Share

Copy link LinkedInXEmail

A KPI is a performance indicator chosen because it tells whether a Goal is advancing, and measured by one or more Metrics. A number nobody would act on is not one.

What you can do with a KPI

• jobSpecify a reading
  Two teams report the same number differently. I need one definition: what it is made of, what establishes it, which way is better, in what form it reads and in what unit.
• tableMeasurement Plan · Measurement Model
• tableEvidence Plan · Evidence

Its KPIs— 42

• Conversion Flow Completion RateWhether the routes to conversion hold the people who start them. When it falls, volume can only be bought with more traffic.
• Assisted-Conversion ActionsWhether demand is reaching a person when it needs one. This is the part of conversion a self-serve flow cannot absorb.
• Lead TotalsWhether the top of the conversion path is being filled. It says nothing about whether those leads go on to convert.
• Conversion TotalsWhether demand is converting at all, in absolute terms. The count every rate below it is a share of.
• Conversion RateWhether the leads already won are being converted. It separates a demand problem from a conversion problem.
• Organic Visibility ShareWhether the brand is findable for the questions its market actually asks, without paying for the position.
• Site VisitsWhether being findable is producing arrivals at all. The crudest reading of discoverability, and the base every rate above depends on.
• Paid Search EfficiencyWhether bought visibility is worth what it costs. The counterweight to visibility won without paying.
• Page Experience (Core Web Vitals)Whether a page is good enough to be found and to be stayed on. Experience is a ranking input and an abandonment cause at once.
• Citation RateWhether assistants answering on the brand's subjects draw on the brand. Being the source is the newer form of being found.
• Mention RateWhether the brand appears at all where questions are now answered. Every later reading assumes this one.
• Share of VoiceWhether the brand holds its share of the conversation against the set it competes with.
• Video Answer AppearancesWhether the brand's own video is what a question returns, on the surfaces that answer with video.
• Schema CoverageWhether the brand states its own facts in a form a machine can read. What is not declared cannot be cited.
• Account Activation RateWhether people who begin a relationship finish starting it. An abandoned registration is an identity the model never gets.
• Attributed-Conversion ShareWhether outcomes can be connected to what produced them. It bounds how much of every efficiency reading can be believed.
• Referral ConversionWhether the sources that send people send people who act.
• Content Engagement RateWhether content is being used rather than merely served.
• Video Completion RateWhether video holds attention to the end, which is usually where its message is.
• Qualified Visit RateWhether arrivals are the right arrivals. Traffic that never qualifies is a targeting cost, not an engagement one.
• Email Engagement RateWhether a permissioned audience still wants to hear from the brand. It decays quietly, which is why it is read as an early warning.
• Fulfilment RateWhether what was agreed actually happened. A conversion that is never fulfilled costs more than one that was never made.
• Reactivation RateWhether lapsed relationships can be recovered, and how readily. The test of whether a relationship ended or only paused.
• Cost per LeadWhat it costs to raise a hand. Read beside conversion rate, or it rewards cheap leads that never convert.
• Cost per ConversionWhat it costs to produce an outcome — the efficiency reading that survives the lead stage.
• RevenueWhether marketing is producing money and not only activity. The one reading here in the unit the business plans in.
• Return on Ad SpendWhether what came back exceeds what went out. The single reading a budget conversation tends to end on.
• Revenue per LeadWhat a raised hand is worth, which is what decides how much a lead may cost.
• Revenue per ConversionWhat an outcome is worth, which is what decides how much an outcome may cost.
• Click-through RateWhether bought attention is being earned once it is placed. It separates a creative problem from a targeting one before either becomes expensive.
• Retention RateWhether relationships already won are being kept. Every acquisition number means less the lower this reads.
• Customer Lifetime ValueWhat a relationship is worth over the whole of its life, which is the number every acquisition cost has to be judged against.
• Advocacy RateWhether satisfied consumers are bringing others in. The one growth reading that does not rise with spend.
• ReachHow much of the addressable audience was touched at all, as distinct from how often it was touched.
• Visibility MomentumWhether presence is improving, holding or slipping. One period's reading cannot tell a gain from a model that was retrained underneath it.
• PositionWhether the brand is the answer or merely inside it. Appearing and being named first are not the same result.
• Hallucination RateWhether what assistants say about the brand is true. This is the exposure that visibility creates when it is wrong.
• Post-Citation Click-through RateWhether being cited sends anyone. The reading that separates visibility from traffic.
• Lapse RateWhether relationships already won are slipping away, read as the moves that leave rather than the ones that stay. Every retention number is the complement of this one.
• Likely-to-Lapse RateWhether the model is seeing trouble before the window confirms it. It is the reading retention outreach acts on, and the one that says whether acting early was worth it.
• Effective FrequencyWhether the people a placement reaches see it often enough to notice and not so often that further exposure buys nothing. Both ends are failures, which is why this reading has a band and not a direction.
• Opt-out RateWhether the people a channel reaches want to keep being reached. It is the relationship read from the side that leaves quietly, and the first reading a permission-based channel owes.

