PAPERBOAT / MARKETING DECISION WORKBENCH
Campaign Measurement Plan Builder
Write the brief before somebody builds the dashboard. Leave with a counting rule, evidence owner, review window and a list of questions the team still needs to settle.
By Robert Urban. Free to use, with no email gate.
What should happen after the report arrives?
“Measure the campaign” sounds like an assignment until you hand it to three people. One starts tagging buttons. Another opens the CRM. Somebody asks what the dashboard should look like. All three can do competent work and still deliver three different answers.
This workbench produces a small, specific document they can agree on first. Choose the kind of outcome, describe one decision, and write down what the measure includes. The brief keeps missing answers visible. It does not connect to your accounts, inspect your implementation or decide that your campaign worked.
Use it before a campaign launch, a reporting rebuild or a handoff between an agency and an internal team. If the question requires a broader investigation, the marketing analytics and reporting service covers that work. Here, the job is to leave with a usable specification. If that specification leads to a two-group conversion test, the Marketing Experiment Sample Planner can help estimate the observations needed under explicit assumptions.
Describe rules, not individual people. Do not enter customer records, personal information or confidential client details. This tool processes entries locally and does not save them between visits. Export is optional.
No brief generated.
Your brief will appear here.
Try the example. Then challenge its denominator.
The fictional example concerns inquiries for a commercial maintenance contract. The business wants to know whether a larger campaign budget deserves a bounded test. The first version of the measure is inquiry qualification rate, not revenue and not a claim that the campaign caused a sale.
The numerator is unique inquiries meeting an agreed contract-fit definition. The denominator is valid unique inquiries from the same acquisition cohort. That sounds straightforward until some records have not been reviewed. Treating every unreviewed inquiry as unqualified would make a slow follow-up process look like weak demand. Quietly removing those records could make the campaign look better than the available evidence warrants.
The example therefore asks the team to show unreviewed records separately and wait for a suitable maturity check. Its review date is deliberately blank. The tool does not know your sales cycle or the availability of the person who must approve the definition. A fictional date would make the example tidier and the handoff less honest.
Load the example to see the output structure. Replace its rules rather than merely replacing the campaign name. A retail order, a booked consultation and a downloaded paper need different evidence.
Four outcomes, four different acceptance checks.
Inquiry: Test the difference between a successful receipt and an attempted submission. Include a failed validation path and a duplicate. Agree how tests and spam are excluded without discarding genuine repeat business.
Purchase: Follow the order record. A confirmation page can be revisited; an order can later be refunded. The value field asks what the number represents instead of labeling every transaction “revenue.”
Appointment: Separate a request, a confirmed slot, a reschedule and attendance. If the business question is about completed consultations, a booking count alone leaves important work unfinished.
Content interaction: Specify an observable action and its boundary. A reader reaching a section is different from understanding it. A download is different from a buying conversation. Keep the downstream question in the brief even if it cannot yet be answered.
The branch changes the evidence prompt and the acceptance-test list. The count, rate and monetary-value options change which definition fields belong in the output. Switching from a rate to a count removes the old denominator from the generated brief; it does not silently keep presenting an abandoned formula.
Take the brief to the people who can disagree with it.
Ask the implementation owner to explain exactly when the proposed event or record exists. Ask the commercial owner whether that occurrence means what the metric label suggests. Ask whoever operates the source system what can arrive late, be overwritten or appear twice. Those are different questions, and the answers need to meet in the same document.
Then run the stated acceptance checks with permitted synthetic records. Record the expected result before testing. A successful path matters, but so does proving that a failed submission does not become a successful inquiry. Reconcile the source record with the reported observation and write down any population differences instead of promising that all systems will match perfectly.
The output is a text brief so it can move into an issue tracker, a project document or an agency scope. Assign the implementation task there. This builder does not create tickets, email colleagues or install tags. After any input changes, export and print are disabled until you rebuild the brief, so you are not handed the previous version by accident.
For teams using GA4, Google’s recommended-event reference distinguishes generating a lead from qualifying one. Treat the reference as implementation documentation, not permission to equate every lead with a sale. The technical owner must check the actual configuration and current requirements.
Keep personal records out of the planning exercise.
Write “sales operations” rather than a colleague’s personal contact details. Describe a source system rather than pasting an export. Use a fictional inquiry to discuss a matching rule. The useful part of this exercise is the definition, not the identity of the customer.
The custom tool runs in your browser and does not add network requests, input storage or share links containing your entries. Downloading creates a file only when you ask for it. Keep the exported brief with the project so the team can refer to the same definitions.
Google’s guidance on avoiding personally identifiable information warns about identifiers in URLs, page titles and user-entered fields as well as obvious contact details. A measurement specification should therefore name what must not be collected, not just what somebody wants in the report.
Before you hand it off
Does a finished brief mean tracking is ready?
No. It means you have written a specification. Your team still needs to implement it, test the successful and unsuccessful paths, and confirm that the resulting data means what the brief says.
Can I export with unanswered fields?
Yes. Open requirements are included explicitly. That can be useful for a planning meeting. It is not a launch approval, and an output with no blank fields is not a quality score.
Will this calculate campaign ROI or statistical significance?
No. It organizes the measurement requirements. It does not analyze observations, estimate lift, select a sample size or turn an attribution claim into proof of causation.
Can the plan cover several measures?
Use one primary measure per brief. Additional diagnostic measures can have their own briefs when they require different sources or counting rules. Keep the relationships in your project document rather than forcing unrelated definitions into one field.
What if I do not know the event name?
Leave the optional technical name blank. The brief retains an implementation question. Choosing a plausible-sounding name is less useful than asking the responsible person to confirm the mapping.
Does the tool remember my work?
No. It does not save entries between visits. Download the brief if you want to keep it. Clearing the plan asks for confirmation; leaving or reloading the page can still lose unsaved work.
Bring the unresolved definition.
If the form is easy to complete but the team cannot agree on what a qualified inquiry means, that disagreement is the next piece of work. I help connect measurement to the commercial decision and the people responsible for it.
Talk through the measurement problem or explore the broader PaperBoat services. You can use the tool without starting a conversation.