Position is a number you can measure

The sentence in the floor guide

Everyone in retail knows the front of a shop is worth more than the back. Almost nobody can say how much more. So the best places on a floor get handed out on seniority and instinct. This lesson turns that belief into a measurement. A handful of garments are moved around a real shop on a careful schedule. Then each place is read off in money.

Wrayburn's visual merchandising guide, which Tamsin Landseer wrote and every shop follows, contains this line:

Hero product goes to the front table.

Nobody at Wrayburn could say what the front table is worth. Not roughly. At all. The guide had been in force for six years, every shop obeyed it, and the number on the other side of the sentence had never been produced. That is normal, and it is why visual merchandising is so often argued in adjectives. When the quantity does not exist, taste is the only currency left in the room.

Bexholme has four kinds of selling position, and this lesson is about measuring the difference between them.

The obvious test, and the reason it lies

The obvious test is to put Cottrane on the front table for a week and count what it sells, then put it back on the four-way for a week and count again, and divide.

That test tells you almost nothing, for two reasons that have nothing to do with visual merchandising.

The week is not a constant. Bexholme counted 3,790 people through the door in week 1 and 4,370 in week 3, a 15% swing between two ordinary weeks in the same phase. A line that gained 15% by moving to the front table in week 3 gained nothing at all.

One line is not the shop. Whatever the tee does at the front, you have learned it about a GBP 16.00 tee.

The design Wrayburn used

Four lines, four positions, four weeks, arranged so that every line sits in every position exactly once, and every week contains all four positions exactly once.

Two rules made the test worth running.

Presentation quantity was held at 40 units in every position. Otherwise the test would have moved depth and position together — the front table holds 90 and the wall bay 42 — and no reading could be attributed to either. Forty units clears the presentation minimum from lesson 1 on all four fixtures.

Nothing else was allowed to change. No price, no promotion, no window, no new colour landing mid-test.

Here is what the tills recorded.

Reading it: the shop's index

Average each position down its own four readings.

No footfall adjustment was needed for that table. That is the entire reason for arranging the weeks the way they were arranged. Each position was measured once in week 1, once in week 2, once in week 3 and once in week 4. Whatever the weeks did, they did equally to all four positions, so the comparison between positions is clean without touching the numbers.

At Bexholme, then, the front table sells 2.19 times what the main walkway sells, and the side aisle sells 0.53 times it. A line moved from the side aisle to the front table sells roughly four times as much. That is the number the floor guide had been missing for six years.

Reading it again: the index belongs to a line, not to a shop

Now take one line at a time. A single line's four readings sit in four different weeks, so this time the footfall does have to come out. Each reading is divided by that week's footfall as a share of week 2, then compared with the same line's own walkway reading.

The shop average of 2.19 sits inside a spread from 1.59 to 2.59, and that spread is not noise-shaped. It is ordered by price, and by how considered the purchase is. The front table converts attention a shopper had not planned to give: a GBP 16.00 tee is bought on that attention, and a GBP 69.00 cardigan is not. Read the side-aisle column the same way in reverse. The cardigan loses least by being pushed into the side aisle, because the customer who wants it goes looking.

A hero position multiplies impulse. It barely moves a considered purchase. Give it to the line whose customer has not decided yet.

Presentation is part of the position

One more thing hides inside P1, and lesson 3 turns on it.

The Bexholme front table is a folded-stack table. Everything measured at P1 was measured folded, on a table, in the first five metres of the shop. That is not a place on the floor. It is a place and a way of showing goods, and the two cannot be separated in the reading.

A tee folds. A shirt folds. A cardigan folds. All three were in the test. Nothing that hangs was.

What the index is actually for

An index is only useful when it turns into money. Multiply it back out.

The best available occupant of the Bexholme front table is Cottrane: 34 a week on the walkway, index 2.59 at P1, so about 85 to 88 a week on the table, at GBP 10.00 of gross margin. Call it GBP 880 a fixture-week. That is the front table's rent: what Wrayburn gives up by putting anything else on it.

Which gives the test that lesson 3 is about. A move is two moves, and you have to price both:

A line earns a hero position when what it gains by moving in is more than what the incumbent loses by moving out.

gain = (new index − old index) × rate of sale × GM a unit

loss = (incumbent's old index − incumbent's new index) × its rate of sale × its GM a unit

The cheap half: adjacency

Position is charged to somebody. Adjacency is very nearly free, and Wrayburn tested that too, in the same four weeks.

In weeks 1 and 2, Yelverdon trousers stood on a four-way in the trouser run, and Cottrane tees were two fixtures away. In weeks 3 and 4, the tees moved onto the two-arm beside the trouser four-way, styled as the outfit both appeared in in the campaign. The till data:

Four extra tees a week is GBP 40.00 of gross margin a week, and it cost nothing. The tees were going to stand somewhere regardless, and the two-arm was already there.

On 86 transactions, though, it is suggestive rather than proved. A two-sided exact test on those counts gives about 0.054, the wrong side of the line most people draw, on one shop and four weeks. Wrayburn kept the adjacency because it is free, not because it demonstrated anything. That is an honest reason to keep something, and a dishonest one to write a policy about.

Help me design a rotation test to measure what my selling positions are worth. I will tell you how many positions I want to test, which lines I can rotate, and how many weeks I have.

Build me a schedule in which every line sits in every position exactly once, and every week contains every position exactly once. Show it as a grid of line by week. Then explain in one paragraph why that arrangement means I do not have to adjust the position averages for weekly footfall, and why I DO have to adjust when I read a single line's own positions against each other.

Then list what I must hold constant for the test to mean anything: presentation quantity, price, promotion, window, and any new colour or size landing mid-test.

Finally, be blunt about the limits of what I will get. Tell me how many readings I will have per cell, whether position and fixture type are confounded in my design, and which of my conclusions I will not be entitled to draw from one shop and one season. If my design cannot support a conclusion I want, say so rather than qualifying it.