Shopify brands facing cost pressure should test and lock their base price by roughly November 10, then freeze testing through Cyber Monday. Judge the winner on revenue or profit per visitor, not conversion rate. ABConvert fits brands above $2M that want price, shipping, and page tests in one app.
Most A/B tests lose. The brands that win BFCM are the ones that found out which price loses in October, while losing was still cheap.
It is the middle of October. Black Friday lands on November 27, and somewhere in your Slack there is a thread from August about raising prices that never got resolved. Freight went up. Duties went up. The finance spreadsheet says you need four or five more points of margin. And every time the decision comes up, someone says what everyone is thinking: what if we raise it and conversion falls off a cliff right before our biggest month?
That thread is where a lot of brands between $2M and $10M are sitting right now. I recorded a conversation this week with Jeffrey Lin, the founder of the Shopify testing app ABConvert, and the most useful thing he said was not about winning tests. It was about losing them. His data says roughly nine in ten variants do not beat the original. The brands that get value from testing are the ones that treat a loss as information, and they need enough calendar left to act on it.
This piece is for the operator who has to make the price call before the promo calendar is final. If you are earlier than that, the skip line above is honest. If you are past $20M and already run a testing program, the calendar and the freeze rule near the end are the parts worth your time.
Your base price has to be settled before Black Friday because every promotion you run that week is calculated from it. A 25% discount on an underpriced product locks the underpricing in for the highest volume days of your year, and it trains your best customers to expect that number when the sale ends. Change the base price after the sale and you are raising prices in January on people who just bought from you at a discount.
The timing is tighter than it looks. Our own read of the BFCM 2026 shopper data found that roughly 80% of shoppers start before Thanksgiving week, with 23% beginning in early November. So the window where you can change a price without a large share of your season’s buyers seeing it twice closes in the first half of November, not on the 26th.
Work backward from that. A price test needs time to cover at least two full weekly cycles, because weekday and weekend shoppers behave differently. ABConvert’s own guidance says most stores reach a readable result in 7 to 14 days and smaller stores may need 2 to 4 weeks. Start a test the week of October 19 and you can read it around November 9, roll out the winner, and rebuild your promo math off the new number before early November traffic ramps.
The second reason is cost. Many brands carried 2025 and 2026 tariff increases on their own margin and told themselves they would deal with it after peak. That is exactly the situation where a guess is most expensive. If you have been through our tariff pricing guide, you have seen the recommendation to test increases in steps rather than announce one. This is the season that advice was written for.
ABConvert is a Shopify app for A/B testing prices, shipping rates and thresholds, offers, page templates, whole themes, and checkout. You split visitors into groups, each group sees one version, and the app reports which version produced more revenue or profit per visitor, using your own product, shipping, and transaction costs when you supply them.
The context worth knowing as a buyer: Jeffrey launched the app on October 31, 2022, bootstrapped, after a classmate told him he never knew what price to charge for new products. It now lists more than 2,000 Shopify stores and over 50,000 tests run. On the Shopify App Store listing it carries the Built for Shopify badge and a 4.7 rating across 83 reviews as of October 14, 2026. Pricing starts at $99 a month for 1,000 test orders, and the bill scales with test orders processed, not with your store’s total volume. One detail matters for this article: price testing starts on the $199 Growth plan, not the $99 Starter plan.
The newer piece is AI. ABConvert runs an MCP connection that lets Claude and other assistants create, preview, launch, and read tests, on Growth and above, currently labelled beta. Jeffrey told me he sees most brands as having a data problem more than an AI problem, and that framing is fair to the tool. The assistant is only as good as the order, behavior, and margin data you give it.
The most instructive price test in ABConvert’s public case studies is one where the first variant lost. Levain Bakery, the New York cookie brand, faced tariff driven cost increases and needed a price that covered them without losing customers. Their first test used a larger increase sized to the cost problem. Revenue per visitor dropped. Their second test used a smaller increase, and revenue per visitor rose slightly. They shipped the smaller one. Their VP of Technology summed it up as: “The first increase was too much, but the second one hit the right range.”
Here is how I read that. The team’s hypothesis was a cost question: how much do we need? The first test answered a customer question instead: how much will they accept? Those are different numbers, and the gap between them is where brands lose money in November. Without the losing test, the obvious move would have been to ship the larger increase, because that was the number the spreadsheet asked for. Jeffrey made the same point in his prep notes for our episode: the test that did not win is the one that kept them from an increase that was too aggressive.
The metric is what made the call possible. To see why, take an illustrative example, assuming a $40 product converting at 3.0%. That is $1.20 in revenue per visitor. Raise it 15% to $46 and suppose conversion falls to 2.5%: revenue per visitor drops to $1.15, a loss. Raise it 7% to $42.80 and suppose conversion only slips to 2.9%: revenue per visitor rises to about $1.24, a win. Judged on conversion rate alone, both increases lose. Judged on revenue per visitor, one of them is the answer.
A second case shows the upside. Bloon Paris, a French chair brand, tested a $50 increase on its US store for one month to offset customs costs. Conversion barely moved and gross margin per visitor rose almost 18%. They raised prices before their sales period so the Black Friday promotions ran on a healthier base. That is one brand’s result over one month, not a benchmark, but it is the exact sequence I would want a $3M brand to copy this year.
A brand above $2M can run two tests between now and Black Friday, launched together in separate visitor groups, and should run nothing new after about November 16. Jeffrey’s working rule is no more than three tests at once, and before peak I would cut that to two so each one gets enough traffic to resolve.
