What would you need to learn before changing this page?
“Which version is better?” sounds useful until you ask what better means. Clearer expectations, more saved enquiries and more profitable work are different questions. Evidence that answers one may leave the others open.
Choose the decision first. Then write down the page difference, the evidence you would need and the conditions that could prevent a useful answer. This guide helps you prepare that comparison brief. It does not run an experiment or establish that a change improves results.
Turn a preference into an answerable question
Imagine a fictional repair-class organiser. The page offers a conversation about available sessions; it does not reserve a place. The current button says Get started. A proposed version says Ask about repair sessions.
Both labels lead to the same intended enquiry step. The organiser wonders whether the new wording would make that step easier to understand.
Write the question as: Does the proposed wording make the information-request step clearer to likely readers? That is different from asking whether it creates more bookings. Neither version promises a booking.
GOV.UK's research-planning guidance starts with questions and assumptions, then chooses activities that can answer them. Apply that sequence before deciding that you need a live test. GOV.UK: Plan user research for your service.
Some changes need correction rather than comparison. If the page incorrectly says a request reserves a place, correct the false statement. Do not keep misleading copy available merely to see whether it produces more clicks.
Describe the difference precisely
Give each proposed version an identifier. Save the exact wording or artifact, not a memory of the page.
| Field | Fictional comparison brief |
|---|---|
| Version A | Button: “Get started” |
| Version B | Button: “Ask about repair sessions” |
| Planned difference | Button wording only |
| Intended unchanged conditions | Offer, surrounding copy, form fields and destination |
| Question | Whether readers understand the information-request step |
| Outside this question | Bookings, revenue, fulfillment quality and causal conversion lift |
“Intended unchanged” needs checking against actual versions. A new headline, different form and revised price would create a different comparison, even if the button was the original reason for the work.
If several changes are necessary, record them as a bundle. You may decide whether to use that bundle, but a result cannot automatically identify which part mattered.
Match the question to the evidence
Use this distinction to select the next piece of work. It is a planning aid, not a ranking of research methods.
Scroll horizontally to compare all columns.
| Question | Evidence that could help | What that evidence alone cannot establish |
|---|---|---|
| What do readers think the action does? | Observations of relevant readers interpreting identified versions | An audience-wide conversion effect |
| Does the intended journey work? | Authorized checks of the actual action and receiving result | Which wording people prefer or why they act |
| Do observed outcomes differ between versions? | Comparable version-linked records with defined units, windows and unknowns | That wording caused the difference |
| Did the change cause an outcome difference? | A suitable controlled evaluation with a documented design and analysis | A conclusion from a casual before-and-after total |
For clarity work, the existing comprehension-check guide supplies a session method. This brief decides whether that method addresses your question. It does not turn preference comments into conversion measurements.
For a controlled comparison, GOV.UK's A/B-testing guidance describes a hypothesis, original and changed versions, planned duration and analysis. It also notes setup complexity and data requirements. That source concerns digital health; its general design distinction is useful here, without transferring its examples or outcomes to a funnel. GOV.UK: A/B testing — comparative studies.
This worksheet does not assign visitors, choose a sample size or supply a statistical stopping rule. Resolve the actual design and available tools before calling a proposed activity an A/B test.
Make three different decisions from three evidence situations
The cases below are invented planning situations. No participants, traffic allocation, submissions or results are reported.
C01: Prepare a clarity check
The organiser has identified static versions A and B. The open question is what readers expect after the button. A static review can address that interpretation without sending an enquiry.
Decision: Prepare a clarity check using the existing comprehension method. Record which version each person sees and whether they have seen another version. Seeing the answer once can affect a later interpretation.
Possible use: Revise wording that creates the wrong expectation. Limit: Even consistent understanding would not prove more bookings, sales or successful form submissions.
C02: Hold a behavioral comparison
The question changes to whether B produces more saved enquiries. The available report contains button clicks, but it cannot connect a displayed version to a confirmed new enquiry.
Decision: Hold the saved-enquiry comparison. Ask the measurement owner to establish whether the necessary version and receiving records can be obtained appropriately. Keep the current evidence labeled unavailable for that question.
Possible use: Specify the missing records and permitted observation window. Limit: Clicks do not substitute for confirmed saves. Treating the missing save data as zero would manufacture a result.
C03: Keep a before-and-after record descriptive
The organiser changes the button, changes the session price and starts a different promotion during the same period. Later enquiry totals are available, but the exposed audiences and their version histories are not comparable.
Decision: Document what changed and what was observed. Do not attribute the difference to button wording. Reframe a future comparison around a question the available evidence can support.
Possible use: Identify uncertainties worth investigating. Limit: A bigger total cannot separate the effects of the page, price, audience or timing. A smaller total would have the same attribution problem.
Write the decision rule before looking for a winner
Record what would lead you to revise, retain or investigate the page. Avoid an automatic “highest number wins” rule when the evidence does not support that conclusion.
For the clarity question, a concrete trigger is an observed expectation that the request reserves a place. That finding would prompt a wording review; it is not a numerical success threshold. For the saved-enquiry question, an unresolved version-to-save link prevents the comparison from answering it.
Also record unacceptable problems: a false promise, an inaccessible action or a broken receiving step. Resolving those problems is useful work even if no performance comparison is possible.
Name an owner, a review point and the next evidence task. Preserve observations that conflict with your preferred version. A documented “not enough evidence” decision is more useful than a winner selected from incompatible records.
Complete one comparison brief
Use the page-comparison planning worksheet. It includes versions A and B, C01–C03 and blank records for your decision, evidence needs, conditions, limits and next action.
For a ClickFunnels AI project, include that brief alongside the workflow and access guidance at /agents. No experiment, analytics or allocation capability is established by this exercise. It does not grant access, change the page or measure a result.
Finish with the smallest useful action: correct a known falsehood, prepare a clarity check, resolve a missing measurement or plan a suitable evaluation. Choose the action that can answer the decision you actually face.