FoxrFoxr
Post-purchase funnels

Split tests

Run A/B and multivariate tests on a post-purchase offer, read the results, and promote the winner.

A split test shows different versions of one offer to different customers, so you can find out which one earns more. Each offer slot can run one test at a time.

There are two kinds:

TestUse it toHow it works
A/B Test (recommended)Compare 2–5 completely different offers — different products, designs, or copy.Each version is a separate layout.
Multivariate Test (advanced)Fine-tune one offer — its discount, timer, shipping, or starting quantity.You list values to try, and every combination gets an equal share of customers. Needs more traffic.

Create a test

On the funnel's Upsells tab, click Create test on the offer you want to test. The offer becomes Layout A — your control.

Step 1 — Select the type of test. Choose A/B Test or Multivariate Test and click Next.

Step 2 — Set up the variations (see below) and click Next.

Step 3 — Name your test. Pick something you'll still recognise in a month, like Free shipping vs 20% off. Click Create test.

The offer card is replaced by a test card with the status Not started. Click Start.

Save the funnel. The test only starts once the funnel is saved.

A/B test: add layouts

  • Click Add another offer and choose what the new layout sells — Specific products, Automatic upsell, Most expensive item in cart, Least expensive item in cart, or a copy of another offer in the funnel.
  • Click Edit offer on a layout to design it in the usual offer editor.
  • Use the Duplicate layout icon to start from a copy, and Remove layout to drop one. Layout A can't be removed.
  • Each layout card shows the share of customers it will get.

You need at least 2 layouts and can have up to 5.

Multivariate test: add variables

Tick the parts of the offer you want to vary, and add up to 4 values for each. Every value is tested against the offer's current setting.

VariableWhat you enterExample
DiscountsPercentage or fixed amount values. Percentages can't go over 100.20, 30 (tested against the current 15%)
TimerWhole minutes.3, 10
ShippingA shipping charge. 0 means free shipping.0, 4.99
QuantityThe quantity the offer opens with.2, 3

The number of combinations is every value (plus the current setting) of every variable multiplied together. A test can split traffic across at most 24 combinations. For example, 2 discount values and 1 timer value make 3 × 2 = 6 combinations.

The wizard also shows how many impressions it recommends — about 2,500 per variation — before you trust the result.

How traffic is split

  • Only a Running test splits traffic. While a test is Not started or Paused, every customer sees Layout A.
  • Each checkout is assigned to one variation, evenly across all of them, and keeps it for the whole post-purchase visit.
  • Assignment is per order, not per customer, so a returning customer can see a different variation next time.

The test card

PartWhat it shows
StatusNot started, Running, or Paused.
Start / Pause / ResumeControls the test. Save the funnel after clicking.
ProgressImpressions so far against the recommended number, and how many variations are being tested.
Variables and LayoutsWhat's being tested.
⋯ menuEdit test or Delete test.
See test analyticsResults per variation.
Select winnerEnds the test with the variation you choose.

While a test exists, you edit its offers through Edit test, not the usual Edit offer.

Read the results

Click See test analytics. Each variation — for example Layout B · 20% off · Free shipping — has a row with Views, Conversion, Revenue, and Per view (revenue per view). The figures cover the whole life of the test.

The variation with the highest revenue per view gets a badge:

  • Ahead so far — no variation has 100 views yet. Treat it as a hint, not a result.
  • Leading — the best performer among variations with at least 100 views.

Pick a winner

Click Select winner. The current leader is pre-selected.

Choose the variation you want to keep and click Use this variation.

Save the funnel.

The winning variation — including any discount, timer, shipping, or quantity values it was testing — becomes the slot's normal offer, and the test is removed. The analytics you collected are kept.

Don't call it too early

If no variation has 100 views, the app warns you that the differences are still mostly noise. Let the test run longer for a result you can trust.

Delete a test

Choose ⋯ → Delete test. Layout A stays in the slot as its offer and every other variation is discarded. The analytics you collected are kept.

Next

What customers see — the post-purchase page from the customer's side.

On this page