Why measurement is planned before it is built
Most measurement failures are planning failures that surface as tool problems. A form submission that fires twice, a campaign that appears under three different names, a conversion that the ad platform counts and analytics does not — each is traced back, eventually, to a decision nobody made. Nobody wrote down which event proves the objective; nobody fixed the naming conventionAn agreed pattern for naming campaigns, ads, events and tracking parameters so that reports can group them reliably.; nobody decided which system is the source of truth when two disagree. The tool is then blamed for a question it was never asked.
Planning first inverts that. The strategy guide to building a measurement plan sets the business framing: what success means, which evidence will be accepted, how credit is shared. This section takes over from there and turns the framing into artefacts that engineers, analysts and channel managers can act on. The order matters because each artefact feeds the next. A KPIA metric tied directly to an objective. tree tells the tracking plan which events exist; the tracking plan tells the naming convention which fields must be consistent; the naming convention tells the data quality routine what "correct" looks like; and the consent design constrains all of it.
The payoff is not elegance. It is that when a stakeholder asks why the numbers moved, the answer takes minutes rather than a week of archaeology, and when the numbers in two systems differ, the difference is explained by a known cause rather than argued about.
The planning sequence
Each artefact has an owner and feeds the one after it. Treat the table as a dependency chain: a change upstream means a review downstream.
| Artefact | Owner | What it feeds |
|---|---|---|
| Objectives and questions | Marketing lead with the business sponsor | The KPI tree |
| KPI tree | Analytics or marketing operations | The measurement plan's metric definitions |
| Measurement plan | Analytics lead | The tracking plan, reporting design and the source-of-truth table |
| Tracking plan | Analytics lead with a developer | Tag Manager configuration, the data layer, QA scripts |
| Naming convention and UTM rules | Marketing operations | Campaign set-up, link building, channel grouping |
| Consent design | Analytics lead with privacy or legal counsel | Tag behaviour, what the plan can promise, modelling expectations |
| Data quality routine | Analytics lead or a named analyst | Weekly checks, audits, the reconciliation method |
| Reconciliation method | Analytics lead with channel managers | Reporting, budget conversations, incrementality questions |
Small teams collapse several roles into one person; that is fine as long as the artefacts still exist as separate documents, because they change at different rates. Objectives change quarterly, the tracking plan changes with every site release, and the naming convention should change almost never.
The core guides
- Measurement plan tutorial — the step-by-step method from objectives to a tracking spec developers can implement: questions, KPI tree, events and parameters, dimensions, platform mapping, identity and consent, owners, hand-off and QA criteria.
- Tracking plans — the document itself: the columns it needs, naming rules for events and parameters, versioning, change control, sign-off and how to keep the plan and the implementation from drifting apart.
- KPI trees — decomposing an outcome into drivers with explicit arithmetic between levels, so that a movement at the top can be traced to the lever that caused it.
- UTM and naming conventions — how analytics reads UTMURL tags identifying traffic source. parameters, what each should carry, case and separator rules, and the governance that stops a convention decaying as more people build links.
- Marketing data quality — the failure catalogue, a weekly check routine, how to run an audit and how to fix causes rather than symptoms.
- Privacy, consent and measurement — what is and is not collected without consent, how consent mode and modelling fit, and how to design measurement that works within the rules of each jurisdiction.
- Why platforms and analytics disagree — counting units, AttributionAssigning conversion credit to touchpoints. windows, consent, timing and a reconciliation method that turns a recurring argument into a known explanation.
The existing KPI framework page remains the short reference for metric definitions; the guides here are the working method around it.
Further guides
Further guides apply the same planning to particular business models, customers, promotions, enquiries and events, add routines that keep the data trustworthy, and cover hiring outside help:
- Measurement for subscription businesses — trial, activation and first payment events, billing as the source of truth, and cohorts and payback by channel.
- Measurement for marketplaces — a funnel for buyers and one for sellers, the matched transaction, liquidity and channel credit for each side.
- New and returning customers — why analytics new users are not new customers, how to define and record customer status, and how to report acquisition and repeat sales by channel.
- Measuring a sale or promotion — judging a promotion after it ends: the baseline, incremental orders, pull-forward and cannibalisation, the margin given away and the new customers it brought in.
