Ad 707 was the top evergreen spender. A copy pointing at the homepage instead of the Tomato collection would show whether one less click to the product helps or hurts.
The landing page only. Same image, copy, audience and ad set.
Inconclusive. Stopped after nine days.
The homepage version got 77 link clicks in nine days. Meta kept the budget on the established ad, and the Tomato season was fading, so Ad 707 was paused on 19 September.
A landing page test inside a shared ad set does not get enough traffic to read. The new version needs its own budget, or a test on an ad that is still rising rather than one at the end of its season.
Retesting the homepage on Ad 1000 (above), the current top ad, with a plan to ring-fence the homepage version if it is starved.
Each day's line is the whole test so far, so it settles as the numbers grow
Shows whether every version is getting its share
The count behind the cost per purchase
The earliest signal: does the ad earn the click
| Version | Spent | Link clicks | CTR | Purchases | Cost per purchase | Purchases per click | ROAS |
|---|---|---|---|---|---|---|---|
| Tomato collection page | £1,307 | 1,084 | 1.34% | 25 | £52.30 | 2.31% | 1.25x |
| Homepage | £168 | 77 | 0.70% | 3 | £55.87 | 3.90% | 1.24x |
| Day | Tomato collection page spend · clicks · purchases | Homepage spend · clicks · purchases |
|---|---|---|
| 13 Sept | £215 · 213 · 5 | £1.87 · 0 · 0 |
| 14 Sept | £102 · 83 · 1 | £22.21 · 10 · 0 |
| 15 Sept | £117 · 64 · 3 | £39.83 · 17 · 1 |
| 16 Sept | £308 · 222 · 4 | £57.69 · 27 · 2 |
| 17 Sept | £277 · 217 · 3 | £2.30 · 1 · 0 |
| 18 Sept | £283 · 281 · 8 | £12.17 · 3 · 0 |
| 19 Sept | £4.46 · 4 · 1 | £5.70 · 3 · 0 |
| 20 Sept | £0 · 0 · 0 | £10.17 · 7 · 0 |
| 21 Sept | £0 · 0 · 0 | £13.54 · 9 · 0 |
| 22 Sept | £0 · 0 · 0 | £2.12 · 0 · 0 |
Spend, clicks and purchases are as reported by Meta for the ads in each test, from the test's start date to today, which is partial. Where versions share an ad set, Meta decides how much each one spends, so compare cost per purchase and purchases per click rather than totals. A test is only called once its rule is met; stopping early on a good week is how false winners get made.