Five social media scheduling key metrics defined by timezone, population and exclusions

Read each row as a separate evidence class: the first two measures describe delivery and monitoring, impressions are platform estimates, and the final pair concern organisational outcomes. Keep those categories apart even when a report shows them together. A platform count is not an England market statistic, and a tagged outcome does not establish that a post caused it.

Set boundaries before calculating

Use the rows as routing rules, not permission to combine unlike evidence. Before calculation, define the event, eligible population, numerator, denominator, unit, date range, timezone, exclusions, source system, export version, owner and uncertainty. If a count has no denominator, write not applicable rather than inventing one.

These fields apply to all five measures. Use Europe/London for reporting and identify any period that crosses a clock change; set start and end dates for each measure before working out the result. The table routes a number to its evidence class, while its event, population and exclusions determine whether the number is usable.

• Rule of thumb: Classify a measure as process, platform or outcome before comparing values, and define its measurement boundary before calculating.

What qualifies as a destination action rate?

Destination action rate records a tagged visitor completing an action defined in advance in the destination service. Include sessions arriving through the scheduled-post identifiers selected for the measure.

Divide valid completed actions within the outcome window by eligible tagged sessions, report a percentage, and use cohort entry dates with a cut-off in Europe/London.

Exclude staff tests, bots, duplicates and untagged traffic. Disclose cross-device loss, which limits coverage, and keep tracking choices visible. The measurement and reporting guidance covers the destination-side record. Online services can use link decoration to identify inbound traffic, for example from a particular email campaign.

If link decoration stores or accesses information on a device, have a qualified practitioner assess the actual technology and purpose. Do not widen collection just to improve a dashboard. Cross-device loss and tracking choices remain limits to disclose when interpreting the resulting rate.

• Rule of thumb: Count only eligible tagged sessions and disclose what tracking cannot capture.

What does an impression count represent?

Platform impressions are estimated counts based on the impression definition used by the platform, not a rate.

Record the exact posts, audience settings, organic or sponsored scope, definition in force on the export date, and account timezone recorded with the export. A numerator and denominator do not apply to this measure.

Keep the export with the relevant definition page and flag processing delays, omitted surfaces or later changes in the definition as uncertainty. Do not compare impression totals across platforms or treat them as evidence about an England population. Benchmark research explains why borrowed benchmarks can mislead.

• Rule of thumb: Present platform impressions as platform-specific estimates, not comparable population counts.

When is credited contribution a ratio?

Attributed contribution ratio records an outcome approved by finance and credited under a named attribution rule. Use recognised outcomes credited in the period as the population.

Divide credited observed contribution after stated reversals, in pounds, by full economic cost for the same scope, also in pounds; use the finance period and cut-off.

Keep taxes, transfers and costs outside the agreed scope excluded. The ratio is credited observed contribution divided by full economic cost. Monitoring that an outcome changed may describe a change, but does not prove an intervention caused it, as explained in quality in policy impact evaluation.

Describe the result as model-assigned contribution, never causal return. The attribution methods explain the rules, and return on investment addresses decision thresholds based on your own objective and evidence. Attribution assigns contribution under a rule; it does not establish causation.

• Rule of thumb: A named attribution rule assigns contribution, but does not prove causation.

What counts as a completed monitoring record?

Monitoring completion is a process measure: it records whether an assigned monitoring record closes with evidence, including any unresolved issues. Use released items assigned a monitoring window as the population; divide complete closures by all eligible released items and report a percentage. The window starts and ends in Europe/London.

Leave out cancelled items that never went live. If access is lost or logs are unavailable, mark the record missing instead of assuming completion. This measure records completion of the operating process, not audience response. Pair it with the quality checklist so each failure has an owner.

• Rule of thumb: Missing monitoring evidence stays missing, not complete.

How is a due-item match established?

Count an item as a match only when the live post agrees with its approved revision, account, audience and scheduled time.

The eligible population is every item due in the period; exact matches are the numerator, and all due items, including failures, are the denominator. Express the result as a percentage.

Exclude only items formally withdrawn before their due time, and show withdrawn items separately. Use the approval record, scheduler log and live capture as sources. Disclose missing captures and timestamp gaps rather than recording a confirmed match. The due-time boundary is Europe/London, and show the result alongside both counts.

How do clock changes affect reporting dates?

Use Europe/London start, end and cut-off times for every measure. Check and state whether the reporting period crosses a clock change, and apply the same care to outcome windows so an event is not assigned to the wrong period.

The clocks change twice a year: forward one hour at 1am on the last Sunday in March, and back one hour at 2am on the last Sunday in October.

The period when clocks are one hour ahead is called British Summer Time. A reporting week that contains a clock change lasts 167 or 169 hours rather than 168, so note the change in the report. Outcome-window cut-offs need equal care: an event can otherwise fall into two reporting periods. See measurement mistakes for common errors.