What this template is for
The template exists for the moment a team agrees on an objective and someone asks "so what do we track?". Without a plan the answer is usually "everything", which means nothing is defined, nothing is owned and the reports that follow cannot explain what they show. The structure follows the measurement model widely attributed to Avinash Kaushik, cited in the sources, in which objectives lead to goals, goals to KPIA metric tied directly to an objective. choices, KPIs to targets and targets to segments; this template adds the implementation layers beneath it, the events and parameters that must exist for each KPI to be computed, and the governance layer that keeps the whole thing current. The measurement plan framework page summarises the idea; the measurement plan tutorial walks through producing each part.
Use the template when a new objective is set, a new site or app is being built, a quarter is being planned, or an inherited analytics setup has no explanation. Do not use it as a tracking plan: the events table here records which events the plan needs and why, while the tracking plan guide owns the implementation detail of each event, its triggers, its data-layer shape and its QA status. Keep the two documents linked rather than merged, so that the plan stays readable by the people who own objectives. The other templates in the measurement templates hub feed this one: the KPI dictionary defines the metrics the KPI column names.
The template
The plan is three tables. The first runs from objectives to segments; the second lists what must be collected; the third says who keeps it alive. Copy all three into one document, one tab or section each.
Objectives, questions and KPIs
| Field | What to write |
|---|---|
| Objective | One business outcome in plain words, with the period it applies to |
| Decision it informs | The choice someone will make differently when this number moves |
| Questions | The plain-language questions a reasonable person asks when the objective moves, one per row |
| KPI | The single metric that answers each question, named exactly as in the KPI dictionary |
| Target | The value the period aims at, with the baseline it was set from and who agreed it |
| Segments | The dimensions reports will cut this KPI by: channel, market, device, customer type, product line |
| Guardrail | A metric that must not worsen while this KPI improves |
Events and parameters
| Field | What to write |
|---|---|
| KPI served | Which KPI from the first table this event feeds |
| Event name | The documented or custom event name, in the naming style the tracking plan uses |
| Trigger | The user action or system state that fires it, in words |
| Parameters | The parameters the KPI or its segments need, with the value each must carry |
| Platform | Where the event is collected and where it is reported: analytics, ad platform, CRM, warehouse |
| Status | Planned, implemented, verified, or retired, with the date of the last change |
Owners and cadence
| Field | What to write |
|---|---|
| Artefact | The thing being owned: an objective, a KPI definition, an event, a report, this plan |
| Owner | One named person, not a team |
| Cadence | How often the owner reviews it and how often the number is reported |
| Audience | Who reads the report and what they decide with it |
| Acceptance criteria | What "working" means: the check the owner performs and the evidence that passes it |
| Last reviewed | The date the owner last confirmed the row still matches reality |
How to fill it
Fill the tables in order and resist filling the events table first, however familiar the event names are. Start with objectives written as outcomes, not activities: "more qualified leads from the new market" is an objective, "launch a campaign in the new market" is a task. Add the decision each objective informs; an objective with no decision behind it is reporting for its own sake and should be removed before it generates a dashboard. Then write the questions. If leads fall, is it fewer visitors, fewer converting, or a change in where they came from? Each question becomes a branch of a KPI tree, and the KPI trees guide explains the arithmetic between levels; the point for the template is that every KPI row should descend from a question, so that a moving number can be explained.
Name KPIs exactly as the dictionary names them, and if a KPI has no dictionary row, write the row before continuing. Set targets from a baseline the plan records, not from a benchmark, and name who agreed each one. Choose segments that correspond to decisions: cut by channel if budget moves between channels, by market if markets are run separately, and no further. Guardrails stop a KPI being gamed; a lead target without a lead quality guardrail invites forms that collect noise.
The events table is where the plan meets GA4Google's event-based analytics platform. or whatever analytics tool is in use. One row per event per KPI served; where an event serves several KPIs, repeat the row rather than merging, so that retiring a KPI shows exactly which events it frees. Parameters should carry the segments: if a KPI is cut by product line, the event must carry a product line parameter, and campaign-level segments depend on the UTMURL tags identifying traffic source. conventions being applied to every link. The owners table is filled last and is the part most teams skip. Each row names one person, a cadence and an acceptance criterion; "the dashboard looks right" is not a criterion, "the daily key event count matches the CRM within the tolerance we agreed" is.
Filled example
The example below fills all three tables for an invented regional furniture retailer with a website and two showrooms. The figures are illustrative and belong to no real business.
Measurement plan for Dresserledge Furniture, first quarter
Objectives, questions and KPIs:
| Objective | Decision it informs | Question | KPI | Target | Segments | Guardrail |
|---|---|---|---|---|---|---|
| Grow online sofa revenue to £180,000 in Q1 | Budget split between paid search and paid social | Are more people arriving, or are more of them buying? | Sessions; purchase conversion rate | 60,000 sessions; 1.4% from a 1.1% baseline (agreed by the ecommerce lead) | Channel, device | Return rate stays under 6% |
| Book 240 showroom appointments in Q1 | Whether to extend showroom hours and staff Saturdays | Which channels and which pages produce bookings? | Appointment bookings; booking rate per landing page | 240 bookings; 2.5% on the showroom pages from 1.9% | Channel, showroom, landing page | Cost per booking stays under £35 |
Events and parameters:
| KPI served | Event name | Trigger | Parameters | Platform | Status |
|---|---|---|---|---|---|
| Purchase conversion rate | purchase | Order confirmation page loads with a transaction id | transaction_id, value, currency, items with item_category | GA4 via GTM; Google Ads via import | Verified 2026-01-09 |
| Booking rate | appointment_booked | Booking form submits successfully | showroom, landing_page, preferred_day | GA4 via GTM; CRM via webhook | Implemented, awaiting QA |
| Sessions by channel | session start (automatic) | Session begins | utm_source, utm_medium, utm_campaign from the convention | GA4 | Verified 2026-01-09 |
Owners and cadence:
| Artefact | Owner | Cadence | Audience | Acceptance criteria | Last reviewed |
|---|---|---|---|---|---|
| Sofa revenue objective | Priya Nair, ecommerce lead | Weekly review, monthly report | Managing director, buying team | Weekly GA4 revenue within 5% of the order system | 2026-01-12 |
appointment_booked event | Tom Okafor, web developer | Checked after every release | Analytics owner | Test booking appears in DebugView and CRM within one minute | 2026-01-09 |
| This plan | Priya Nair | Quarterly, and after any site release | Marketing team | Every event row is "verified" or has a dated owner note | 2026-01-12 |
Common mistakes
- Starting with events. A plan that begins with the events the team already knows ends up measuring what is easy rather than what the objectives need.
- Objectives with no decision. If nobody would do anything differently when the number moves, the row is reporting, not measurement; remove it.
- KPIs named loosely. "Conversion rate" without a dictionary row means every dashboard will compute it differently; link each KPI to its dictionary entry.
- Targets copied from benchmarks. Set targets from the recorded baseline and name who agreed them; external figures belong in a hypothesis, not a target.
- Segments nobody acts on. Each segment should correspond to a budget, a team or a product decision; the rest is noise that slows every report.
- One owner called "marketing". Teams do not review documents; people do. One name per row.
- Merging the plan with the tracking plan. The plan explains why each event exists; the tracking plan explains how. Merged, neither audience reads it.
