Retail pricing strategies that hold together

The price file nobody decided on

A 96-store grocery operator could tell me the margin on any single item in about four seconds, and could not tell me why the item beside it was priced where it was. That is the normal condition. Most retail pricing strategies exist as 41,000 individual decisions sitting in a price file, each defensible on its own, none of them adding up to a structure anyone chose. Blended margin lands where it lands, and every quarter somebody proposes fixing it with an across-the-board increase.

I spent most of my career on the other side of that table, arguing one rung of a ladder I could only see a sliver of. What I learned late is that the buyer was not negotiating my price. She was defending an architecture: a set of zones, a short list of items shoppers actually watch, and a margin mix that had to land on a number by the end of the year. The individual price was the output.

This page is that architecture from the retailer's seat. It covers four decisions in the order they have to be made, with the arithmetic shown. The supplier-side view of one category's ladder lives on category pricing and price architecture, and this page does not repeat it.

What a retail pricing strategy has to decide

Four decisions, and they are not interchangeable. Getting the order wrong is how a chain ends up with a coherent-looking rule set producing an incoherent shelf.

Everything downstream inherits from these. A promotion is a temporary departure from the architecture. A markdown is a scheduled exit from it. A price-optimization model, covered separately on price optimization models, is a way of searching inside the constraints these four decisions define, which is why a model bolted onto an undecided architecture produces recommendations merchants override.

Step 1: build price zones from measured response

A price zone is a group of stores that share a price. Most chains have them and most chains inherited them, usually from a market map drawn before half the current store base existed.

The rebuild is not complicated. For each store, you already hold the two things that matter: how the store's shoppers responded the last time a price moved, and whether the chain's own competitive flag says there is a discounter within the trade area. The first is measured, the second is maintained by hand and usually accurate. Cluster on those, not on geography.

Here is what the 96 stores looked like after the rebuild.

Read the two index columns against each other rather than down the page. The gap between zone 1 and zone 4 on key value items is 6 points. On everything else it is 12. That spread is the entire zone strategy: the items shoppers compare stay close to parity everywhere, and the items they do not compare carry the geographic difference. A chain that widens both columns equally has not built zones, it has built four price levels and told its most competitive market it is expensive.

The measured half is what makes the zones defensible. Across a fixed 200-item basket, base elasticity in zone 1 came back at -2.1 and in zone 3 at -0.9. Same items, same weeks, same chain. Zone 3 shoppers are less than half as price-responsive because they have nowhere convenient to go, and pricing them identically to zone 1 gives away margin to buy units that were never at risk. For how to measure that number without contaminating it, see price elasticity.

Two rules keep zones from rotting. Re-cut them annually rather than continuously, because a store that flips zones mid-year produces a shelf-price change nobody can explain to the store manager. And never let a zone ordering invert: if zone 3 is meant to be the highest, no item in zone 3 may price below zone 1, no matter what a category-level rule says. That inversion is the single most common thing a competitor screenshots.

Step 2: pick key value items on comparability, not just frequency

A key value item is an item whose price the shopper remembers and compares. The standard method is to rank by trip penetration, take the top 200, and call them KVIs. Frequency is necessary and it is not sufficient.

The soft drink is the instructive row. It appears in more trips than the rotisserie chicken and it is not a key value item, because a single-serve bottle bought cold at the register is not a price a shopper carries between stores. Nobody drives across town for it and nobody remembers what it cost. It sits at 41% margin precisely because it fails the second screen.

The second screen is comparability: can a shopper actually put your price next to somebody else's without effort. Three things make an item comparable, and an item needs at least two of them. It is sold in an identical form everywhere (a gallon of whole milk, a dozen large eggs). It is bought on a plan rather than on impulse. And it is priced in a unit the shopper can convert without arithmetic, which is why per-pound produce qualifies and a 4-pack of anything usually does not.

Run both screens and the list comes out short. At this operator it came out at 320 lines including size variants, which is 0.8% of the file carrying 18% of sales. Those 320 lines are where the chain's price image is decided, and the discipline that follows is simple to state: their price is a strategic choice reviewed on a calendar, and no automated rule is allowed to move them.

Step 3: set the margin mix across the basket

This is the step most chains skip, and it is the one that turns four decisions into a number the CFO can hold you to. Assign every line to a role, then let the roles carry different margin rates on purpose.

Every row is share of sales times the margin rate, so the last column sums to the blended margin: 2.16 + 4.62 + 14.26 + 4.24 + 2.08 = $27.36 per $100, or 27.4%. The zone table above arrives at the same place from a different direction: (0.25 x 24.8) + (0.45 x 27.1) + (0.25 x 29.6) + (0.05 x 31.4) = 27.4%. Two cuts of the same chain reconciling is the check that the architecture is real rather than a slide.

Now the reason to build the table at all. Suppose finance wants 37 more basis points of blended margin next year. There are two ways to get them.

Take it on the background group, which is 46% of sales: a move from 31.0% to 31.8% delivers 46 x 0.008 = 0.37 points. That is roughly a 1.2% shelf-price increase spread across 33,600 lines nobody price-checks.