The price test should cover your top three to five revenue SKUs, with one variant sized to your cost problem and one smaller. That mirrors the Levain sequence, compressed into a single test so you are not out of calendar when the first variant loses.
The second slot goes to your free shipping threshold. Jeffrey’s approach is to look at where your order values cluster, not just the average, and test whether a higher threshold paired with a better offer pulls those orders up. The 2026 shopper data puts unexpected costs at the top of mobile cart abandonment at 46%, so the threshold is doing conversion work as well as order value work. Our older guide on setting a free shipping minimum from your median order value covers the math. On ABConvert, shipping tests require a Shopify plan with Carrier Service API access, so check that before you build it.
The freeze is the part brands skip. Testing guidance from GemPages on running reliable Shopify tests says not to start a test during an unusual promotion, and BFCM is the most unusual promotion of the year. Traffic mix, intent, and discount stacking all shift. Jeffrey’s own warning is that the most common way a test lies is being called too early, and a test that starts November 20 has no room to be called late.
ABConvert is the wrong choice for a store under about $1M a year that has never run a test, because a 10% win rate needs traffic and repetition to pay off, and the price testing you would buy it for starts at $199 a month. Jeffrey said as much in his intake for our show: the app is not built for fresh brands or teams that will not make testing a regular habit.
There are other limits a buyer should weigh. Checkout content, cart, and upsell tests require Shopify Plus. The MCP connection is in beta, and ABConvert’s own documentation still lists some test types, including checkout and visual editor tests, as setup through the app rather than through Claude. And a single 2025 App Store review complained about a ticket closed before the issue was confirmed fixed. ABConvert says it changed that workflow, and the other reviews praise support, but if hands-on help matters to you, ask about it on the demo.
Shopify Rollouts now schedules, gradually publishes, and A/B tests whole themes and checkout configurations, and it costs nothing extra on the Grow plan and up. If your question is whether a new theme or a checkout upsell block performs better, start there before paying for an app. It does not test prices, shipping, or offers, which is the gap the apps fill. Intelligems is the most established alternative for price and offer testing, and Shoplift publishes a solid walkthrough of how price tests should hold one shopper to one price. If you already run either one, do not add ABConvert on top of it before BFCM. Two testing tools in one store is premature complexity, and it is the pattern I have watched most often at this stage.
For transparency: Jeffrey is a guest on an upcoming eCommerce Fastlane podcast episode. That does not change this assessment, and there are no affiliate links in this article.
If you are doing $2M to $10M a year, the one thing to do this week is enter real product costs for your top five SKUs into whatever testing tool you use and launch a two variant price test by October 19. Size one variant to your cost problem and one smaller, judge on revenue or profit per visitor, and put November 9 on the calendar to read it. Whatever wins, rebuild your Black Friday discounts from that number, and use the guardrails in our BFCM profit leak breakdown so the new margin is not given straight back in a sitewide code.
If you are between $1M and $2M, run one test only, on price for your single best seller, and accept that it may not resolve in time. If it does not, keep your current price through BFCM and run the test properly in January. A flat result is still more information than you have today.
If you are under $1M, do not add a testing app this season. Jeffrey’s advice for smaller stores is one big swing test focused on quality, and I agree, but I would run it in January when there is no promo traffic distorting it. Before BFCM, spend the time on all in pricing at checkout and a clear free shipping threshold.
If you are above $10M and already run a program, your job in the next six weeks is the freeze. Write down the date no new tests start, tell the agency and the growth team, and hold it.
Raise prices before Black Friday if you are going to raise them at all, ideally by mid November. Every Black Friday discount is calculated from your base price, so an increase after the sale means charging recent discount buyers more in January, which feels worse to them than a change made earlier. Roughly 80% of shoppers start before Thanksgiving week, so a change in the first half of November reaches fewer buyers twice. Test the increase first on your top products, judge it on revenue or profit per visitor, and ship the version that holds up rather than the number your cost spreadsheet asks for.
A Shopify price test usually needs two to four weeks, and it should cover at least two full weekly cycles because weekday and weekend shoppers behave differently. ABConvert’s guidance says many stores see a readable result in 7 to 14 days, while smaller stores often need 2 to 4 weeks. Bloon Paris ran its $50 increase test for a full month. The most common mistake is stopping as soon as a winner looks significant, because early leads often reverse. Before Black Friday, start by about October 19 so the result is in hand around November 9.
A realistic win rate for Shopify A/B tests is around 10%, meaning about one variant in ten beats the original. ABConvert founder Jeffrey Lin says that is what the platform sees across its customers, with expert teams reaching about 15%. That rate is why testing only pays off with a steady rhythm of well prepared tests, rather than one test expected to change the business. It is also why a losing test is useful: Levain Bakery’s first price test lost, and that result is what pointed the team to the smaller increase they shipped.
No, Shopify Rollouts does not test prices. Rollouts schedules, gradually publishes, and A/B tests whole themes and checkout and customer account configurations, with experiments available on the Grow plan and above. It reports a fixed set of metrics and does not cover price, shipping, offer, or audience targeted tests. For those you need a dedicated app such as ABConvert, Intelligems, or Shoplift. Rollouts is a strong first choice if your question is about a theme or a checkout block, and it costs nothing extra on qualifying plans.
If you only have time for one test before BFCM, test price on your top revenue products. Price is the decision your whole promotional calendar depends on, and it has the largest effect on margin when a variant wins. Run two variants, one sized to cover your cost increase and one smaller, and judge the result on revenue or profit per visitor rather than conversion rate. If the test has not resolved by about November 9, keep your current price through Cyber Monday and run the test again in January rather than guessing during peak.