Why it matters

What goes wrong without it, as the commitments that license this kind say it, each standing on its sources.

• Declare measurement grain before calculationA rate means nothing unless everyone agrees what one instance of its numerator and its denominator represents. Marketing reporting fails less often because a number is wrong than because two numbers computed at different grains were divided — sessions over people, clicks over sends — and nothing in the output says so. Declaring the grain first is what makes the calculation reproducible by someone who did not write it.
  Declaring the Grain
• Campaign is the accountable initiativeMarketing organisations accumulate layers — programme, portfolio, initiative, workstream — each added by a team that wanted one more level of roll-up. Every layer is another place spend and outcome can be attributed, and after two or three of them it stops being answerable which level owns a number. One accountable unit that may nest inside itself gives the roll-up without the ambiguity.
  Understand Campaign Hierarchy
• Make outcome quality, calculation method and attribution inspectableAn outcome number is only as good as the inputs, exclusions and window that nobody wrote down. Media measurement standards exist because outcome claims circulate detached from their method, and a figure with no stated method cannot be challenged, reproduced or compared against anyone else's. Keeping what a measure means apart from how it is computed is what makes either one reviewable.
  Outcomes and Data Quality Standards · Declaring the Grain · Measuring Visibility in the AI Era · Data-driven multi-touch attribution models · Causally motivated attribution for online advertising · A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook · Why Three Exposures May Be Enough · Digital Audience-Based Measurement Standards
• A declared path must be walkableEvery other check asks whether a relationship is well formed: typed, bounded, licensed, inverted. None asks whether anything travels it. So the model could pass every check with a verb declared for thirty-eight kinds and asserted by none, and say nothing - which is what happened. A reader meeting that has no way to tell a deliberate absence from an oversight, and neither does the author six months on. The distinction is schema completeness against population completeness: the first was already enforced here and the second had no instrument at all. It matters most where a claim is a chain rather than a link. Performance asserted from delivery alone, with no step that reaches a person, is the specific failure measurement standards exist to prevent, and it is invisible to a check that only reads one edge at a time.
  Shapes Constraint Language (SHACL) · Outcomes and Data Quality Standards
• A move that cannot be counted is a pictureA journey that names its moves and cannot count them is a funnel with arrows: it can be drawn and not measured, which is the failure a state machine was adopted to end. Customer-migration models make the point exactly. A transition rate is the moves out of a condition over the people in it, and the people in it are the fact that put them there; so the events added to give every move a trigger are not extra instrumentation but the missing denominators of rates the model already carried and could not derive. A law that every move has a trigger, satisfied by facts nothing observes, would be the same picture with more arrows.
  Modeling customer relationships as Markov chains · Declaring the Grain · The Goal Question Metric Approach · Counting Your Customers the Easy Way: An Alternative to the Pareto/NBD Model
• A reading names its unit, and the unit fits its formA number without its unit is not a reading. Metrology defines a quantity value as a number and a reference together, and the unit codes exist because a value that travels without its unit is misread at the far end: a rate and a count that print the same figure have said different things, and percent is a name for the number 0.01, not a kind of thing. This model already fixed a KPI's direction and form as facts in the graph; the unit is the third fact of that family. Declaring which forms a unit fits is what a unit ontology does when it binds units to kinds of quantity, and it turns unit-form agreement from a review into a check.
  International Vocabulary of Metrology: Basic and General Concepts and Associated Terms (VIM), JCGM 200:2012 · ISO 80000-1:2022 Quantities and units, Part 1: General · The Unified Code for Units of Measure, version 2.2 · QUDT Ontologies: Quantities, Units, Dimensions and Data Types, release 2.1 · ISO 4217 Currency codes
• A target is a level on a reading, due within a windowGoal-setting research finds that specific, difficult goals produce higher performance than vague or absent ones, and that the effect depends on the goal being measurable and time-bound; the management literature made the same point as a mnemonic in 1981, and the frameworks that join strategy to metrics give each goal a magnitude and a timeframe. This model held goals as evergreen outcome areas and KPIs as readings, and left the level a campaign was actually committed to nowhere: a campaign could be measured by a KPI without ever saying what result would count as success. A Target makes that commitment a term of its own, joins it to the goal and the reading it belongs to, and lets the build refuse a campaign that promises a reading it does not take.
  Building a Practically Useful Theory of Goal Setting and Task Motivation: A 35-Year Odyssey · There's a S.M.A.R.T. way to write management's goals and objectives · Aligning Organizations Through Measurement: The GQM+Strategies Approach