In short: a measurement plan links the company's objectives to the actions observable on the site. It specifies what is collected, when, with what consent, in which tool and how to verify the data. It avoids confusing traffic, click, accepted form, qualified request and sale.
Installing an analysis tool does not automatically create a reliable measurement. Events can be duplicated, blocked, triggered before success, or separated from the business result. The right starting point is therefore not the list of tags, but the decisions that the team wants to make.
Starting from business decisions
Ask which questions need to be answered: which pages generate relevant requests? Which channels bring confirmed sales? Where do users abandon a subscription? What content assists a journey without being the last click?
For each question, name the possible decision. If no action changes according to the result, the collection may be useless. This rule limits dashboards filled with numbers that no one uses.
Build a value chain
Classify the signals by level. A page view indicates an exhibition; a click shows an intention; an accepted form confirms a technical transmission; a qualified request adds a commercial judgment; a paid sale materializes a result.
These levels are not interchangeable. A campaign can increase shipments while decreasing the quality of requests. Linking the site to the CRM or to the sales system allows a more useful reading, provided that access, purpose and data protection rules are defined.
Define an event dictionary
For each event, document a stable name, exact condition, allowed parameters, exclusions, truth source, and owner. For example, a request should only be counted after the server's positive response, not at the click of the button.
Plan for deduplication when an event can go up from the browser and server. Google's recommendations on GA4 events distinguish recommended events from custom events; the choice must remain consistent between pages and tools.
Avoid unnecessary personal data
A measurement tool usually does not need the name, email, phone number or free message of a prospect. Do not place this information in a URL, page title, event parameter, or advertising tool.
Use controlled technical identifiers when the case is legitimate, document their duration and limit access. The principle is simple: collect the data necessary for the decision, not everything that is technically available.
Integrate consent from conception
The behavior of the tags must reflect the actual choice of the user. A refusal should not be presented as an acceptance, and unnecessary tracers should not leave before the choice when consent is required.
The CNIL details the rules applicable to cookies and other tracers and specifies the possible special conditions for some audience measurement solutions. The exemption is not automatic: it depends on the purpose and configuration. The measurement plan must therefore identify the tools, their role and the regime chosen, with validation adapted to the context of the company.
Choose a source of truth by result
The browser can confirm that an interface has displayed a success. The server knows if a request has been accepted. The CRM indicates whether she has been qualified. The payment platform confirms a transaction. Assign each result to the most reliable source and keep the others as intermediate signals.
For forms, check four steps separately: interface-side validation, HTTP acceptance, storage or transmission, then notification or consultation by the team. This distinction avoids declaring a conversion when a request is lost after the confirmation screen.
Prepare the campaign nomenclature
Set source, support, campaign, and content settings before launch. Use short, consistent, and documented values. Variations in capital letters, accents or abbreviations quickly create several lines for the same campaign.
Keep a shared reference table with the date, owner, landing page, and goal. Never place personal information in UTM settings: it can appear in history, logs and third-party tools.
Design an action-oriented dashboard
Start with a few indicators: volume of useful results, rate between two clearly defined stages, lead quality, cost when the expense is known and evolution over time. Always display the period, scope and filters.
Separate acquisition, behavior and business result. An advertising report can assign a conversion according to its own window; an analysis tool or a CRM can use another rule. These differences are not necessarily mistakes, but they must be understood before comparing the numbers.
Test before going online
Prepare a protocol with accepted, rejected and erroneous cases. Control the consent granted and then refused, the reloading of the page, the double click, the navigation between domains and the common blockers. Check in the debugging tool and then in the final destination.
For a campaign, our guide to the Google Ads landing page explains how to check the continuity between ad, page and conversion. A technically active marking does not compensate for an ambiguous offer or an unusable form.
Monitor quality over time
A measure can be derived after a redesign, a change of form, an update of the consent banner or the addition of a provider. Organize periodic checks on priority events and document versions.
Add proportionate alerts: complete drop in requests, abnormal increase, no event after a deployment or strong difference between the site and the CRM. An alert signals that it is necessary to investigate; it does not prove the cause alone.
Respect the limits of attribution
A conversion can follow multiple searches, devices and offline exchanges. No model perfectly reconstructs the entire course. So present attribution as a reading method, with its window and limits, not as an absolute truth about the influence of a channel.
Match the quantitative data to the feedback from the field. The question "how did you know us? ", trade and grounds for refusal can explain what an automatic report does not see.
The minimum sheet per event
To make the plan verifiable, keep:
-
the business objective and the associated decision;
-
the name of the event and its exact trigger;
-
the authorized parameters and the excluded data;
-
the expected consent status;
-
the source of truth and the deduplication rule;
-
the recipient tools and the managers;
-
the test scenario and the date of last verification.
BlackStone Digital can integrate measurement into an SEO and paid-search strategy or a new conversion journey. To define the need, tell us which actions genuinely matter to your business and which tools you already use.
