In short
Learning goal: by the end of this lesson you can design a cart test that tries to raise order value, judge it on revenue or profit per visitor, and plan it with the traffic the cart actually gets.
Who this is for: Shopify brands doing roughly $1M to $10M a year with a cart page or cart drawer that most buyers pass through.
What the cart is for
By the time a shopper opens the cart, they have chosen at least one product. Two things can happen next: they buy what's there, or they leave. Every cart test tries to add a third: they buy a bit more.
That's why cart tests are risky. Anything that adds a decision (an add-on, a bundle, a threshold to reach) can raise the order value of the shoppers who buy and lower the number who do. Extra costs are a known reason for leaving: leaving aside shoppers who were just browsing, "extra costs too high (shipping, tax, fees)" was the most common reason US online shoppers gave Baymard Institute for abandoning an order, at 40% (Baymard, list updated 22 September 2025, read 12 October 2026). A cart test that makes the shipping cost more visible, or harder to avoid, plays with exactly that.
What to measure at this step
| Metric | Definition | Use it as |
|---|---|---|
| Revenue or profit per visitor | Revenue (or gross profit) from shoppers who saw the cart, divided by those shoppers | The primary goal |
| Average order value (AOV) | Revenue divided by orders | A secondary goal: explains an RPV change |
| Checkout-start rate | Share of cart viewers who start checkout | A diagnostic |
| Attach rate | Share of orders that include the offered add-on or bundle | A diagnostic |
| Threshold reach | Share of orders at or above a free-shipping minimum | A diagnostic |
| Checkout completion | Orders divided by checkouts started | A guardrail |
AOV alone is not the goal. A test can raise AOV by losing small orders: fewer shoppers buy, the ones who do spend more, and revenue per visitor falls. Revenue per visitor catches that (revenue per visitor vs conversion rate). Profit per visitor goes further and subtracts the discount you gave to get the larger basket.
Test ideas, each with its mechanism
Free-shipping threshold. Mechanism: shoppers add items to avoid a shipping fee. It works when the gap to the threshold is small for many baskets and there's an obvious item to add. It costs you the shipping on every order above the line, including the ones that were already there. Part 6 of the BFCM series, free-shipping thresholds for BFCM baskets, has the break-even maths, and the free-shipping threshold calculator does it for your numbers.
Progress bar. Mechanism: makes the gap visible ("$12 away from free shipping"). It only helps if the threshold exists in checkout, and the bar's amount must match what checkout gives in every currency you sell in. Benchmyrk's shipping tests count the minimum on the cart subtotal before discounts; a Shopify free-shipping discount counts discounted prices, so a $100 basket at 20% off counts as $80 there.
Cart add-on. Mechanism: one relevant, low-price item that completes the purchase (batteries, a care kit, a refill). It works when the add-on is genuinely related. A random best seller is a distraction.
Bundle. Mechanism: a reason to buy two or three things together, usually a discount on the set. It raises AOV when the set is something shoppers wanted; it gives margin away when they'd have bought the set anyway.
Order offer. Mechanism: a spend tier ("$10 off $100") rewards a bigger basket. Judge it on profit per visitor: the discount comes out of margin on every qualifying order. Part 4, Black Friday offers judged on profit, has the break-even table.
Attach rate vs revenue per visitor
An add-on test always "works" on attach rate: if you show an add-on, some shoppers take it, and the control shows no offer at all. That is not a result. The questions are whether the shoppers who took it spent more in total, and whether showing it cost you any orders.
Read attach rate to understand what happened. Read revenue per visitor, or profit per visitor, to decide. Check checkout completion as a guardrail: an add-on that slows the cart down on mobile can cost more orders than it adds revenue.
Plan with the cart's own traffic
Only visitors who see the cart enter a cart test. That's a small share of your sessions, and revenue goals need more visitors than conversion rate because order values vary a lot.
Illustrative inputs, not data from a real store. Computed with the same sample-size formula Benchmyrk uses to plan revenue goals (two-sided, 95% confidence, 80% power, two versions, even split).
Inputs: 1,100 shoppers a day view the cart (lesson 1's example store), 18.6% of them buy, average order $82 and a spread of $55 between orders.
| Smallest change in revenue per visitor to detect | Visitors per version | Days |
|---|---|---|
| 10% | 10,667 | 20 |
| 5% | 42,667 | 78 |
| 3% | 118,520 | 216 |
A cart test at this traffic can detect a 10% change in revenue per visitor in about three weeks. Most add-ons and thresholds move revenue per visitor by less than that. So test changes big enough to matter (a threshold, a bundle offer), not small ones (the color of the progress bar). If a 10% change isn't plausible, the likely outcome is "inconclusive", and you should decide in advance what you'll run in that case. The sample size and test duration guide explains the inputs.
Common problems
- Cart drawers vs cart pages. Many themes open a drawer instead of a page. Test where shoppers actually see the cart, and check that the change survives the drawer re-rendering after each add.
- Markets and currencies. A threshold or progress bar in one currency must convert, or be set, for every market.
- Discount stacking. An add-on discount, a bundle discount and a shopper's own code can combine in ways you didn't plan. Check the combinations at checkout before launch.
- Two tests on the same cart. A shipping test and an offer test both change what the shopper pays. Run them one at a time, or as one combined test.
- BFCM baskets. Sale-week baskets differ. Test thresholds and bundles in a normal period and decide before the sale.
Exercise (10 minutes, on your own store)
- Take the last 28 days of orders. Write down the median and average order value.
- Count orders within $15 of your free-shipping threshold (or of a round number like $100, if you have none). That's the group a threshold or progress bar can move.
- Pick your three best sellers and write down the most natural add-on for each.
- Write one hypothesis with its mechanism: "Offering [add-on] in the cart for [product] buyers will raise revenue per visitor, because it completes the purchase."
- Count your daily cart viewers and check a 10% change in the sample-size calculator.
You don't need Benchmyrk for any of this.
Do it in Benchmyrk
Check yourself
- An add-on test shows a 12% attach rate and revenue per visitor is inconclusive. What do you report? Inconclusive. Attach rate is a diagnostic, and the control shows no offer, so some attach rate is guaranteed. Revenue per visitor decides.
- AOV rose 8% and revenue per visitor fell 3%. What happened? Fewer shoppers bought. The ones who did spent more, but not enough to make up for the lost orders.
- Why plan a cart test with cart viewers, not store sessions? Only shoppers who see the cart enter the test. Planning with all sessions would make the test look several times faster than it is.