Price Experiment Use Cases
Eight price experiments worth running in Elevate A/B Testing, each with a hypothesis, a setup sketch, and what a win looks like.
Concrete price experiments worth running, and what you learn from each. Every one includes the hypothesis to record before you launch, how to set it up, and what a win looks like.
Start with the one that matches a question you already argue about internally. Those are worth running first, because you already care about the answer.
Test a price increase
The most valuable price experiment, and the one almost nobody runs. Most stores assume a higher price costs them sales, and never check whether it costs fewer sales than the extra margin is worth.
Hypothesis. Demand for this product is less price sensitive than we assume, so a ten percent increase will reduce conversion by less than ten percent and increase revenue per visitor.
Setup. One product with steady traffic. Two variations. Autofill: Price Increase, 10, Percentage. Goal metric: Revenue per visitor.
A win looks like. Conversion drops slightly and revenue per visitor rises.
Watch out for. Judging this on conversion rate. The variation will convert worse. That is not the question.
Find the discount depth that actually pays
You discount because it works. The open question is whether it needs to be that deep.
Hypothesis. A fifteen percent discount converts nearly as well as twenty five percent, so the shallower discount earns more profit per visitor.
Setup. Three variations: current discount, one shallower, one deeper. Goal metric: Profit per visitor. Add COGs first, or profit cannot be calculated.
A win looks like. The shallower discount matches or beats the deeper one on profit per visitor. You keep the volume and the margin.
Watch out for. Running this during a promotional period. Your baseline is not normal.
Test the strikethrough
A compare at price sets an anchor. Whether the anchor helps, and how high it should sit, is testable without changing what you charge.
Hypothesis. Showing a compare at price alongside the selling price increases perceived value and lifts conversion.
Setup. Two variations at the same selling price. Autofill: Compare Increase. Goal metric: Revenue per visitor.
A win looks like. Same price, more revenue. Your margin did not move.
Watch out for. Compare at prices you cannot substantiate. Check what your market requires before anchoring against a price you never charged.
Find your charm pricing
Whether prices ending in 99 outperform round numbers is one of the oldest arguments in retail, and cheap to settle on your own store.
Hypothesis. Prices ending in 99 convert better than round numbers at the same approximate level.
Setup. Two variations a few cents apart, for example 49.99 against 50.00. Autofill rounds up, down, or to nearest, so it will not produce a .99 ending. Across a catalogue, set these by CSV. Goal metric: Revenue per visitor.
A win looks like. A measurable difference from a price change too small to affect margin.
Watch out for. Expecting a large effect. Run it catalogue wide rather than on one product.
Price differently by market
If you sell internationally, your prices in each currency are probably a conversion of your home price rather than a decision.
Hypothesis. Buyers in this market will accept a higher price than a direct currency conversion produces.
Setup. Add the currency to the experiment and set prices for it. Selecting the currency scopes the experiment to that market automatically, so no audience filter is needed. Goal metric: Revenue per visitor. Requires a Premium Elevate plan.
A win looks like. A market specific price that earns more than the converted one.
Watch out for. Running out of traffic. One market is a fraction of your visitors, so this needs a longer run.
Close the subscription gap
Subscription pricing is usually set once, as a percentage off the one time price, and never revisited.
Hypothesis. A smaller subscription discount will not reduce signups enough to offset the extra revenue per order.
Setup. Requires Recharge. Two variations with different subscription prices at the same one time price. Goal metric: Revenue per visitor, with Subscription percentage of orders as a secondary read.
A win looks like. Subscription share holds steady while revenue per visitor rises.
Watch out for. Reading this on signup rate alone. A cheaper subscription always wins that metric and can still lose money.
Price a category, not a product
Single product tests answer a narrow question slowly. Category wide tests answer a broader one faster, because more traffic enters.
Hypothesis. This entire category is underpriced relative to what buyers will accept.
Setup. Filter by collection or tag and select all, up to 500 products. Autofill applies a percentage change across every product at once. Goal metric: Revenue per visitor.
A win looks like. A pricing rule you can apply to a whole category, rather than a single product answer you cannot generalize.
Watch out for. Mixed margins inside one category. If cost of goods varies widely, use Profit per visitor.
Test the price and the message together
Sometimes the question is not what the price should be, but how it should be framed.
Hypothesis. A lower price presented with an explicit savings message will outperform the same lower price shown on its own.
Setup. Three variations: current price, lower price, lower price plus a savings message added through the optional content layer. Goal metric: Revenue per visitor.
A win looks like. The message adds lift on top of the price change, which tells you presentation is worth investing in.
Watch out for. Two variations that each change two things. Keep one as price only, or you will not know which half did the work.
Choosing between these
One high traffic product and a pricing argument
Test a price increase
A discount you have never questioned
Find the discount depth that actually pays
Low traffic and a full catalogue
Price a category, not a product
International sales priced by conversion
Price differently by market
Recharge subscriptions
Close the subscription gap
Whichever you choose, record the hypothesis in the experiment before you launch. Results are much easier to act on when you can see what you expected.
Last updated