Grant measurement mistakes, from missing baselines to claiming causation from timing

The most damaging measurement mistakes happen before the report is written. This list covers ten design and reporting errors found through review of current UK government evaluation guidance. It does not rank grant providers or claim a measured prevalence across English businesses.

What to take away

• Measurement mistakes often start before reporting, so define baselines and metrics early.
• Activity and spend are not outcomes; build a results chain to show real change.
• Timing alone does not prove causation; use comparison evidence and cautious contribution language.
• Correct errors by finding the control failure and tracking whether the remedy works.
• Report missing data and adverse findings honestly to support learning and fair review.

1. Starting without a baseline

Without a defined starting value, the team cannot measure change reliably. Capture the period, source, filters and unusual conditions before delivery begins.

2. Treating activity as outcome

Money spent, equipment bought and sessions delivered are not proof of improved productivity, skills or sales. Build a results chain from inputs and activities to outputs and outcomes.

The 2026 Magenta Book separates process, impact and value-for-money questions. Use that distinction to label evidence accurately.

3. Leaving metrics undefined

"Jobs", "businesses supported" and "growth" can have several meanings. Write the unit, inclusion rule, period, geography, source and owner before counting.

4. Double counting

One participant may attend several sessions or appear in partner records. Decide whether the measure counts unique people, attendances or completions. Test duplicates without exposing unnecessary personal data.

5. Changing the target after results

Keep the approved target and original method. If circumstances justify a new target, record the approval and show both. Rewriting history prevents a fair variance review.

6. Claiming causation from timing

An outcome improving after funding does not prove that the project caused it. The government's Quality in Policy Impact Evaluation guidance says outcome monitoring alone cannot establish attribution.

Consider comparison evidence and alternative explanations. Use cautious contribution language when the design cannot estimate causal impact.

7. Ignoring missing data

Report the eligible population, responses, missing values and reasons where known. A high satisfaction percentage from a small, self-selected group may not describe everyone reached.

Set follow-up and validation rules before analysing results. Do not remove inconvenient records without an approved reason.

8. Mixing cost scopes

Do not divide a full-year project cost by one quarter's outputs or compare a net cost with a gross benchmark. Align period, activities, VAT treatment, population and accounting basis.

The Cabinet Office Model Grant Funding Agreement illustrates how financial reporting and performance measures may both form part of an award. Reconcile them to the same baseline.

9. Collecting data without a use or lawful basis

Every field creates work and may create privacy risk. Collect only what supports an agreed decision or duty. The ICO guidance hub is the primary starting point for UK data-protection guidance.

10. Hiding weak or adverse findings

Selective reporting prevents learning and may mislead the funder. Present limitations, unexpected effects and underperformance with evidence and corrective action.

Apply a correction routine

Give every issued figure a version, reporting date and owner. If an error is found, identify all reports, dashboards and decisions that used it. Correct the source, repeat the calculation, obtain review and record the reason and date.

Do not erase the earlier report or describe a material change as a routine refresh. Tell the funder through the required route when an issued claim or result was wrong. The agreement and circumstances determine whether a formal notification is needed.

After correction, find the control failure. A wrong formula, ambiguous definition, late source or unauthorised edit calls for a different remedy. Track whether the remedy works during the next reporting cycle.

Add a measure specification, evidence register and independent calculation check to the reporting process. This draft contains no live internal links and requires methodological and agreement-specific review before publication.

Before you act

• Define a baseline before delivery begins.
• Distinguish process, impact and value-for-money questions.
• Write unit, inclusion rule, period, geography, source and owner.
• Test for duplicates without exposing personal data.
• Keep the approved target and original method.
• Collect only data with a use or lawful basis.

Common questions

Why is a baseline necessary before measuring grant outcomes?

Without a defined starting value, the team cannot measure change reliably. The article advises capturing the period, source, filters and unusual conditions before delivery begins. This ensures that any later comparison reflects real change rather than guesswork or shifting definitions.

How can a grant provider avoid double counting participants?

Decide whether the measure counts unique people, attendances or completions. Test duplicates without exposing unnecessary personal data. This prevents inflated figures and ensures that reported numbers reflect distinct individuals or events accurately.

What should be done when an error is found in an issued figure?

Identify all reports, dashboards and decisions that used it. Correct the source, repeat the calculation, obtain review and record the reason and date. Do not erase the earlier report or describe a material change as a routine refresh. Tell the funder through the required route.