Operations
Part of Reputation management operations: from source record to publication
Reputation management service standards: targets from a log
Reputation management service standards for England: define each clock, measure the median and 90th percentile from your own log, then set a target.
A response-time target is only as honest as the case log behind it. The sound order is to define the event, measure the current operation, then authorise a target the team can support without weakening consumer, privacy or publication controls.
A target set the other way round, from an attractive promise or an industry average, measures nothing the team actually does. How a case moves through its states is a separate question, covered in the operating workflow. Here the case log supplies the numbers.
This is a working method, not an approved service standard. Consumer-law and data protection review should sign off any target before the team adopts it.
What to take away
- Measure the current operation first, then authorise targets it can support.
- Fix the definition of each clock before collecting a single time.
- Use your own median and 90th percentile over a defined period, never an industry average.
- Report a distribution and the unresolved queue, not a lone average.
- Keep quality gates separate from speed, so a fast case still fails if a gate fails.
- Leave target and tolerance blank until workload evidence and specialist review exist.
Fix the definition before measuring
A time means nothing until the event, the clock and the owner are fixed. Fill these seven fields before measuring, and name the review sources in scope, such as Google, Trustpilot or Facebook. Verify the product and account configuration against each one before promising coverage.
Narrow Service Definition Fields
- Eligible public sources
- Eligible item and duplicate rule
- Operating window and Europe/London calendar
- Public response authority
- Complaint and urgent-risk hand-offs
- Excluded work and unsupported channels
- Source outage and supplier failure treatment
- Eligible public sources.
- Eligible item and duplicate rule.
- Operating window and Europe/London calendar.
- Public response authority.
- Complaint and urgent-risk hand-offs.
- Excluded work and unsupported channels.
- Source outage and supplier failure treatment.
State whether the organisation merely replies on an external platform or publishes reviews itself. The distinction changes the controls the team must configure, and so which events can honestly start a clock.
Teams can assess reputation-management software such as Birdeye, Reputation.com and Yext against the same clocks. This is not a ranking or a claim that each product supports every workflow.
Define each clock in full
Each clock needs nine entries: start event, stop event, valid pause, excluded event, timezone, system of record, owner, target and tolerance. The first seven can be filled in today. Target and tolerance stay blank until workload evidence exists.
| Clock entry | Question to settle | Trap if left open |
|---|---|---|
| Start event | Which moment begins the wait? | An alert receipt is counted as source availability |
| Stop event | What counts as done? | A draft is counted as a reply |
| Valid pause | Which waits stop the clock? | Every difficult item gets paused |
| Excluded event | Which items never enter? | Duplicates inflate volume |
| Timezone | Which calendar applies? | Mixed zones shift the results |
| System of record | Where is the time kept? | Two logs disagree |
| Owner | Who answers for a miss? | Nobody acts on a long wait |
An alert receipt may not equal source availability. A draft awaiting legal advice is not the same state as an untouched case, so name it as a valid pause. Pausing every difficult item would make the reported time meaningless, so list the permitted pauses and audit their use.
Name the role that owns each clock, with the authority to act when it runs long.
Set the target from your own log
Set the target from the team's own case log, not from an industry average. The evidence basis is the measured median and 90th percentile over a defined period. Both come from cases the team actually handled, under its own controls.
As an illustration only, a completed clock might read: first acknowledgement within 4 working hours, plus or minus 2 working hours. Set your own values from your own evidence.
- Choose a defined period and write down its start and end.
- List every eligible case in the period, including those still open.
- Calculate the elapsed time for each closed case using the clock definition.
- Find the median, the time within which half the cases were finished, and the 90th percentile, the time within which nine in ten were finished.
- Record the missing cases and the reason each is missing.
- Pass the figures to specialist review, then set target and tolerance.
The median describes the typical case, while the 90th percentile describes the slow end. Together they show whether a target would be met by most cases or only by the easy ones. A tolerance is the margin around the target within which a case still counts as on time, and it should come from the spread you observed.
Keep the target and tolerance blank until workload evidence and specialist review exist. A target written before that point is a promise, not a standard.
Report a distribution, not a lone average
Report the distribution and the unresolved queue rather than a lone average. An average hides the slow tail, and a closed-case figure leaves out the cases still waiting. Show both, with the number of cases behind each, so the next period can be compared like for like.
Report a distribution with channel, risk route and case type visible where volumes permit responsible reporting. Do not publish a percentage with no denominator, eligibility rule, period or missing-data note. A figure without those four notes cannot be checked.
The GOV.UK Service Standard, written for public services, asks teams to define what success looks like and publish performance data. A private organisation cannot imply that GOV.UK assessed or endorsed its service, but it can borrow the discipline and say what success means before it reports any figure.
Keep quality gates apart from speed
A case cannot earn release by being fast. Require separate results for source preservation, routing, evidence, review integrity, personal-data handling, claim clearance, destination approval and correction readiness. Each result is PASS, FAIL or UNRESOLVED.
Speed targets sit beside the gates, never above them. The CMA's fake reviews guidance explains how businesses should comply with consumer protection law on consumer reviews, and no response clock excuses a case that fails that check.
The Digital Markets, Competition and Consumers Act 2024 provides the current UK statutory framework for fake reviews. Set a zero-tolerance control only after qualified consumer-law review defines the covered events and escalation path.
For personal information in case files or responses, set the control against the UK GDPR's data minimisation principle and the Data Protection Act 2018. The ICO's guidance supports limiting data to what the purpose requires, and the local standard should name who decides and how a challenge is recorded.
Define service failures before launch: missed eligible item, unsupported claim, unauthorised disclosure, wrong-account publication, unrecorded incentive, inaccessible internal task, lost correction link, expired approval or unavailable evidence. Give each a containment owner, notification route, recovery record and restart authority.
Rehearsing the failures in risk scenarios shows whether the containment owner can act before a clock expires.
Before you act
- Complete all seven fields of the service definition.
- Record every clock with its start, stop, pause, exclusion, timezone, system of record and owner.
- Measure the median and 90th percentile from your own log over a defined period.
- Report the distribution and the unresolved queue with a denominator and a missing-data note.
- Record PASS, FAIL or UNRESOLVED for each quality gate.
- Define service failures, containment owners and recovery records before launch.
Common questions
What should a team measure before authorising targets?
Its current operation, over a defined period and from its own case log. Targets follow only once the team knows what it can support without weakening consumer, privacy or publication controls.
Why is a lone average response time not enough?
It hides the slow end of the work and leaves out cases still open. Report a distribution and the unresolved queue, and never publish a percentage without a denominator, eligibility rule, period or missing-data note.
Can a buyer copy a supplier's uptime figure into its own standard?
No. If a monitoring or publishing supplier is involved, do not copy its marketing uptime into the buyer's standard. Ask for scoped evidence and record what the buyer verified.
