LearnGrokWeekly optimisation goes wrong when the team jumps from a poor headline metric to a budget cut. These prompts give you a repeatable route from the export files to a recorded spend decision. They are for performance marketing leads who own the weekly review, but need channel, creative and analytics work to line up.
Use the prompts in order when the data is messy. If your reporting is already reconciled, start with the diagnosis and return to the audit only when something does not add up.
Key point
Start with the numbers, not the narrative
A budget decision is only as sound as the conversion, spend and comparison data behind it.
1. Check whether the weekly pack can support a decision
Paste the campaign export, conversion data, delivery notes and planned budget into Audit the weekly data pack. Include both the current period and a comparable period. A seven-day export compared with a partial week, for example, can create a false decline before you have considered any campaign change.
The prompt asks for reconciliation rather than a verdict. That matters where platform conversions, analytics events and CRM outcomes differ. It also asks whether the result is safe for a directional move, a limited test, or no move at all.
Do not fill gaps with estimates just to get to a recommendation. Ask the analytics owner for the missing event count, the CRM join, or the dated change log. Product behaviour and available features can vary, so check the current xAI documentation if you need to confirm how your workspace handles uploaded material.
Check
A useful audit names the missing field
You should see the exact metric or comparison that is absent, why it matters, and the next person or report to check.
2. Separate symptoms from likely causes
Run Diagnose campaign underperformance once the audit says the comparison is usable. Give it the campaign brief as well as the results. A higher CPA means something different for an acquisition campaign with a new offer than for a mature retargeting campaign with unchanged settings.
Keep the diagnosis at the level where you can act. If the export shows that conversion rate fell only on mobile landing pages after a page release, that is more useful than “the funnel is weak”. If spend rose but delivery notes show an audience expansion on the same date, retain that as a hypothesis, not a proven cause.
Read the evidence against or missing column carefully. It is the protection against a confident story built from a single metric. A weak CTR, for instance, can point to creative fatigue, but it can also be caused by a placement or audience mix change.
3. Locate the break between ad and outcome
Use Find creative and funnel breaks when ad engagement and downstream outcomes disagree. It works best with consistent creative IDs across the ad report, page analytics and CRM export. Include the ad copy and visual description, not only the creative label. The replacement angle needs to relate to the message actually shown.
The prompt produces actions for each creative: keep, reduce, pause or test a variation. Treat pause as an operational instruction only when its evidence is strong enough and the budget decision permits it. Where a creative has low click-through rate but high-quality leads, the quality outcome should carry more weight than the click metric.
Watch out
Do not optimise to the easiest metric
A cheaper click is not a better campaign if qualified outcomes, purchases or conversion value deteriorate.
4. Make one accountable budget decision
Paste the audited results and diagnosis into Make the budget reallocation decision. Define the total available budget and every constraint before you paste the rows. This prevents a recommendation that increases promising campaigns but silently exceeds the weekly cap.
The output is a change log, not a vague suggestion to “shift budget to winners”. Each row has a current amount, proposed amount, guardrail, owner and reversal trigger. Review the budget accounting section before sending it on. The proposed total must equal the money you can actually allocate, unless the log explicitly records unallocated spend.
Expected impact is deliberately qualitative. The evidence may support a high-confidence reduction in an underperforming unit without supporting a precise forecast of the resulting conversions. Keep those two things separate.
5. Leave a record that can be checked next week
Run Write the weekly optimisation record after the decision is approved. Use the record as the source of truth for channel edits, creative production, page changes and measurement fixes. The rollback condition is important. It tells the operator when a change should be reversed rather than debated after performance has already worsened.
Check
The record is ready when every material change has a measure
Each budget or creative action needs an owner, due date, success measure, guardrail and a stated reversal action.
How to spot a bad answer
Treat an answer as unreliable if it does any of the following:
- States that one change caused a result when several changes happened in the same period.
- Treats unavailable CRM quality data as if it confirmed lead quality.
- Recommends moving spend while leaving the total budget unreconciled.
- Calls a creative a winner from clicks alone despite downstream conversion data.
- Uses a target, benchmark or external market explanation that you did not supply.
- Hides conflicting figures rather than naming the source of each one.
Correct the input first, then rerun only the affected prompt. If the issue is an unresolved tracking conflict or an unjoinable dataset, stop the reallocation. Hold budget where practical, run a limited controlled test where necessary, and assign the data fix in the weekly optimisation record.
