What cross merchandising is
Cross merchandising is placing two items from different categories next to each other so that buying one prompts buying the other: salsa on a clip strip beside the tortilla chips, bagged salad in the case next to the rotisserie chicken, lime wedges on the beer cooler. The tactic is old and the vocabulary is settled. What is not settled, in most stores, is which pairs actually work.
Every merchandiser has a list of pairings that feel obvious. Some of them are worth a display and some of them are two popular items that were always going in the same basket, and the difference is worth real money in a fourteen-store chain, let alone a large one.
The pairs a buyer expects, and what the baskets say
Ridgeline Markets is an illustrative fourteen-store grocery operator running 61,400 baskets a week. Pull the candidate pairs off the transaction file and score each one on lift, which is how much more often the two appear together than their individual rates would predict.
| Pair | Baskets with both | Support | Lift | Buyer's intuition |
|---|---|---|---|---|
| Pasta and jarred pasta sauce | 2,410 | 3.9% | 2.7 | strong |
| Beer and salty snacks | 1,880 | 3.1% | 1.4 | strong |
| Tortilla chips and jarred salsa | 1,290 | 2.1% | 3.1 | strong |
| Rotisserie chicken and bagged salad | 970 | 1.6% | 2.9 | weak |
| Bananas and peanut butter | 640 | 1.0% | 1.1 | strong |
| Cold brew and protein bar | 415 | 0.7% | 2.2 | none |
Two rows in that table are the reason to run the analysis at all.
Beer and salty snacks has the second-largest raw co-occurrence in the set and a lift of 1.4, which is barely above chance. Both items are in a lot of baskets. Putting them together buys very little, because the shoppers buying both were already buying both.
Bananas and peanut butter is the pairing every merchandiser suggests and it lifts at 1.1, which is noise. It sounds right and it is not there.
Meanwhile rotisserie chicken and bagged salad lifts at 2.9 despite nobody naming it, and cold brew and protein bar lifts at 2.2 off a small base, which makes it the interesting test rather than the obvious build. The mechanics of support, confidence and lift, and where each one misleads, are covered in market basket analysis.
Testing one adjacency
A lift number identifies a candidate. It does not prove that moving the item will change anything, because the shoppers may already be finding both. So test it: build the salsa clip strip on the tortilla chip endcap in seven stores, leave seven alone, and read four weeks.
| Store group | Salsa units/store/wk before | After | Change |
|---|---|---|---|
| Test (7 stores) | 118 | 141 | +19.5% |
| Control (7) | 121 | 124 | +2.5% |
The raw test change is 19.5%, but the control moved 2.5% on its own, so the adjacency is worth 16.6%, not 19.5%. That is 19.6 incremental salsa units per store per week, or 549 units across seven stores in four weeks.
Note what did not happen: tortilla chip units were flat in both groups. A clip strip lifts the item on the strip, not the item it hangs from, which is the usual result and the reason cross merchandising is scored on the attached category rather than the anchor.
What the facing cost
Roll the test out to all fourteen stores and the salsa side is straightforward. At 19.6 incremental units per store per week, 52 weeks and fourteen stores, that is 14,269 units. Salsa retails at $3.79 with a 34% margin, so $1.29 of gross profit a unit, or $18,407 a year.
Then subtract the thing that lost the space. The clip strip displaced a single-serve nut facing that was doing 4.1 units per store per week at $2.99 and a 54% margin, $1.61 of gross profit a unit. Fourteen stores over a year is 2,985 units and $4,806.
| Line | Gross profit |
|---|---|
| Salsa gained, 14 stores, 52 weeks | $18,407 |
| Nuts displaced | -$4,806 |
| Net | $13,601 |
Still a good decision, and roughly a quarter smaller than the number the salsa figure alone would have supported. Any cross merchandising program that reports only the lifted item is reporting a gross number as if it were net.
Where cross merchandising goes wrong
No space is free. Every clip strip, shipper and case-end takes a facing from something. If the displacement is not measured, the program looks better than it is.
Occasion pairs travel, category pairs do not. Chicken and salad is one dinner decision that happens to ring as two lines. Bananas and peanut butter is two separate errands. The first responds to proximity and the second mostly does not.
Compliance decays. A clip strip that nobody refills is a facing you paid for and are not using. Cross merchandising has a higher execution failure rate than a planogram change because it lives outside the set and outside the reset cycle.
Dayparts hide inside the average. A pair lifting 2.2 across the week may be 4.0 in the morning and 1.0 at night, which changes where the display should go, not whether it should exist.
Where Scout fits
Scout reads the transaction file, so it can rank candidate pairs by lift across your own stores, size the incremental units after a test, and hold the displaced item in the same report so the net number is the one you see. It models the assortment and space decision. It is not a planogram or space-planning system: it will tell you the salsa clip strip is worth $13,601 net and it will not publish the set to your stores.
The short version
- Cross merchandising pairs items from different categories in one location so that buying one prompts buying the other.
- Rank candidates on lift, not on co-occurrence. Beer and salty snacks had the second-highest raw count in the worked example and a lift of 1.4.
- A lift number is a candidate for a test. Test against control stores, or the category's own drift gets counted as your result.
- Always net off the displaced facing. In the worked example $18,407 of salsa gross profit became $13,601 once the nuts that lost the space were counted.