Five CRM measurement mistakes, from a moving denominator to a borrowed benchmark

Small business CRM measurement mistakes usually begin before a chart is drawn. The team has not defined the eligible work, has changed the measurement rule, or is asking an observational report to prove causation.

Method: Relationship Brief selected five errors that would change interpretation or undermine a decision, then matched each to a current official method or regulator source.

Date: Research was completed on 6 September 2026.

Scope: Product-neutral CRM reporting for an ordinary small business operating in England, with UK-wide data-protection and PECR boundaries where relevant.

Exclusions: No supplier dashboard, dataset, customer result, benchmark, price, conversion rate, financial return or legal compliance was assessed.

Non-ranking: The list follows the reporting chain. It is not ordered by prevalence or harm. There was no supplier funding, affiliate relationship or hands-on testing.

What to take away

1. Letting the denominator move

A progression rate becomes easier to improve if unassigned, unresolved or failed records quietly disappear. Define eligibility before calculating the result. Preserve the population, numerator, denominator, period, exclusions and correction policy with the query version.

The UK Statistics Authority's Standards for Official Statistics require official producers to use suitable data and sound methods and to explain uncertainty. A private business is not claiming Code compliance, but those principles support a repair: publish the complete metric card and counts behind the rate.

2. Treating missing tracking as no behaviour

No recorded email open does not prove that a person did not read the message. Pixels, redirects, devices, settings and consent choices affect the event available to the CRM.

The ICO's final storage and access technologies guidance was updated on 29 April 2026 and covers PECR plus relevant UK GDPR issues. Record the tracking mechanism, permission or exception analysis and missingness. Do not compensate for gaps by assuming an unobserved event.

3. Joining incompatible periods

A new workflow, field definition, integration or query can create a break that looks like performance change. Keep a dated method history and annotate the series. If the old and new definitions cannot be reconciled, begin a new series rather than drawing one continuous line.

The Government Analysis Function's quality and change guidance tells official-statistics producers to explain adjustments, revisions and method changes. It also says the advice may help other authors. The CRM repair is to preserve both versions and state the effect on comparability.

4. Calling attribution causation

A last-touch rule assigns credit to a recorded event. A before-and-after chart records association. Neither creates the unobserved outcome that would have occurred without the CRM change.

HM Treasury's May 2026 Quality in Policy Impact Evaluation framework says outcome monitoring alone does not establish that an intervention caused change and explains the role of a counterfactual. It is policy guidance, not a private-company instruction. The repair is still clear: label credit rules and observations honestly, and involve an evaluator before making a causal claim.

5. Borrowing an unrelated benchmark

A global vendor average, a customer case study or a UK software-adoption survey does not establish an England small-business CRM norm. Population, geography, period, event, numerator, denominator and collection method must match before the comparison is meaningful.

ONS survey uncertainty guidance explains how sample estimates can differ from population values through sampling and non-sampling errors. An internal CRM dataset has different issues, but the source supports investigating coverage, response and processing bias. Withhold the external benchmark when comparability or uncertainty cannot be defended.

Fix one mistake at the source before redesigning the dashboard. Rebuild the affected result, document whether the decision changes and publish the correction. A clean chart cannot rescue a weak definition.

Before you act

Common questions

What is the first step to avoid a moving denominator in CRM measurement?

Define eligibility before calculating the result. Preserve the population, numerator, denominator, period, exclusions and correction policy with the query version. The UK Statistics Authority's Standards for Official Statistics support publishing the complete metric card and counts behind the rate.

Why is a missing email open not proof of no behaviour?

Pixels, redirects, devices, settings and consent choices affect the event available to the CRM. No recorded open does not prove a person did not read the message. Record the tracking mechanism, permission or exception analysis and missingness, and do not assume an unobserved event.

When should an external benchmark be withheld from a CRM report?

Withhold the external benchmark when comparability or uncertainty cannot be defended. Population, geography, period, event, numerator, denominator and collection method must match before comparison is meaningful. ONS survey uncertainty guidance supports investigating coverage, response and processing bias.