Skip to content
allanninal.dev/build
Collaborate

Part 5 of 7 · Shelf price checker series ~5 min read

The gap between the till and the shelf

The first full walk finds the backlog. Every walk after it is about something smaller and more useful: how long a price change takes to reach the shelf, which changes never make it, and whether the morning list is in an order anybody can actually follow.

Key takeaways

  • Every price change opens a clock, and a photo of the new label stops it.
  • Order the swap list by the store’s walking route, not by product code.
  • Changes keyed outside the normal price change night are the ones that lag.
  • Walk the changed bays weekly and every bay monthly. One catches lag, the other drift.
  • Report lag by store, not a count of wrong labels, which falls when fewer bays are walked.

From a price change to a confirmed label

A price change followed from the till to a confirmed shelf-edge labelA horizontal row of five boxes joined by arrows. Price change, keyed in the till system with a time. Till updated overnight, in every store at once. Swap list, ready by half past five in walking order. Label swapped and photographed in the bay. And Clock stopped, when the read agrees with the till. A note says the clock stops on a photo that reads the new price, not on somebody ticking a list, because a tick records only that a person meant to swap the label.EVERY PRICE CHANGE OPENS A CLOCKPrice changekeyed in the tillsystem, with a timeTill updatedovernight, everystore at onceSwap listby 05:30, inwalking orderLabel swappedand photographedin the bayClock stoppedthe read agreeswith the tillThe clock stops on a photo that reads the new price, not on somebody ticking a list. A tick records that a person meant to swap the label.
Fig 1. One change as one line. The system always knew when the first box happened; the last box is the one nobody used to record.
  • Machine learning
  • Management
  • Analytics
  • Front-end & mobile

The swap list is a route, not a report

A price change list sorted by product code, or by department, sends somebody from aisle two to aisle nine and back to aisle three. Stores already walk themselves in a fixed order, for fills and for date checks, and the bay numbers from the walk give the checker that order for free. The morning list starts at the first bay the person reaches and ends at the last, with overcharges marked so they are not left for the afternoon.

Each line carries what to look for as well as what to print: the old price that should be coming down, the new price going up, and the unit price that has to be on it. When the label has been swapped the person photographs the bay, and the line closes from that photo rather than from a tick.

How long the labels actually took

One week of price changes grouped by how long the new label tookFour vertical bars showing one week's price changes at the largest store, grouped by when the new label was confirmed by a photo. Same morning accounts for one hundred and twenty-six changes. Within two days accounts for forty-one. Within a week accounts for seventeen. Not by the next walk accounts for twenty-eight. A note says the last bar was not random: nineteen of its twenty-eight changes were keyed outside the Tuesday night run and never appeared on a Wednesday morning list.050100150200~126Same morning~41Within two days~17Within a week~28Not by next walkPrice changes in one week at the largest store, by when the new label was confirmedThe last bar was not random. 19 of its 28 changes were keyed outside the Tuesday night run, and never appeared on a Wednesday morning list.
Fig 2. One week’s 212 price changes at the largest store, grouped by how long the shelf took to catch up.

Out-of-cycle changes are the ones that lag

The chain changes prices on Tuesday night, and the process around that night is good. In the first week measured, 126 of the week’s 212 changes had a confirmed new label the same morning. What it was not built for is the change keyed on a Thursday afternoon because a supplier cost moved, or the Friday markdown on a line that is not selling.

The store still printed its list once a week, on Wednesday morning, so those changes never appeared on one. Of the 28 changes still showing the old label at the next walk, 19 were keyed outside the Tuesday run. The fix was not more effort. It was a list on every morning that had a change behind it, however short, and a clock on each change rather than an assumption about each cycle.

How often to walk

A full walk of the largest store took two people about two and a half hours. Doing that every week is not realistic, and it is not needed. The measure that matters week to week is lag on changes the system already knows about, and that only needs the changed bays photographed: in the largest store, about a hundred and twenty photos on a price change morning rather than 583.

The full walk still earns its place once a month, because some faults are not caused by a price change at all. A label knocked onto the wrong product during a fill, a promotion label from a cycle nobody remembers, a line delisted with its label left behind: none of these open a clock, and only walking every bay finds them.

What the lag record holds, per change and store

  • Changed at. When the till price version took effect.
  • Cycle. Whether it was keyed in the scheduled run or outside it.
  • Listed at. When it first appeared on that store’s swap list.
  • Confirmed at. The time of the photo that read the new price.
  • Lag. Confirmed minus changed, in hours.
  • Still open. Changes with no confirming photo yet, and how old they are.

What the manager actually reads

Not a count of wrong labels. That count goes up when more bays are photographed and down when fewer are, which quietly rewards walking less. What each store manager gets weekly is lag: the share of changes confirmed the same morning, the oldest change still open, and the number of overcharges found on the monthly walk that no price change explains.

That last figure is the one worth watching. It measures the shelf drifting on its own, mostly through labels moved during filling, and it turns a vague request to check the labels into a specific conversation about which aisles and which shifts.

The next post prices it and the one after gives the service names, the tables and the IAM. The bill is dominated by one thing, the vision read per photo, which is exactly why the weekly walk covers only the bays that changed.

All posts