A/B testing is often held up as the gold standard for pricing decisions. But in practice it rarely lives up to the promise.

Results take months to materialise, by which point the market has moved on. You have to choose just two prices to test, when the real question is about the whole price-response curve. You often can’t show different prices to different customers without damaging trust. And in many markets, running a pricing experiment simply isn’t feasible at all.

Causal inference offers something better: a way to understand the underlying factors that drive willingness-to-pay, how they interrelate, and how to use that understanding to find exactly the right price.

The challenge

You have historical data on prices and sales. You can see that when prices were higher, sales were lower. But you can’t simply conclude that raising prices will reduce sales, because past price changes weren’t random.

Maybe you raised prices when demand was strong. Maybe you lowered them to clear inventory. The data contains the ghosts of past decisions.

The approach

Causal pricing analysis typically involves three steps:

1. Map the causal structure. What factors influenced both your pricing decisions and your sales outcomes? Competitor prices? Seasonality? Inventory levels? Marketing spend?

2. Find exogenous variation. Is there any source of price variation that’s independent of demand? Cost shocks that got passed through? Exchange rate movements? Regulatory changes?

3. Estimate and validate. Use that variation to estimate price elasticity, then stress-test the conclusions against alternative assumptions.

What you get

Not a precise point estimate (that would be false precision), but a defensible range with explicit assumptions that you can use to inform a decision.

More importantly, you get a structural understanding of your pricing dynamics. You learn which customer segments are most sensitive, which factors matter most, and where the opportunities are. That understanding compounds: it makes every subsequent pricing decision faster and better-informed.

Flexibility is the point

Causal pricing adapts to your situation. Whether you have rich transactional data or only aggregate trends, whether you operate in one market or fifty, whether you are setting prices for the first time or optimising an existing structure, the framework scales to fit.

It’s not a last resort when experiments aren’t feasible. For many businesses, it’s the more practical and more informative approach from the start.