KPI Dictionary Template: Get Teams Using the Same Numbers

What belongs in a KPI dictionary template?

Preql’s Finance KPI Dictionary Template describes a KPI dictionary as a shared metric list with definitions, calculations, sources, and owners. Start with the recommended fields below. Add timing, denominator, and exclusion rules wherever two analysts could interpret the same label differently. The dictionary defines the number; it does not display the current result.

Keep one row per metric in the main sheet, with an expanded specification sheet for calculations that need more detail. Your operational dashboards show results. The dictionary must let someone explain those results without tracking down the analyst who built the report.

Copy these four completed entries as a starting point

These examples are starting definitions, not universal standards or targets, following StartLab’s guidance on adapting KPI examples. Replace the illustrative source-field labels with actual system fields. Confirm the proposed owner roles and refresh schedules with the people doing the work, then enter your revision date and approval reference. Do not publish the examples unchanged.

For these ratios, return “not applicable” when the denominator is zero. Silently showing zero percent tells a different story. Before adopting the entries, settle how reopened opportunities, cross-device users, accounting adjustments, and rescheduled tasks affect the calculation.

Which extra fields keep calculations consistent?

Expand an entry wherever the analyst implementing it would otherwise have to guess. Record the decision supported, unit, counting level, reporting window, date basis, inclusions, exclusions, latency, interpretation, and effective date. A readable description is not enough. Different teams need instructions that produce the same result from the same records.

Bernard Marr’s A Sample KPI Template recommends connecting indicators to strategic objectives and documenting both intended and prohibited uses. Make that restriction explicit. A task deadline metric can inform capacity planning without becoming an individual performance ranking.

Separate refresh cadence, latency, and reporting dates

Refresh cadence tells you when the report recalculates. Data latency tells you how long underlying events take to become available. In a hypothetical setup, a report refreshes every morning but uses accounting data finalized only after monthly close. The timestamp is fresh. The inputs are not.

A reporting window is different again: “October” names the period, while “closed_date” decides which opportunities belong in it. An opportunity created in September and closed in October falls into different groups depending on the chosen field. Use your daily versus weekly reporting rules to set the review schedule, not as a substitute for defining the period and date basis.

Finish the win-rate definition before building the chart

For the starter win-rate entry, specify one opportunity per record and a calendar-month window assigned by closed_date. Limit the denominator to eligible won or lost opportunities. Then settle reopened records, cancellations, the timezone, and missing closure dates. Each unresolved choice leaves the report builder to make policy through filters.

Take an illustrative dataset with 40 wins, 60 losses, and 100 still-open opportunities. Closed-opportunity win rate is 40%; dividing wins by all opportunities produces 20%. Neither calculation explains itself through the label “win rate.” They answer different questions. StartLab’s KPI Dictionary Template for Small Businesses recommends distinct names when calculations differ materially.

The formula is not agreed until the population underneath it is agreed.

How do you reconcile conflicting HR and finance numbers?

Align the business question, reporting window, date basis, population, and exclusions before comparing totals. Then have the accountable owner test the agreed definition against both reports, investigate record-level differences, and record approval. Do not pick the report with the more authoritative-looking format. A shared number must be reproducible, not merely accepted in a meeting.

Worked example: one label, two populations

Consider a hypothetical month-end reconciliation. HR’s “headcount” report shows 104 active people, including four contractors. Finance shows 100 employees and excludes contractors. Both use the same snapshot date, and there are no other discrepancies. Nobody has made an arithmetic error. The reports count different populations.

For workspace planning, publish “active workforce count,” including contractors. For employee-only planning, use “active employee headcount.” Put employment-status rules and snapshot timing in both entries so the distinction survives the next reporting cycle. Broader HR operations metrics need the same discipline. A shared label proves nothing about who was counted.

Follow this approval-and-reconciliation sequence

1. Collect the existing definitions and reports. Ask HR, Finance, and other users for formulas, filters, source fields, and exported records covering the same window. Start with what each report actually does.
2. Agree on the decision. Establish whether the metric supports payroll planning, workspace capacity, hiring, or another purpose. Separate different questions rather than forcing one number to serve them all.
3. Write the calculation rules. Specify numerator, denominator, grain, date basis, inclusions, exclusions, and missing-data treatment. List unresolved choices where reviewers can see them, not inside report logic.
4. Assign accountability and reviewers. Name one definition owner, the person implementing the calculation, and the teams whose approval is required. A list of interested departments is not an owner.
5. Run both calculations under the agreed rules. Compare record IDs and subtotals, not just final totals. Check timing, joins, duplicates, cancellations, and source freshness until every difference has an explanation.
6. Approve, publish, and connect. Record approval, version, and effective date. Link each relevant report to the entry and document remaining limitations before calling the definition shared.

Treat this sequence as a working process, not a guarantee that the numbers will match on the first pass. Capture owners, handoffs, and exceptions in a workflow documentation template. The test is whether someone can apply the approved rules to actual records and explain the result.

Apply the same naming discipline outside HR. Website visitor-to-lead conversion and lead-to-opportunity conversion measure different funnel transitions. Create separate entries. Letting both teams publish an unexplained “conversion rate” simply moves the disagreement into the next review.

How many metrics should you maintain, and how often?

Following Preql’s guidance, start with leadership and board reporting, and review entries quarterly. Prioritize places where conflicting definitions already affect decisions. Expand only as ownership and review capacity allow, and review entries after a material calculation or source change. Preql recommends starting in a spreadsheet and moving to tooling when definitions need to stay synchronized with live reports.

Preql’s Finance KPI Dictionary Template recommends prioritizing leadership and board metrics and suggests this starting range for many teams. Use it to limit scope, not to fill a quota. We would rather maintain a smaller dictionary with accountable owners than a larger one nobody trusts.

Quarterly reviews, dated change logs, and links from reports to dictionary entries are also part of Preql’s guidance. Make the review concrete: check owner assignments, source fields, exclusions, and report links. Then ask whether the metric still supports the decision named in its entry. An unused definition does not earn its upkeep.

Keep the revision date separate from the effective date. Approval today does not mean the new definition applies to today’s reporting period. Record whether historical results will be recalculated, and flag comparisons that cross a definition change. Quietly comparing old-definition results with new-definition results defeats the point of the dictionary.

When spreadsheet maintenance stops keeping pace with live reports, evaluate how tooling connects each definition to its implemented calculation and exposes changes. More documentation is not the same as better control. The useful test is whether teams can see which definition a report follows and when that definition changed.

How Cogniver helps teams approve metric definitions

Cogniver gives metric-definition changes a defined approval path. Build a document approval workflow visually, require the proposed definition document before approval proceeds, and add multi-step reviews. Use branching and merging to involve HR, Finance, and Operations where their judgment is needed instead of sending every proposal through the same chain.

Each workflow gets its own isolated AI agent, trained by organization admins on that workflow’s rules and configuration. The agent answers questions, routes requests, and chases approvers. That takes follow-up off the definition owner’s plate while keeping conversation memory separate between workflows and companies.

Groups and grades from Cogniver’s org chart drive approver resolution. Reviewers can be required to enter values at their step, and later routing can use those values. Put those capabilities around your dictionary to require the document, reach the accountable reviewers, and carry each proposed change through to a decision.