Start with the record, not the impression
Weeks three and four are for reading what the tests actually produced and deciding what to do next. A five-hundred-dollar budget is a planning scenario, not a forecast or a promise of results. The task is not to spend the full amount or to find a winner at any cost. It is to use the information available to make a careful next decision while keeping the work within each platform’s terms.
Before changing campaigns, gather the records for every test. Bring together the diary, tracker, platform reports and any notes made during setup. If information is missing, mark it as missing rather than filling the gap with a guess. A result that cannot be interpreted reliably is not a strong basis for expanding spend.
Use this checklist:
- [ ] Find the original test plan, including its purpose and the criteria written before launch.
- [ ] Gather the diary entries and tracking records for each variant.
- [ ] Match records to the right test, audience, creative, landing page and time period, where those details were part of the setup.
- [ ] Note any interruptions, delivery problems, edits or changes that could affect interpretation.
- [ ] Separate observed facts from explanations or assumptions.
Keep the account itself in view as a working tool. Confirm that access is under your control, the account is being used as intended, and the activity remains consistent with the platform’s requirements. Do not treat account access, tracking gaps or a delivery issue as something to conceal or bypass. If the account or campaign needs attention, address it through appropriate platform processes before drawing conclusions from its performance.
Check whether the test is readable
A test can produce activity without producing a dependable answer. Before comparing variants, check whether the information is complete enough to support the comparison. Look for mismatched tracking, missing diary notes, unclear naming, or a change that made one variant materially different from the plan. These do not automatically mean the test failed, but they affect how much confidence to place in its read-out.
Use a simple classification for each test: readable, limited, or not interpretable. “Readable” means the records broadly match the plan and the prewritten criteria can be applied. “Limited” means there is some useful information, but a caveat—such as an interruption or incomplete record—narrows what can be concluded. “Not interpretable” means the available information cannot answer the question the test was meant to address.
Checklist:
- [ ] Can you identify what question the test was designed to answer?
- [ ] Do the diary and tracker describe the same test and period?
- [ ] Were variants run in a way that still allows a fair comparison?
- [ ] Were any edits or delivery changes recorded?
- [ ] Is there enough usable information to apply the original decision criteria?
Do not repair a weak read-out by inventing certainty. Record the limitation, decide whether a cleaner repeat is worthwhile, and avoid treating a tracking defect as evidence that one variant performed better. If the test is not interpretable, the useful decision may be to fix measurement or stop that line of testing—not to increase spend.
Apply the criteria you wrote in advance
Compare each variant against the criteria set before the test began. Those criteria might cover a relevant action, the quality of that action, or a practical guardrail such as whether the test stayed within its planned scope. The specific measures depend on the campaign and platform. Keep the comparison tied to the original question rather than selecting whichever metric looks most encouraging afterward.
For each variant, write down the observed result beside the relevant criterion. Then note whether it met the criterion, missed it, or remains unclear. Avoid combining unlike measures into a single impression. A variant that attracts attention is not automatically better if the test was intended to assess a later action; likewise, an apparent advantage is not meaningful if the records are incomplete.
Checklist:
- [ ] Apply the same criteria to every variant in the comparison.
- [ ] Keep the original test objective visible while reviewing the metrics.
- [ ] Mark each criterion as met, not met, or uncertain.
- [ ] Record the limits of the comparison, including tracking or delivery issues.
- [ ] Do not add a new success measure simply because the original one is inconvenient.
The point is consistency, not rigidly defending a plan that no longer makes sense. If the objective itself has changed, close out the current test honestly and write a new plan for the next one. Do not relabel the old test after seeing its outcome. A clear record helps distinguish a genuine change in strategy from a result-driven change in the definition of success.
Retire what failed its stop condition
If a test reached a stop condition written in advance, follow it. A stop condition might reflect a guardrail, a clear mismatch with the intended audience or offer, a measurement problem, or another reason specified in the plan. Retiring a variant is a normal outcome of testing; it does not mean the whole campaign or account must be abandoned.
Checklist:
- [ ] Find the stop condition in the original plan.
- [ ] Confirm whether the available evidence shows that it was reached.
- [ ] Pause or retire the affected variant as the plan requires.
- [ ] Record why it was stopped and what the evidence can—and cannot—show.
- [ ] Keep the decision separate from unrelated tests that remain readable.
Do not keep a failing variant active just because money has already been spent. Past spend cannot be recovered by spending more. At the same time, do not stop a test solely because a short-term result feels disappointing if the agreed condition has not been met and the data remain too limited to interpret. Apply the rule you set, note any uncertainty, and choose a deliberate next step.
