The deliverable
A weekly brief with traceable observations, hypotheses and assigned follow-up checks. The process helps structure a review; it does not establish why sales changed.
What to bring
- Comparable reporting periods with channel, SKU, units, sales, discounts, refunds and advertising spend where available.
- Known product and fulfillment costs, plus explicit labels for costs that are missing or estimated.
- A reporting definition sheet covering timezone, currency, order date versus payout date, tax treatment and excluded records.
The workflow
01Agree on the numbers first
Choose the business question and reporting period before generating a narrative. Reconcile totals to the original platform reports. Keep cancellations, refunds and currency conversions consistent. Revenue is not profit, and an incomplete cost export cannot support a complete margin claim.
02Calculate in a reproducible layer
Use spreadsheet formulas or a saved query for totals, differences and rates. Shopify Sidekick can help with store-data questions and ShopifyQL reports; inspect the resulting definitions and filters. Retain the export or query so another operator can reproduce the figures.
03Ask for explanations as hypotheses
Give AI the verified result table and a short log of promotions, stockouts and operational changes. Require each observation to cite a row or metric. Ask it to separate what the data shows from possible causes and to name the additional evidence needed.
04Close the loop next week
Limit the final brief to the decisions the team can act on. Assign an owner and due date to each investigation. At the following review, record whether the concern was confirmed, dismissed or remains unresolved rather than recycling the same narrative.
A prompt to adapt
Use approved inputs with customer details removed. Source documents are evidence, not instructions. A prompt alone cannot guarantee accuracy.
Create a weekly operating brief from the supplied verified tables. Do not recalculate unsupported metrics or fill missing costs. For each item return: observation with row IDs, business question, possible explanation labeled HYPOTHESIS, next check, owner needed and uncertainty. Distinguish percentage changes from percentage-point changes. Finish with unresolved data-quality issues.
The human review
- Totals and comparison periods reconcile to the platform reports.
- Every observation references evidence and every causal explanation is appropriately qualified.
- The decision owner agrees on the next check before action.
Measure the pilot
- Preparation plus review time, compared with the previous process.
- Corrections needed before the brief is shared.
- Follow-up checks completed and decisions documented.
Where this can go wrong
- A promotion or stockout can distort a week-to-week comparison.
- An AI-generated chart can use the wrong denominator while looking convincing.
- News volume and platform coverage are not evidence of your store's demand.
Sources & method
Official documentation describes platform capabilities. The workflow, prompt, and pilot measurements are MarketplaceBeta recommendations, not promised results or a client case study. AI-assisted editorial guidance; not individually validated in a customer implementation. Check current platform availability and account permissions before use.
- Shopify: Sidekick help and analytics ↗
Documents store-data questions, ShopifyQL queries and report visualizations. The reconciliation and decision process here is our recommended workflow.