Bedding and sleepwear switch-off: how we counted, and how to set it up
For the analyst who wants to know how a before-and-after test was read, what each window held, and what the brand had running before it switched the agent off.
- Every definition behind the numbers on the case study
- The setup in the store, step by step
- What to check in your own data
Open the method and setup notes
Enter your work email and the notes open on this page. We also send you the link, so you can come back to it.
Thanks. The notes are open below.
We use your email to know who reads our research. Privacy policy
What was measured
Period and population
A European bedding and sleepwear brand on WooCommerce with 67,000 products, with Meta as its main acquisition channel. The switch-off ran in stages from 26 June to 15 July 2026. The store comparison sets the last two weeks of full service, 15 to 29 June, against the two weeks of fewer recommendations, 30 June to 15 July. The Meta comparison sets 1 to 15 June against 1 to 2 August, with a later check in early September.
What counts as a result
Items per order, average item price and average order value, all from completed store orders. Recommendation impressions per day come from the agent. In the ad account: purchase ROAS, cost per purchase, purchases per day, spend per day, CPM and CTR, as relative changes.
How an order is credited
A June order counts as "with a recommended product" when it contains a product the shopper took from a recommendation; "from an ad only" when it followed an ad click and nothing else; "no touch" when it had neither.
Controlled or observed
A before-and-after test, not a holdout. The brand switched the agent off; nobody was held back. The season was ruled out by the brand's 2025 pattern: the summer dip began on 16 June in 2025; in 2026 the account held its June level until 15 July, the day the script was removed.
What the numbers do not say
Absolute ROAS is not shown, because the ad account counted 2.11 purchase events per real order. Relative changes inside the same account are unaffected by that.
The terms on the case study
| Term | How it is defined here |
|---|---|
| Items per order | The number of products in a completed order, averaged over the window. It fell 10.2% while the average item price moved 0.9%. |
| Average order value, indexed | The average order in the window, with the first window set to 100. Fewer items per order explains 93% of its fall. |
| Recommendation impressions | A recommendation block shown to a shopper, counted per day. They fell 37.1% between the windows; banners on the same pages rose 8.7%. |
| Order with a recommended product | A June order containing a product the shopper took from a recommendation. 31% larger than a no-touch order, 51% larger than an ad-only order. |
| Purchase ROAS | Revenue Meta credits to the campaign divided by spend, as a relative change only. −33.7% between the windows. |
| CPM and CTR | Cost per thousand impressions and click-through rate: what Meta does before the click. Both held (−1.4%, +2.9%), so the loss sits after the click. |
| Half-life | The time in which the remaining effect halves: 15.9 days, measured on 60,000 users. The effect faded; it did not stop. |
What was switched on, in order
What ran before the test
Recommendations on the site, the abandoned-cart and abandoned-browse emails, and the audience and catalog sync into Meta. Coverage rose from 54% of sessions in April to 90% in June.
26 June: bulk email stops
One-off campaigns stop. The automations with personal recommendations keep running.
30 June: recommendations cut back
Recommendation impressions fall 37% over two weeks. Banners on the same pages run normally, so the site itself works.
15 July: the script is removed
Every personal function stops, on the site and in the inbox. From here the Meta account is read against June.
The read-out
Two equal store windows for the basket; two ad-account windows for the return; a check in early September; the 2025 season as the control for timing.
To check it in your own store
Pull items per order and average order value by week from your platform, and recommendation impressions from the agent. If the basket moves while the item price holds, the change sits in what shoppers add, not in what they pay.
In your ad account, read ROAS, cost per purchase, CPM and CTR over the same weeks as relative changes. If CPM and CTR hold while ROAS falls, the loss is after the click. Compare the same weeks of the year before to rule out the season. A holdout group gives the same answer without switching anything off.
See which of last week's visitors you missed
Thirty minutes, with your store and ad accounts open. Then 30 days free. A holdout group decides: if the agent doesn't add orders in 30 days, you don't pay.
Book 30 minutes
You'll talk to one of us.