Your sales
One product at a time, in one currency. Use net price per unit (after discounts). Rows on the same date are combined. Rows with zero units are dropped. At least 12 rows, three price levels and a price that moved by a few percent are needed.
Upload or paste your sales, then click Estimate.
Result
Price curve
What a price change would do
Per period, from the fitted curve. Rows marked "outside" are beyond the prices you have charged: read them as a guess, not a result.
| Change | Price | Units | Revenue | Profit |
|---|
Back-test
How it is calculated
Model: ln(units) = a + e × ln(price), plus optional controls: a linear trend and one annual sine and cosine pair. Fitted by ordinary least squares; the standard error of e is heteroskedasticity-robust (HC1) and the 95% interval uses Student's t. Because the model is in logs, e is the elasticity directly. An 80% interval and the profit figures come from 400 pairs-bootstrap resamples with a fixed seed, so a result is reproducible. Profit: units at price p are predicted with the controls at their average and a smearing correction for the log; profit is (p − cost) × units. The best price is searched only between the lowest and highest price in your data. When demand is elastic the textbook price cost × e ÷ (e − 1) is shown as a reference. Back-test: the first 75% of rows fit the model, the last 25% are predicted, and the average log miss is compared with the same model with the price left out.
Not modelled: promotions and advertising you do not add to the data, competitor prices, stock-outs, cross-effects between products, customers who stock up, and the long-run response to a price change. Elasticity from history is correlation, not an experiment. Unverified: checked against simulated series with known elasticities in unit tests, not against a published dataset or a commercial tool.
Common questions
What is price elasticity of demand?
It is how much your unit sales change when your price changes, expressed as a percentage ratio. An elasticity of -2 means a 10% price rise is followed by roughly a 20% fall in units. Between 0 and -1 demand is inelastic: raising price loses fewer units than it gains per unit, so revenue rises. Below -1 it is elastic: raising price loses more sales than it earns.
How do I estimate price elasticity from my own sales?
Use a history in which your price actually moved, and regress the log of units on the log of price. The slope is the elasticity. This tool does that with robust standard errors, optional controls for a steady trend and for the time of year, and a bootstrap that shows how uncertain the answer is. You only need three columns: date, price and units sold per day or week.
How much price variation do I need?
At least 12 rows, three distinct price levels and a typical swing of about 3% or more. If your price never changed, nothing in your data can say what would happen if it did, and the tool refuses instead of guessing. A deliberate test with two or three prices held for a few weeks each is worth far more than years of a flat price.
What is the profit-maximising price?
With a constant elasticity e below -1 and a unit cost c, profit is highest at c times e divided by e minus 1. For example an elasticity of -2 and a cost of 10 gives a price of 20. This tool does not trust that formula outside your data: it searches only the prices you have actually charged, and warns if the best price sits at the edge of them.
Why can the result be wrong?
Elasticity from history is correlation, not an experiment. If you raised prices in busy seasons, discounted while also advertising, or a competitor changed at the same time, the slope mixes those effects with price. The tool controls for trend and season, flags a slope that is positive, and back-tests whether knowing the price predicts later sales better than ignoring it, but it cannot see promotions or competitors you do not give it.
What should I do with the answer?
Treat it as a hypothesis for a small, reversible price test, not as a number to bet the business on. If the interval says demand is elastic and the best price is well inside your tested range, try moving a few percent toward it for a few weeks and compare. If the interval is wide or the tool says unidentified, run a cleaner test first.
Is my sales data uploaded anywhere?
No. The file is read and analysed in this browser tab. Nothing is sent to a server, there are no accounts or cookies, and the tool never stores your data. It works offline once the page has loaded.