Falling sales make a markdown feel like action. It is action. It’s just usually action on the price stage, taken before you know whether the broken stage is traffic, conversion, basket, or local demand.
Call the control point sales decomposition: store sales are traffic × conversion × basket. That isn’t decorative retail algebra. It tells you whether fewer people are arriving, fewer arrivals are buying, or buyers are taking less home.
A broad markdown can make one of those numbers look briefly less ugly while taking a bite out of gross-margin dollars. Very helpful, if your objective was to hide the diagnosis from yourself.
The Split
Start with the boundary. Footfall means people or visits counted at a stated location boundary under a stated convention.
A door counter that counts entries and exits as visits, or counts your staff as shoppers, damages the traffic measure. It doesn’t rewrite the POS record. Passing pedestrians are further upstream still. Only some passers-by come in.
Keep the three operating measures separate:
- Traffic → door entries or purchasing parties under one documented convention → opportunity to transact.
- Conversion → purchasing parties divided by that same traffic denominator → whether the visit becomes a transaction.
- Basket → merchandise sales per transaction, with items per transaction alongside it where the mix matters → what the converted visit is worth.
The corpus’s day-29 store illustration has 400 passers, 52 entries, 10 purchasing parties, and 13 items sold. Those numbers answer different questions.
Ten transactions divided by 52 entries is a store-conversion measure. Thirteen items divided by ten transactions is an items-per-basket measure. Dividing your online revenue by 52 door entries isn’t a creative version of either. It’s a category error with a spreadsheet attached.
For the last eight weeks, pull these measures by store area or source before you approve a broad price change.
Keep your web orders separate from store transactions even when both draw on one inventory pool. In the rain-jacket cohort, three website orders reduced available stock without creating a store entrant. A shared SKU is not permission to blend the funnel.
The Board
Use one board to make your boundary visible. The cells are control instructions, not universal benchmarks, and the source and denominator travel with every rate.
| Store area or source | Footfall source | Conversion | Basket | Sell-through interval | Markdown | Shrink | Rent-to-sales |
|---|---|---|---|---|---|---|---|
| Store doorway | Door counter with a stated entry, visit, and staff convention | Purchasing parties ÷ door entries | Merchandise sales and items ÷ transactions | Receipt cohort over a stated selling window | Approved price instruction in POS, shelf, and site | Physical count against book stock | Store sales attributed under the lease definition |
| Street or centre | Passer count or area index, labelled as upstream opportunity | Do not calculate from street counts | Store transactions only | Same receipt cohort; do not assign passers to units sold | No automatic response from an area index | Counted at the store, not in the street series | Test the local sales denominator against fixed occupancy |
| Website | Sessions under a separate channel rule | Online orders ÷ online sessions | Online merchandise sales ÷ online orders | Shared cohort, but online units identified separately | Website price updated with the approved change | Pick, carrier, return, and warehouse events kept distinct | Channel attribution follows the lease and management policy |
The distinction matters because outside traffic can diverge sharply by place. In the cited UK series for September 2025, footfall was down 4% overall while town and city centres were up 4% and district or local centres were down 10%.
That series is built from aggregated device data and excludes people identified as living or working in the retail area, so it isn’t interchangeable with your door counter. It’s a clue about opportunity around your location. It isn’t an excuse to stop counting the threshold.
Which Stage Broke
Use the shape of the movement to find the stage that failed. The purpose is to stop your markdown meeting from becoming a ritual sacrifice of margin.
Sales fall; conversion and basket hold → traffic or catchment is broken
Compare your own door traffic with customer-origin records, commuter patterns, demand generators, barriers, and nearby substitutes.
A catchment is not a circle you drew around the shop. The relevant burden includes travel time, delivery, delay, and inconvenience. Median card-transaction distance in the cited research ran about 4 km for food stores, 12 km for eating and drinking, and more than 20 km for durable goods. The point is variation by occasion, not a radius you can paste onto every retail plan.
Traffic holds; conversion falls → offer, assortment, capacity, or local competition is broken
Inspect availability, broken size or colour runs, advice, queueing, fitting-room coverage, replenishment, and what your customers can buy nearby.
Research cited in the corpus found average in-store conversion of 43% across eight brands and four location types. That sample is context, not a target. A specialty shop, a convenience store, and a showroom don’t share the same customer mission.
Traffic and conversion hold; basket falls → mix, attachment, or achieved price is broken
Check items per transaction, full-price versus markdown mix, returns, and whether a high-value item has quietly left your basket.
More transactions can coexist with fewer margin dollars. The till is perfectly capable of applauding the wrong thing.
Markdowns rise; margin dollars fall → price response is masking the traffic diagnosis
With an invented $40 landed cost and $100 ticket, the spread is 150% markup on cost but 60% gross margin on sales.
Selling at $80 after a 20% discount cuts gross-margin dollars from $60 to $40. A third of the original dollars, gone. A lower ticket may release stock. It doesn’t make your lost margin reappear.
Sell-through improves after a count adjustment → the denominator is broken
Your cohort needs units received, units sold at each price, units remaining, returns, and adjustments.
In the 24-jacket example, 18 sales and five ending units can display 78.3% sell-through while sales against the original receipt are 75.0%. The missing unit is not a customer.
Rent percentage rises; rent cheque holds → fixed occupancy is being mistaken for a rent increase
In the corpus example, $90,000 of annual occupancy on $900,000 sales is 10%. A 15% sales fall produces $765,000 sales and 11.8% occupancy against the same rent.
At a 40% gross margin, that decline also removes $54,000 of gross profit. Your denominator moved. The landlord didn’t suddenly become more expensive in arithmetic.
The Price File
Markdown is a commercial choice, not a demand measurement system. Give it an owner.
A usable delegation names the merchandise scope, permitted reduction, effective dates, and exception authority. Your approved change then reaches the POS, the shelf label, and the ecommerce page together. Preserve former price, new price, time, location, and reason.
Don’t use a former price as a mood board. Under 16 CFR 233.1 a former price has to be bona fide — openly offered in the regular course of business for a reasonably substantial period . No universal number of days is supplied.
“Was” and “save” are evidence claims. If your record doesn’t support the comparison, advertise the current price and don’t invent a before story.
The next layer is inventory. A markdown changes achieved price. It doesn’t change the historic landed cost attached to your unit. Shrink changes usable quantity without a sale. A physical count finds your book-to-physical gap; it doesn’t establish theft.
The cited NRF average of 1.6% shrink is a multi-brand, multi-location figure, so treat it as context rather than a default for your store.
The reason code, count evidence, approver, and posting time are your diagnosis. A generic shrink code is where a receiving failure goes to retire.
The Catchment
Local demand is not city population with a nicer map.
Read your store against the people who can actually substitute: customer origins, delivery addresses, enquiry logs, referrer sources, workplace movement, competing centres, and barriers such as a highway or a bad crossing. A retail park, a central high street, and a neighbourhood centre can lose different traffic in the same week. They run on different trip purposes.
Don’t let the local story become a tidy excuse either. More nearby customers attract more competitors. Density can compress the upper end of prices as easily as it widens choice.
Pair any area signal with your own threshold data and conversion. If traffic is down across the centre but your conversion holds, protect price while you test the catchment. If traffic holds and conversion fails, fix the offer before you turn every label red.
Traffic tells you how many chances the store receives. Conversion tells you whether the store uses them. A markdown can move units. It cannot tell you which of those two broke.
Count your threshold. Read your catchment. Price the evidence, not the panic.