The operating review
Leave the weekly review with a decision you can explain
A weekly storefront review should connect a reported change to its evidence, uncertainty and next owner. Use SmaugBrain Agent work to organize the material and challenge explanations. Keep the commercial decision with the person responsible for the store.
本页内容
Begin with one operating question
Choose a question such as whether a landing page needs clearer delivery information. Do not begin by asking for an explanation of every number. Record the reporting window, comparison period and decision deadline before analyzing changes.
Ask for alternative explanations and missing evidence. A tidy summary should not turn a correlation, tracking change or incomplete export into a confident diagnosis.
Bring evidence that can be compared
These are operator-supplied review inputs, not automatic storefront connections.
| Material | Context to retain | Limitation to record |
|---|---|---|
| Traffic CSV | Dates, timezone, channel and landing page | Tracking changes, missing consent or attribution differences |
| Order summary CSV | Order definition, currency, cancellations and reporting basis | Incomplete periods or different treatment of refunds |
| Product and stock notes | SKU, variant and the time of the observation | A snapshot cannot establish current availability |
| Page captures and URLs | Market, language, device and capture date | The supplied view may not represent every visitor |
| Customer feedback excerpts | Question, context and frequency where known | Remove identities; a few comments are not a representative survey |
Ask for a decision memo
Keep four kinds of statement distinct in the requested output.
Observed changes
Attach each number or page finding to its source file, row reference or capture. Preserve definitions and show any calculation.
Uncertainty
List missing fields, conflicting periods and explanations that require checking. Mark an unsupported conclusion as unresolved.
Proposed action
Name the affected page or process, the intended change, its owner and the evidence that would justify proceeding.
Review record
State what the operator approved, deferred or rejected. Retain the decision and a later comparison method with the original materials.
Example: more sessions, unchanged orders
Invented figures for explanation, not a customer result or a forecast.
Week A: 1,000 sessions; 20 completed orders. Week B: 1,200 sessions; 20 completed orders. Calculated order-to-session ratio: 2.00% and approximately 1.67%. Before interpreting: check period length, session definitions and order status. Possible explanations: a channel mix change, a tracking change, or a customer journey issue. Next evidence: comparable landing-page data and a documented page review. Decision: do not claim a cause or change spend from these totals alone.
Prioritize by consequence, not by how dramatic a chart looks
Use the review to assign responsibility; it is not an automated incident queue.
| Priority | Example trigger | Human next step |
|---|---|---|
| Immediate review | A reported checkout blockage or misleading delivery promise | The store owner verifies the issue and decides whether to pause affected activity. |
| Next operating cycle | A verified content gap or recurring support question | Assign a scoped content change and review it before publication. |
| Investigate first | An unexplained metric movement or mismatched exports | Request comparable evidence; defer claims and commercial changes. |
Keep execution authority with the operator
The current system does not automatically synchronize Shopify data, manage inventory, issue merchant refunds or execute advertising changes. It also has no personal-center task center or frontend task scheduler. A weekly review must be started by an operator.
Prepare the approved CSV excerpts, page materials and question for the configured Agent workspace. Input and tool support depend on that workspace. The account area itself is not a merchant-data importer. Never include store credentials or unnecessary customer identifiers.
Close with one checkable commitment
Finish with a small set of decisions that have owners, approval boundaries and a review date. For a content change, retain the approved wording and page reference. For an unresolved number, name the additional export needed.
Compare the next reporting period using the same definitions. Keep revenue conclusions separate from Agent availability and AI resource use; neither an active runtime nor a higher Token count demonstrates a business improvement.
Keep the two sets of records separate
Do my account order totals describe store sales?
No. SmaugBrain account orders describe your plan purchases. Store sales come from the records you supply. Token statistics describe reported AI usage, not store revenue or profitability.
Review AI usage separately from store performance
Review your account’s reported model calls and Token usage for the relevant time range. These account totals describe AI resource use, not per-Agent billing or commercial results.
View Token statistics