Feature

Automatic Promotion Detection for Ecommerce Pricing

Running a sale creates a spike in units sold at a lower price. If that data goes into your elasticity model, it makes your customers look far more price-sensitive than they actually are — and Zorin will recommend prices that are too low. Promotion detection fixes this automatically.

Start free trial

The problem with promotional data

Price elasticity is estimated from the relationship between price and units sold. When you run a promotion, sales spike — not because your customers suddenly became more price-sensitive, but because of marketing, urgency, and the temporary discount itself. Including those data points tells the model the wrong story.

The result: a model that overestimates price sensitivity and recommends you lower prices further than your actual customers require.

How Zorin handles it

1

Z-score detection

Zorin calculates the residuals between your actual sales records and the fitted log-log demand curve. Records where the z-score exceeds 2.0 are flagged as statistical outliers — likely promotional periods.

2

Automatic or manual

Run auto-detection with one click, or toggle individual records manually. You stay in control — if Zorin flags something that was not a promotion, you can unflag it.

3

Excluded from model fitting

Flagged records are automatically excluded when you fit or re-fit the elasticity model. The underlying data is preserved — only the model training ignores it.

Why this matters for your recommendations

More accurate elasticity estimates

Removing promotional noise gives the model a cleaner view of how your customers respond to price at normal trading conditions.

Prevents artificially low price recommendations

Without promotion detection, the model may recommend lower prices than your customers actually require to buy at normal demand.

Especially important for seasonal stores

If your store runs regular promotions — seasonal sales, flash discounts, clearance events — this data would otherwise dominate your model and skew every recommendation.

Your promotion data is not deleted

Flagged records are only excluded from model training. Your full order history remains intact for reporting and profit tracking.

Price recommendations that reflect real demand.

7-day free trial. No credit card required.

Start free trial