- Measuring phone calls and chat enquiries — counting calls and chats as enquiries, attributing them with tracking numbers or tagged links, judging their quality and joining them to the CRM.
- Measuring webinars and events — registration sources, attendance and engagement, offline lead capture, follow-up and the pipeline that follows.
- Keeping personal data out of analytics — where personal data leaks in through URLs, form fields, search terms and identifiers, how to find it, stop it at the source and handle what got in.
- Keeping a measurement change log — a tutorial for recording what changes in tags, consent, sites, campaigns and definitions, where to keep the log and how analysts use it.
- Hiring and evaluating an analytics agency — what an agency, freelancer or consultant is hired to do, what the business keeps, scoping, questions, warning signs, accepting tracking work and ending the arrangement without losing the property.
Measurement roles and governance
Measurement has three jobs that are often confused: deciding, implementing and verifying. The person who decides what to measure (usually the marketing lead or an analytics lead acting for them) should not be the only person who verifies that it was measured, because the same assumptions that shaped the plan will shape the check. The person who implements (a developer, or an analyst in Tag Manager) needs a specification precise enough that "done" is unambiguous, which is what the tracking plan provides.
Governance is light but non-negotiable. A measurement plan has a named owner and a review date. Changes to events, parameters or naming go through the tracking plan's change control rather than directly into a container. UTM links are generated from a shared builder, not typed by hand. A weekly data quality check has a named person and a written checklist. When tools disagree, the measurement plan's source-of-truth table says which number is used for which decision, so the discussion is about causes rather than credibility.
A subscription business adopts the sequence
A meal-kit company with 3 marketing channels and a 4-person team had been reporting from the ad platforms, GA4 and the billing system, and the three never agreed. They spent 2 weeks on planning before touching a tag. The KPI tree reduced 23 tracked metrics to 1 outcome (paid first boxes), 3 outputs and 7 drivers. The tracking plan listed 11 events, each with an owner and a QA step. The naming convention was published as a one-page sheet with a link builder. Six weeks later, 2 of the 3 systems agreed within a documented tolerance of 5%, the billing figure was declared the source of truth for revenue, and the weekly check took 20 minutes instead of an afternoon. Nothing about the tools had changed; the plan had.
Common measurement problems
Most problems arrive as a symptom in a report. The table maps each to the guide that diagnoses it.
| Symptom | Usual cause | Guide |
|---|---|---|
| Nobody agrees what the main number is | No KPI tree; several metrics competing for the top | KPI trees |
| Developers implement something other than what was asked | No tracking plan, or one without acceptance criteria | Tracking plans |
| One campaign appears under several names | Hand-typed UTMs, mixed case, no builder | UTM and naming conventions |
| Conversions double-count or vanish after a release | No weekly check; change control skipped | Marketing data quality |
| Analytics shows far fewer users than expected in some markets | Consent declined; modelling not understood | Privacy and consent |
| The ad platform reports more conversions than analytics | Different counting units, windows and attribution | Why platforms and analytics disagree |
| The plan exists but nobody can find it | No owner, no review date, no link from the brief | Measurement plan tutorial |
| A large share of sessions sits in the Unassigned channel | Mediums outside the documented vocabulary, half-tagged links, sessionless server events | GA4 unassigned traffic |
| Analytics figures fell on the day a cookie banner launched | Expected consent loss plus defaults, blocking or update-order errors | Data drop after a cookie banner |
| Names or email addresses turn up in page URLs or search terms | Forms, confirmation pages, search boxes or campaign links passing personal data into analytics | Keeping personal data out of analytics |
| A metric moved and nobody can say whether the measurement changed | No record of changes to tags, consent, settings, sites, campaigns or definitions | Measurement change log |
| Channels that bring phone calls and chats look weak | Only form submissions are counted as enquiries | Measuring calls and chats |
| New users in analytics are reported as new customers | Analytics counts devices and browsers; customer status is not recorded from the order system or CRM | New and returning customers |
| Tracking built by an outside supplier that nobody in the business can check or change | The property, container or documents held by the supplier; no tracking plan to accept the work against | Hiring and evaluating an analytics agency |
If the symptom is not in the table, start with the data quality guide: its failure catalogue is the longest list in the section, and most unfamiliar problems turn out to be a familiar one wearing a different report.
