Practical guide
What is Perfect Store
Perfect Store is an approach to work in the retail outlet in which a manufacturer or a distributor describes its own presence standards - assortment, display and planogram, additional points of sale, POSM, price - and regularly assesses every outlet against that list.
- AI Shelf Recognition for Retail Execution
- Trade Marketing Promotion Execution
In detail
Perfect Store helps a manufacturer or distributor manage brand execution at an outlet: what should be available, how it should be displayed and what supports the assessment. It is an execution standard for a chosen format, not a rating of the whole store. Align placement requirements and corrective actions with outlet conditions and agreements with the retailer.
Assess assortment, availability, display, price and promotions using separate criteria. Each needs a defined subject, a pass condition, an evidence source and a rule version. Evidence supports these criteria; it need not earn a separate score. MML, OSA and display checks can overlap, so avoid unintentionally counting the same fact more than once.
Assign standards by outlet format and category. Not applicable differs from not checked: record the reason under an agreed rule and leave missing observations unknown. Do not remove inconvenient checks after a visit to improve the score. New weights or a different outlet mix can change the result without changing the underlying facts. Use the same criteria and weights with comparable coverage for comparisons. If the evidence cannot support this, report scores from different versions separately and retain the original history.
For binary criteria, one option is to divide the total weight of passed checks by the total weight of applicable checks and express the share as a percentage. Weights must be positive, all applicable checks completed and their total weight nonzero. Missing evidence means an incomplete assessment; no applicable criteria means no score. Other scales need their own method. The score and the decision that the standard is met are distinct: agree the threshold and any blocking critical failures in advance. A high total does not override such a failure.
Choose evidence for the question: a SKU list, shelf observation, a readable price label or active promotion terms. A photo describes the area in frame at capture time, not upload time. Link it to the outlet, visit and criterion; an old image or another processing run does not confirm a new state. GPS provides device-location context, not product availability. Use recognition within the confirmed project scope, with image-quality checks and manual review of uncertainty.
An assessment should lead to action: record the gap, cause, owner, target completion time and recheck condition. If a missing agreed price label is a critical failure, a well-executed promotion does not make the outlet compliant. Keep the original gap and new evidence of correction; closing a task does not replace a shelf check. Execution scores, resolved gaps and commercial results are different measures. A higher score does not establish sales growth or promotion ROI without separate analysis.
How it is applied
- Prepare an anonymised outlet-format example, a checklist with its version and weights, and one disputed or critical criterion. Include a permitted photo with capture time, the original assessment, the current collection process and staff roles. A full customer dataset is not needed for an initial discussion.
- Select requirements that can be checked: which SKUs, display location, price label or promotion terms apply to the outlet. Define the observable fact, permitted deviation and evidence for each criterion.
- Before the visit, agree applicability, weights, the pass threshold and critical failures that block a positive outcome. Record the version, unknown statuses and comparison rules. Check a high-score example with a critical failure separately; a new method must not overwrite earlier assessments.
- In a trial assessment, give the same example to different reviewers. If they disagree, clarify the criterion and the evidence needed rather than averaging disputed answers. Review incomplete photos, missing labels, unknown statuses and non-applicable criteria. For recognition, separately check supported products, capture requirements and manual review of uncertainty.
- For each gap, define an action, owner and target completion time. Specify the new evidence that will confirm resolution and who will review it. Retain the original failure; uploading the old photo again does not resolve it. Define an escalation route when the field rep cannot address the issue on site.
- Discuss the example with Smartup through the Trade Marketing and AI Shelf Recognition pages. Compare the original score, critical failure and evidence of correction to choose between visit controls, data rules or photo workflows. Solution scope, integrations, recognition capabilities and launch terms are agreed separately.