cannot cover the service. Show missed statutory deadlines separately.
What should happen before a metric definition changes?
Estimate how the old and new versions affect the trend. Run both briefly when feasible, mark the break and preserve the prior logic. A small metric-change panel involving service, quality and data owners should approve new calculations, thresholds and retired measures.
