We are suggesting changes to one matched half and leaving the other half alone so your own till numbers tell us what happened, rather than us guessing. This week, consider the suggestions in Arm A and the solo section, and leave every Arm B price exactly as it is.
These are the suggested first changes. On the cautious basis they are worth about £52 a year in total. That figure is not the price rise multiplied by today's units: VAT comes off first, then the cost of the stock, then the margin on the units we expect to stop selling.
| Line | Till code | Current | This week | Sold per day |
|---|---|---|---|---|
| Pasta spirals small packThis is a first step. A second step may follow after the read. | 9900000000001 | £2.51 | £2.61 | 5.25 |
| Still water small packThis is a first step. A second step may follow after the read. | 9900000000021 | £0.84 | £0.89 | 3.88 |
| Butter small packThis is a first step. A second step may follow after the read. | 9900000000016 | £2.21 | £2.31 | 2.85 |
| Chocolate biscuits small packThis is a first step. A second step may follow after the read. | 9900000000012 | £2.67 | £2.77 | 3.92 |
| Apple juice small packThis is a first step. A second step may follow after the read. | 9900000000008 | £1.77 | £1.87 | 4.17 |
| Long grain rice small packThis is a first step. A second step may follow after the read. | 9900000000002 | £1.62 | £1.72 | 2.43 |
Arm B is the held comparison. Changing even one of these prices weakens the read. Holding it back leaves about £63 a year on the table for now, on the same cautious basis.
| Line | Till code | Current | This week | Sold per day |
|---|---|---|---|---|
| Fruit sweets small pack | 9900000000010 | £1.31 | Leave unchanged | 5.53 |
| Tomato soup family pack | 9900000000019 | £2.73 | Leave unchanged | 5.03 |
| Sparkling water small pack | 9900000000007 | £2.80 | Leave unchanged | 5.10 |
| Tomato soup small pack | 9900000000003 | £1.41 | Leave unchanged | 3.88 |
| Sparkling water family pack | 9900000000023 | £2.98 | Leave unchanged | 2.75 |
| Long grain rice family pack | 9900000000018 | £1.70 | Leave unchanged | 4.02 |
Each line here is too large to hide inside an arm, so it gets its own read on 3 November 2026. A suggestion is still the owner's call.
No line is large enough to need a separate read.
We will compare 56 days before and after the changes. The held arm helps us judge whether the changed arm moved differently from ordinary trading. It does not promise an outcome, and it does not identify customer response before the read.
Arm A before-and-after: 22.50 units per day, about 1,260 units across 56 days, with a noise floor of about ±4.0%. This is the before-and-after floor for this arm, not the threshold for comparing Arm A with Arm B. It cannot tell us what happened on each line.
Arm B before-and-after: 26.31 units per day, about 1,473 units across 56 days, with a noise floor of about ±3.7%. This is the before-and-after floor for this arm, not the threshold for comparing Arm A with Arm B. It cannot tell us what happened on each line.
Arm A against Arm B: the combined one-standard-error comparison floor is about ±5.4%. About ±10.6% is a rough 95% reference before overdispersion or department confounding, not a conclusion threshold. We will read the result with those limits rather than treating one standard error as proof.
Residual arm difference: yearly value 17.5%; daily units 14.5%; department coverage 4 in Arm A and 3 in Arm B, with 3 departments shared. These are recorded limits, not hidden precision.
0 lines fell below one unit per day. Together their yearly gap is £0 at current volumes, before any customer response. They are worth doing later and are not being ignored. An eight-week read is too noisy to judge them honestly in this batch. At the boundary, one sale a day gives 56 units in the window and about ±18.9% noise on a before-and-after ratio, so it cannot support a line-level conclusion.
0 lines, with a yearly gap of £0 at current volumes, before any customer response, carry a different-product warning. They are not safe to put into this experiment until the matched names have been checked, so they are disclosed here but not allocated to either arm.