AI Pricing Recommendations
Oriens analyses your product price history to surface actionable pricing suggestions — when to raise, lower, or hold your prices based on patterns in your own data.
What the AI does
The AI Pricing Recommendations feature examines the price history Oriens has collected for each of your products and applies a set of statistical models to identify meaningful patterns. It then translates those patterns into plain-English suggestions with a specific target price, an explanation of the reasoning, and an estimated impact on revenue or margin.
The types of patterns the model looks for include:
- Seasonality: Products whose sales velocity historically improves or declines at certain times of year may benefit from seasonal price adjustments. If a product consistently sells well in Q4 with no price change needed, the model may suggest a modest increase heading into that period.
- Price change response: If your price history shows that previous price reductions on a product were followed by periods of stronger sales, the model will note that pattern. Conversely, if past price increases didn't appear to affect velocity, the model may suggest testing a higher price again.
- Relative price drift: If a product's price has been static for a long period while your store's average prices have shifted, the model may flag it as a candidate for realignment.
- Recovery opportunity: Products with a high number of price alert subscribers relative to their sales rate are flagged as having significant pent-up demand — a meaningful price drop could convert a large number of waiting shoppers at once.
What data it uses
Oriens' AI model works exclusively with your own store's data. Specifically:
- The price history Oriens has recorded for each product and variant since your install date
- The number of active price alert subscribers per product
- The timestamps of price changes (to identify patterns over time)
The model does not use:
- Competitor pricing data
- External market indices
- Sales volume or order data from your store (Oriens does not request order read permissions)
- Data from other Oriens merchants
This means the AI is grounded entirely in what has happened with your own prices over time. Its recommendations are as good as the history available — which is why longer install tenure leads to more actionable and confident recommendations.
How confidence scores work
Every recommendation card includes a confidence rating of High, Medium, or Low. This reflects how strongly the data supports the suggestion.
| Confidence | What it means | Recommended action |
|---|---|---|
| High | A clear, consistent pattern is present across multiple data points and time periods. The model has high signal and low noise for this product. | Strong candidate for action. Review the reasoning and apply the change if it aligns with your business context. |
| Medium | A pattern is present but with some ambiguity — either the history window is shorter than ideal, or the signal is present but not consistent across all periods. | Use as a starting point. Consider testing with a smaller price change before committing to the full suggested adjustment. |
| Low | A potential pattern was detected but the data is limited or the signal is weak. The recommendation is speculative. | Treat as a hypothesis worth monitoring rather than a directive. The confidence will improve as more history accumulates. |
What "Insufficient data" means
Some products will show an "Insufficient data" message instead of a recommendation. This appears when either or both of the following conditions are true:
- Fewer than 30 days of price history: The model needs at least a month of tracked data to establish any meaningful baseline. Products installed recently will show this message until enough time has passed.
- Fewer than 3 price changes recorded: If a product's price has never changed (or only changed once or twice) since Oriens was installed, there isn't enough variation in the data for the model to draw conclusions. A product that has always been £49.99 has no price movement to analyse.
In both cases, no action is required — the message will automatically resolve as history accumulates. Products that have been live on Oriens for 60–90 days with at least a handful of price adjustments will typically produce their first recommendations.
Why promotional prices are excluded from the model
Sales and temporary promotions create sharp, short-lived price dips that would distort any model trying to find meaningful trends. A product that normally sells at £80 but was put on a 48-hour flash sale at £40 shouldn't have that £40 price treated as a signal to lower the permanent price.
Oriens detects a promotional price by checking whether a compare-at price is set in Shopify at the same time as the price change. When a compare-at price is set (meaning Shopify shows "Was £80, Now £40"), Oriens tags that price period as promotional and excludes it from the AI model's training data for that product.
The actual recorded price history still shows the sale price — you can see it in the price history chart. It just doesn't feed into the recommendation model. This ensures the AI is analysing your everyday pricing behaviour rather than the noise of campaigns.
How to read a recommendation card
Each product in the AI Analysis section either shows a recommendation card or an "Insufficient data" state. A recommendation card contains four key pieces of information:
Target price
The specific price the model suggests testing, expressed in your store's default currency. This is presented as a range (e.g., "£74–£79") rather than a single price point to allow for your own judgement. The midpoint of the range is the model's strongest suggestion.
Confidence
High, Medium, or Low — as described above. Displayed as a coloured badge: green for High, amber for Medium, grey for Low.
Reason
A plain-English explanation of why the model is making this suggestion, referencing the specific pattern it detected. For example: "This product's price was reduced to £69 for 3 weeks in March 2025. No comparable promotion has occurred since, suggesting the market may support the higher price again."
Estimated impact
A rough directional estimate of what applying the recommendation might do to revenue or margin, expressed as a percentage range (e.g., "+5–12% margin per unit"). This is illustrative, not a guarantee — it's based purely on the price change itself, not demand elasticity, which Oriens cannot measure without order data.
What to do with a recommendation
Oriens recommendations are suggestions, not automated actions. Oriens will never change your prices without you explicitly doing so in Shopify. The AI tells you what to consider — you decide whether and how to act.
A typical workflow looks like this:
- Review the recommendation card for a product. Read the reason carefully — does the pattern it describes match your memory of what was happening at that time?
- Consider factors the model doesn't know about: upcoming campaigns, supplier cost changes, competitive context, inventory levels.
- If the suggestion makes sense, go to your Shopify admin and update the product price manually.
- Oriens will record the price change and, over time, the model will incorporate the outcome of that change into future recommendations for that product.
There is no "apply" button that changes prices automatically. This is intentional. Pricing decisions have real business consequences and should always involve a human review step.