Replenishment Recommendations
Replenishment recommendations show, for each product, how often your customers typically reorder it — calculated automatically from order history. This helps, for example, with sending well-timed reorder reminders or suggesting subscription intervals.
/shop/recommendationsWhat can I do here?
- View all computed replenishment recommendations: recommended reorder interval per product
- Search by product
- See each recommendation's confidence (reliability) at a glance
- Trigger a recomputation of all recommendations
Step by step
- Find a recommendation for a product: Enter the product name in the search field to filter the table.
- Assess confidence: Look at the bar/percentage in the "Confidence" column — the higher it is, the more reliable the prediction.
- Recompute recommendations: Click "Recompute". This can take a few minutes; a notice appears once the process has started.
Fields explained
| Field | Meaning | Notes/Effect |
|---|---|---|
| Product | Name of the product the recommendation applies to | — |
| SKU | Product's item number | — |
| Interval | Recommended gap between two orders (e.g. "every 3 weeks") | Calculated from the actual ordering behavior of multiple customers |
| Confidence | How reliable the prediction is estimated to be, as a percentage | Increases with more data points and more consistent ordering behavior |
| Sample size | Number of customers/orders the calculation is based on | More data points tend to make the recommendation more reliable |
| Computed at | Timestamp of the last calculation of this recommendation | Changes on every recomputation |
Values & statuses
Confidence level (recommendation_confidence_level)
| Value | Meaning | What happens |
|---|---|---|
| High (above 70%) | Prediction is based on enough, fairly consistent reorder intervals | Recommendation can be used with high confidence |
| Medium (40–70%) | Some data is available but less clear-cut | Use the recommendation as a guide, not a fixed rule |
| Low (up to 40%) | Little or very inconsistent order data | Treat the recommendation with caution |
Frequently asked questions
Why is there no recommendation for some products?
A recommendation requires at least 5 customers who each ordered the product at least twice. If that's not the case, no recommendation is computed.
How often are recommendations updated?
Recommendations are only updated via a manual recomputation using the "Recompute" button, not continuously in the background.
What does a low confidence value actually mean?
A low confidence means that customers' order intervals for this product vary a lot, or that only few data points exist — so the prediction is correspondingly less certain.
What can I use these recommendations for?
For example, for reorder reminders to customers, or to suggest matching intervals when setting up Subscription Plans.