If a stop condition was reached because the test could not be measured properly, the next step may be to correct the setup and run a new, clearly defined test. That is different from continuing the same test as though the data were sound. Where platform tools or account access are involved, use legitimate settings and support routes; do not evade checks or disguise activity to preserve a test.
Decide whether a winner is ready for another test
A variant that meets the written criteria may be a candidate for further testing, not a guarantee that it will keep performing. Ask whether the result is based on a readable comparison and whether the test was broad enough to answer its original question. If the evidence is thin, mixed or affected by a caveat, describe the outcome as provisional.
Checklist:
- [ ] Did the candidate meet the criteria that mattered most?
- [ ] Is the comparison readable and consistent with the plan?
- [ ] Were there important caveats that could change the interpretation?
- [ ] Would repeating the test answer a useful question?
- [ ] Can the next test change one meaningful element while keeping the purpose clear?
A repeat is most useful when it tests a specific uncertainty. That could mean checking whether the same approach holds under a clean setup, or comparing a controlled variation that addresses an unresolved question. Avoid changing several elements at once if you need to understand what caused a difference. If no worthwhile question remains, there is no obligation to run another round simply because budget is available.
A “winner” is therefore a working label for a candidate that currently fits the criteria. Keep the label provisional in your notes. It should not become a reason to scale automatically, overlook account or policy concerns, or assume that a result will transfer to a different audience, placement, platform or offer.
Rebalance the remaining budget and protect the reserve
Review what remains in the planning budget after accounting for spend already committed or recorded. Rebalance only the amount that is genuinely available for a next step. The five-hundred-dollar scenario is a ceiling for planning, not a target that must be used. If the evidence does not justify another test, keeping funds unspent is a valid decision.
First list the remaining work: a repeat test, a measurement fix, a new controlled comparison, or no further spend. Then estimate what each proposed step would require based on your own campaign setup and platform interface. Do not rely on an assumed standard allocation or an expected return. The right allocation depends on the question, the test conditions and the limits you set.
Keep the reserve untouched unless a decision rule says otherwise. A reserve is not spare money to use whenever a result is inconclusive. If the original plan defines a condition for releasing it, check that the condition has actually been met and document the reason. If there is no such rule, leave it out of the next round or revise the plan before committing it.
Checklist:
- [ ] Reconcile recorded spend and any planned commitments.
- [ ] Separate available funds from the reserve.
- [ ] Assign remaining funds only to a defined test or necessary correction.
- [ ] Confirm that the proposed spend stays within the planned limit.
- [ ] Record why the reserve remains untouched or why a written rule permits its use.
Avoid shifting money to the most promising-looking variant without considering whether the test can still answer a useful question. Rebalancing should follow the evidence and the plan, not the urge to recover earlier spend.
Write the next-round plan before making changes
Before launching or editing anything, turn the read-out into a short decision record. Include the test question, the candidate or variants, the criteria, the stop condition, the tracking plan, and the budget boundaries. Note what will stay constant and what will change. This creates a clean starting point for the next read-out and reduces the temptation to reinterpret results later.
Checklist:
- [ ] State the question the next round is meant to answer.
- [ ] Identify what is being repeated, changed or removed.
- [ ] Define the measures and decision criteria in advance.
- [ ] Set a stop condition and a spending boundary.
- [ ] Confirm tracking and diary practices before launch.
- [ ] Check that the account, campaign and materials are being used in line with platform terms.
If the next step involves a purchased advertising account, treat it as a product and operational tool, not as a shortcut around platform requirements. Keep access properly controlled, maintain accurate records, and resolve account or verification issues through legitimate channels. Do not base a test plan on hiding ownership, avoiding checks or working around restrictions. A compliant setup is part of whether the test can be run and interpreted responsibly.
Choose: continue, change or stop
Close the review with one explicit decision: continue, change or stop. Continue when the test is readable, the candidate meets the prewritten criteria, and another defined test would add useful information. Change when the objective remains worthwhile but the setup, measurement or comparison needs correction. Stop when a stop condition was reached, the question no longer matters, or the available evidence does not justify further spend.
Write the reason beside the decision and identify the next action—or state that there is none. If you continue, specify what the next test will answer. If you change, state what will be corrected before spending again. If you stop, record what was learned without presenting an uncertain result as a success or failure.
The purpose of weeks three and four is not to force a winner from a small-budget plan. It is to turn a diary, tracker and platform read-out into a defensible choice. A careful decision can mean using less of the budget, leaving the reserve untouched, or ending the test cycle. That is a complete outcome when it follows the evidence and the rules set for the work.