What you will learn
- The difference between price, presentation and offer
- Price-point, bundle, framing and shipping-threshold tests
- Why profit per visitor is the metric that matters
- Guardrails that protect trust during a price test
- How to sequence price tests so results stay readable
- When not to run a pricing experiment
Price, presentation and offer are three different tests
Many pricing tests fail because they mix three separate questions. The price is the number itself. The presentation is how that number is framed, for example with a crossed-out comparison price, a per-unit breakdown, or a bundle total. The offer is what the buyer gets for the price, such as two units, a bonus item, or free shipping. Changing all three at once makes the result impossible to interpret.
Keep them separate. Test a price point on its own first, then test presentation, then test the offer. That order also protects the customer experience: a store that constantly changes prices and bundles together looks unstable, while a store that tests one clear change at a time looks like it is refining its range.
Four pricing tests worth running
A price-point test compares two or three round, plausible prices for the same product, with everything else unchanged. It answers whether buyers in this market are sensitive to a small step up or down. A bundle test compares a single unit against a two- or three-pack at a per-unit saving, and measures average order value alongside conversion.
- Price point: the same product at two different prices, margin tracked throughout.
- Bundle: single unit versus a multi-pack, judged on order value and profit per order.
- Discount framing: the same saving shown as a percentage, an amount, or a bundle saving.
- Shipping threshold: free shipping above a set order value, judged on order value and margin after shipping cost.
Measure profit per visitor, not just conversion
Conversion rate alone will almost always favour the cheaper price, which makes it a poor guide for pricing decisions. Profit per visitor combines the two numbers that matter: how many visitors buy, and how much profit each order leaves after product cost, shipping, fees and discounts. A test that raises conversion but cuts profit per visitor is a step backwards, however good the sales chart looks.
Record product cost and fulfilment cost before the test starts, and keep them fixed in the calculation. If shipping costs or supplier prices change mid-test, note the change, because it can distort the comparison just as surely as changing the page layout would.
Guardrails that protect trust
Pricing tests should never feel like a trick. Avoid showing different visitors wildly different prices for the same product at the same time without a clear reason, such as a bundle or a quantity difference, because customers compare notes and screenshots travel fast. Keep any comparison price honest: it should reflect a price the product has genuinely been sold at, or a clear bundle value, not an invented figure.
Be careful with frequent price swings for returning customers. If a regular buyer sees a product jump up and down within days, the store looks unreliable even if each individual price was reasonable. Short, documented test windows with a clear decision at the end are fairer than open-ended drifting prices.
How to sequence price tests
Run price tests one at a time and in a sensible order: first the base price, then the bundle or quantity structure, then the framing, then the shipping threshold. Each result informs the next, so the store ends up with a price, a bundle and a threshold that were each chosen on evidence rather than copied from a competitor.
Between tests, return to the winning version and let it run normally for a short period. That baseline period confirms the result holds outside the test window and gives the next test a clean starting point. Rushing from one test into the next without a baseline makes later results harder to trust.
When not to test price
Do not run a price test during a stock shortage, a supplier price change, or a major sale, because the external change will swamp the effect you are trying to measure. Do not test price on a product with almost no traffic either; the sample will be too small to support a decision, and the time spent would be better used on traffic or on the product page itself.
Also avoid testing price as a substitute for fixing a weak product or a confusing page. If visitors do not understand what the product does, or do not trust the store, a cheaper price rarely fixes the underlying problem, and it trains buyers to wait for discounts instead.
Example: A store sells a single kitchen tool at $29. It tests $29 against $34, with the same page, images and shipping terms, for three weeks. The $34 version converts slightly less often, but profit per visitor is higher because each sale leaves more margin. The store keeps $34, then separately tests a two-pack bundle, rather than changing the price and the bundle in the same test.
Frequently asked questions
What is a Shopify pricing experiment?
A Shopify pricing experiment is a controlled change to a price, its presentation, or the offer around it, measured against profit per visitor so the store can tell whether the change actually earns more money.
Should I measure conversion rate or profit in a price test?
Profit per visitor matters most. Conversion rate alone favours cheaper prices, even when a cheaper price leaves less profit after product, shipping and fee costs.
Is it fair to show different prices to different visitors?
Keep tests modest and honest: test clear price points, bundles or quantity differences, keep comparison prices genuine, and avoid frequent swings that make returning customers feel misled.
What pricing tests work best on Shopify?
Price-point tests, bundle tests, discount-framing tests and free-shipping threshold tests are the most useful starting points, run one at a time in that order.
When should I not test price?
Avoid price tests during stock shortages, supplier price changes or major sales, and on products with too little traffic to produce a readable result.