Take it on the KVIs, which are 18% of sales: you need 18 x 0.0206 = 0.37 points, so the margin rate has to go from 12.0% to 14.1%. At constant cost that is a 2.4% price increase on every item shoppers actively compare, applied to the 320 lines that set what the store feels like. Same 37 basis points, and one of the two is invisible.

That comparison is the whole argument for having roles. Without them, an across-the-board increase takes both, and the chain pays for the background margin with its price image. It is also why the premium and impulse rows are worth growing as a share of sales: at 38.5% and 52.0%, every point of mix that moves toward them lifts blended margin with no price change anywhere. Mix is the only margin lever the shopper cannot see.

Step 4: write the rules that keep the file coherent

Rules are what stop 41,000 independent decisions from drifting apart again. Five earn their place.

Price families. Items that differ only in a dimension the shopper treats as free carry one price. Six flavors of the same yogurt at the same size is one price, always. A file that lets flavor drift produces the shelf tag that reads $1.29, $1.29, $1.39, $1.29 and teaches shoppers the price is arbitrary.

Pack ladders. Price per unit of measure must fall as pack size rises, without exception. Here is a real break from the same file:

The 32 oz costs more per ounce than the 20 oz, and the state-mandated unit price on the shelf tag says so in print. Moving the 32 oz to $5.19 restores $0.162 per ounce and the ladder reads correctly. This is the retailer's half of the price pack architecture decision, and it is checkable mechanically, which means there is no excuse for a violation surviving a cycle.

Margin floors and cost pass-through. Every line carries a floor below which no rule may price it, and a policy for what happens when cost moves. A 4% cost increase on a background item passes through in full at the next cycle. The same increase on a KVI does not pass through automatically, because the whole point of the KVI list is that its price is a decision rather than an output. Chains that pass cost through uniformly discover their price image moved without anyone approving it.

Endings. Pick the endings the file is allowed to use and enforce them. The 9-ending effect is one of the better-evidenced findings in pricing, and the sourcing for it sits on the elasticity page. What matters here is consistency: a file that mixes .99, .95, .89 and .00 has thrown away a signal for nothing.

Suggested retail price is an input, not a rule. Suppliers publish a suggested retail price, and following it is a choice the retailer makes item by item. Follow it on background items where the supplier's price research is better than yours and the item carries no image. Ignore it on KVIs, where your zone structure and your shoppers' measured response beat a national recommendation. Where a supplier enforces a minimum advertised price, that is a genuine constraint rather than a suggestion, and it belongs in the rule set as a hard floor.

Where retail pricing strategies go wrong

Cost-plus on everything. A single margin target applied to every line produces the architecture by accident. It prices milk against its own cost rather than against the shopper's memory, and it prices a differentiated premium item at the same rate as a commodity that has four substitutes on the same shelf.

Zones that never re-cut. A store that opens across from a new discounter in March is still in the isolated zone in December, at index 109 on background items, quietly losing the basket rather than the item.

Letting the promoted price become the reference price. Run a KVI at a promoted price 30 weeks out of 52 and the everyday price stops being the reference. The shopper's remembered price is the deal price, and the everyday margin on the item is fiction. This is the retailer-side version of the trap described on discount pricing.

Reading dollars without units. A category growing dollars while units fall is a category taking price faster than demand supports. The dollar line looks like a pricing win for two or three quarters and then private label takes the gap.

Rebuilding the architecture and never auditing the file. Rules that are not checked are suggestions. The pack-ladder break above survived four cycles in a file everyone believed was governed, because nothing was actually comparing price per ounce across sizes. The same discipline that catches it is what a pricebook audit does on the item file generally.

Where Scout fits

Scout sits on the chain's own POS, which is what makes the measured half of this possible. It cuts the zones from observed response rather than from a market map, computes base elasticity per zone per item, holds the role assignment and the margin-mix table as a standing view rather than a quarterly rebuild, and checks the structural rules against the live file so a pack-ladder break surfaces in the cycle it appears rather than four cycles later.

Scout also carries the execution end: it can hold the item file, maintain cost and retail, and push the resulting price file to the POS, so the architecture you designed is the one that reaches the shelf tag.

The honest limit is what the data is. POS is transactions, not people. Scout can tell you that zone 1 shoppers respond at -2.1 and zone 3 shoppers at -0.9, because it watched what the registers did. It cannot tell you who those shoppers are, and any pricing story that leans on shopper demographics needs loyalty-linked data or a survey rather than a POS read.

The short version

• Decide zones, item roles, margin mix and structural rules in that order. The individual shelf price is the output of those four, not an input to them.
• Zone on measured response. Base elasticity of -2.1 in the competitive zone against -0.9 in the isolated one is what justifies a 12-point background index spread and a 6-point KVI spread.
• Screen KVIs on comparability as well as frequency. The 20 oz soft drink is in 11.6% of trips and still carries 41% margin, because nobody compares it.
• Take margin where the shopper is not looking. Thirty-seven basis points costs a 1.2% increase on background lines or a 2.4% increase on the 320 items that set your price image.
• Retail pricing strategies fail on coherence far more often than on level. A file where the 32 oz costs more per ounce than the 20 oz is not a pricing mistake a shopper forgives, and no zone or margin decision repairs it.