◆conversionworks

Recommendations rail

"Picked for you" on the radio-controlled.co.uk homepage, since 24 Oct 2026 · data to 10 Nov 2026 · refreshed 12 Nov 2026

People shown the rail
458
1,151 homepage loads, 713 sessions
People who scrolled to it
60
13.1% of people shown
Click-through, people
6.6%
30 of 458 people clicked a card
Click-through, loads
4.2%
48 clicks on 1,151 rail loads

Does it pay? Shown the rail against the holdout

Conversion: placed an order
Holdout (no rail)
9.4%
12 of 127 people
Shown the rail
5.0%
23 of 458 people
Too early: under 25 per groupp = 0.06
Added to basket
Holdout (no rail)
19.7%
25 of 127 people
Shown the rail
17.0%
78 of 458 people
No significant difference yetp = 0.49

This is the only fair comparison: the same kind of visitor, split at random by client ID. Everything counts from a person's first homepage visit in their group, whether or not they clicked. Of the 30 people who clicked a card, 5 went on to order, but people who click are already keener, so that says little about what the rail itself caused.

GroupPeopleCartedBoughtConversionOrdersRevenueRevenue per person
Shown the rail
Have browsing to personalise on; rail shown
45878235.0%24£4,343£9
Holdout
Have browsing to personalise on; rail withheld at random
12725129.4%13£1,934£15
No signal yet
Nothing to personalise on; no rail. Not comparable
7611150.7%5£602£1
By day
DatePeople shownRail loadsScrolled into viewClicksClicks per load
10 Nov 202634541035.6%
9 Nov 20262959711.7%
8 Nov 202648941933.2%
7 Nov 20263763911.6%
6 Nov 20262656800.0%
5 Nov 202631561058.9%
4 Nov 20263063221015.9%
3 Nov 202636541123.7%
2 Nov 202635751711.3%
1 Nov 20263461923.3%
31 Oct 20263054300.0%
30 Oct 20262871011.4%
29 Oct 20263592077.6%
28 Oct 20263154000.0%
27 Oct 202641710912.7%
26 Oct 20263583000.0%
25 Oct 20264570034.3%
24 Oct 20261521000.0%

From the GA4 BigQuery export, refreshed daily. Test browsers excluded; the Woo orders themselves are not reconciled here, so GA4 may miss a few. A person's group is fixed per browser, so one customer on two devices counts twice.