Reputation management return on investment: a measurement protocol

No return can be calculated for fictional Harbour Lane Furniture Repairs Ltd. There is no intervention, baseline outcome, counterfactual, cost record or observed financial contribution. This page supplies a decision protocol for reputation management return on investment with blank variables. It does not turn ratings, mentions or response counts into money.

What to take away

State the causal question

Write one question: [What incremental change in a defined eligible outcome was caused by the specified review-response intervention, compared with what would otherwise have happened, during the stated period?]

Name the population, England geography rule, start and end dates, intervention version, comparison condition, exclusions and decision owner. If the question concerns complaint resolution, do not quietly substitute sales. If it concerns a public response, do not include review invitations or advertising unless they are separately specified.

Preserve four evidence layers

First, operational records show what the team did: eligible items captured, responses approved, complaints handled and corrections completed. These records normally sit in the review platform or reputation-management software itself, such as Google Business Profile or Trustpilot. Second, that supplier or platform may credit later events under its own rule. Third, finance can reconcile actual orders, refunds, cancellations, costs and contribution for an eligible population. Fourth, a credible evaluation may estimate the incremental difference against a counterfactual.

These layers are not interchangeable. A sale after a response is temporal sequence. A platform label is attribution. Neither establishes causation.

HM Treasury's Magenta Book, updated on 15 May 2026, covers evaluation scoping, design, conduct, use and interpretation. Although written for government, it is an authoritative method source for distinguishing monitoring from evaluation. Under the Magenta Book, sign-off for the design must come from a qualified evaluator, such as a Government Social Research analyst or an independent methodologist.

Fill the input register

For each outcome and cost, record [definition] [eligible population] [event] [numerator] [denominator] [currency and VAT] [period] [source] [owner] [uncertainty]. The source field names the record and where it lives: platform APIs or exports from Google Business Profile or Trustpilot, the CRM, the finance system, the complaints queue and supplier invoices.

Full economic cost includes supplier cash — licence seats, review-volume tiers, onboarding, support, migration and exit — plus staff effort, collection and monitoring, response work, complaint handling, implementation, legal and privacy review, accessibility, security, incident preparation and renewal. GOV.UK's invoice-content guidance supports checking the service, supply date, amount and VAT information. It does not establish economic cost or recoverability.

Choose a defensible comparison

The counterfactual might require a randomised design, a credible phased introduction or another justified quasi-experimental approach. Before-after change alone is vulnerable to seasonality, product quality, service failures, advertising, platform changes and wider events. Predeclare assumptions, contamination, spillovers and stopping criteria. If the available data cannot support a counterfactual, report the limitation and stop at descriptive evidence.

Only after qualified review may the buyer populate:

incremental net contribution = observed eligible net contribution minus counterfactual eligible net contribution

ROI = (incremental net contribution minus full economic cost) divided by full economic cost

Both expressions remain blank. Finance defines contribution after relevant refunds, variable costs and accounting treatment. Tax treatment requires separate advice.

Hypothetical illustration: if observed eligible net contribution is £120,000 and the counterfactual is £100,000, then incremental net contribution is £20,000. With full economic cost of £15,000, ROI is 33 per cent ((£20,000 - £15,000) / £15,000). These figures are illustrative only.

Run sensitivity around buyer-approved uncertain inputs and show where the sign or decision changes. Do not publish a point estimate without uncertainty, attrition and missing-data limits.

Set a rule before seeing results

Record the minimum evidence quality required to continue, the authorised review date and the person who can stop the intervention. Leave the numerical threshold blank. Predeclare how missing cases, changed platform definitions, concurrent advertising, service failures and corrections will be treated. If those factors destroy comparability, do not repair the result by choosing a more favourable window.

Build a sensitivity table with assumptions down the rows and buyer-approved alternatives across the columns. Show the decision under each combination, including an outcome where no incremental effect can be established. A model that changes conclusion under a small defensible assumption change needs more evidence or a smaller decision, not a confident headline.

Keep non-financial gates outside ROI. Consumer-review integrity, privacy, the Privacy and Electronic Communications Regulations 2003 (PECR), enforced by the Information Commissioner's Office, employment, accessibility, security and legal failures cannot be traded for a positive modelled return. The intervention stops when an independent gate fails even if the financial rule would otherwise pass. The verdict here is HOLD: operational measurement may proceed, but a financial return for this work is not established.

Before you act

Common questions

Why are the formulas left blank for Harbour Lane Furniture Repairs Ltd?

The company is fictional, so the page demonstrates the protocol with blank variables rather than producing a financial result.

What is the difference between the four evidence layers?

Operational records show what the team did, a supplier or platform may credit later events under its own rule, finance reconciles actual orders and costs, and a credible evaluation estimates the incremental difference against a counterfactual.

Why is before-and-after change alone not enough to show causation?

It is vulnerable to seasonality, product quality, service failures, advertising, platform changes and wider events. The text says to predeclare assumptions, contamination, spillovers and stopping criteria, and if the data cannot support a counterfactual, report the limitation and stop at descriptive evidence.