Chapter 2: The cost of bad data
Associates could get back nearly six working weeks a year with data they could trust
The hidden labor cost of bad data, in hours and dollars.
Nobody budgets for the time wasted due to incorrect data.
It never appears as a line item. It’s a restock that took longer than it should have, a recount nobody asked for, a walk to the backroom to find out whether the system was telling the truth.
Add it up and associates estimate they would get back nearly six working weeks a year each if the stock information driving their work was always accurate, hidden inside normal shift work for as long as the gap has existed.
4.4 hours
Recoverable per associate per week if stock information were always accurate, by their own estimate.
1 in 8
Roughly one working hour in every eight is recoverable, on a typical 31 to 40 hour week.
:format(webp))
Bad data never shows up as a line item, but retailers pay for it every day, and they pay first in labor. Every recount, every trip to the backroom, every 'let me go check' is an associate doing by hand what the systems should have done. Labor is the biggest cost a store controls, and associates are already stretched thin, so you can't afford to have them spend hours a week fixing the record. Capturing the shelf accurately does more than cut out-of-stocks. It gives those hours back to selling and serving customers.
Greg Buzek, President, IHL Group
Where the hours go
We asked associates how much time they would get back in a typical week if their stock information were always accurate. The average answer, across 2,200 of them, was 4.4 hours.
Recoverable associate hours per week and average by country
| Answer | Share |
|---|---|
| None | 4 |
| < 1 hour | 13 |
| 1 - 2 hours | 25 |
| 3 - 5 hours | 29 |
| 6 - 8 hours | 14 |
| 9 - 10 hours | 8 |
| 11 - 13 hours | 3 |
| 14 - 16 hours | 1 |
| 17 - 19 hours | 1 |
| > 19 hours | 2 |
| Market | Hours |
|---|---|
| Germany | 5.1 |
| US | 4.9 |
| Italy | 4.1 |
| UK | 4.0 |
| Canada | 3.9 |
| Poland | 3.8 |
| France | 3.2 |
Survey question: "If the stock information in your store were always accurate, how much time do you think you'd get back in a typical week?" Left: share of all associates giving each answer. Right: average hours by market.
While it’s an estimate, not a timesheet, it comes from the people who live with the problem every shift, and the shape of the answers is hard to argue with.
Most associates put it at three hours or more, and nearly three in ten say six hours or more: close to a full shift every week spent making up for stock information that’s wrong, searching for products the system says are there, recounting what it can’t be trusted on, and re-checking what it already claimed to know.
The total varies by market but never disappears. In every market the average is above 3 hours per week. Associates in Germany put it highest, at around 5.1 hours a week, with the US close behind at 4.9. France is lowest at around 3.2.
Associates asked what they’d change
A lot of times the back-stock isn’t in the system so I go and have to go physically check more often than I’d like.
Store Associate, US, Grocery Supermarket
What the recoverable hours are worth
These hours come with a price, and it scales with headcount.
Annual value of recoverable hours, by retailer headcount
Based on average US associate pay per hour ($18)
| Headcount | Value |
|---|---|
| 1,000 associates | 4.1 |
| 10,000 associates | 41 |
| 100,000 associates | 410 |
Calculated from the survey average of 4.4 recoverable hours a week, average reported US pay of $18 an hour, and a 52-week year.
Take a US associate. At the survey average of 4.4 hours a week and the survey’s average US hourly pay of $18, over a 52-week year, that’s around $4,100 a year per person.
A retailer with 10,000 store associates is looking at more than $41 million a year in recoverable labor hours; one with 100,000, more than $410 million. The largest grocers employ substantially more than that. The same hours cost a German retailer around €3,500 per associate a year, and a UK retailer around £2,700.
These figures are estimates, built from the hours and pay associates reported. They are not a savings target and not a case for fewer people. They are paid hours spent making up for a record that is wrong, on recounts, stock checks and note taking, when they could be completing other tasks.
The view from HQ. Asked about bad data's impact
:format(webp))
Inaccurate data made measuring on-time in-full at a store-level granularity much harder.
Grocery IT Director
:format(webp))
Unreliable inventory data has created a cascade of operational bottlenecks.
Retail Technology Partner
The tasks that only exist because the system can’t see the store
Store operations work is perpetual and even perfect data won’t change that.
Hours a week associates spend on each routine task
| Task | Hours |
|---|---|
| Customer questions tech should answer | 4.1 |
| Fixing product locations and layouts | 3.9 |
| Looking for gaps on the shelf | 3.7 |
| Picking online orders | 3.6 |
| Fixing prices and shelf labels | 3.3 |
| Finding products the system misplaced | 3.1 |
| Removing or marking down expired stock | 3.0 |
| Recounting or correcting stock records | 3.0 |
Survey question: "In a typical week, how much time do you spend on each of the following tasks?" Average hours per associate across all eight tasks asked about.
Shelves need filling, prices need changing, products expire. We asked associates how long eight routine tasks take them in a typical week. But three of them only exist because unlike the associates, the system can’t see the true picture of the store, creating wasted time, frustration, and inefficiency.
- Looking for gaps on the shelf: 3.7 hours
- Looking for products the system says are in stock but that aren’t where they should be: 3.1 hours a week
- Recounting or correcting stock records because the information can’t be trusted: 3.0 hours
The customer pays too. Associates spend around four hours a week answering customer questions they say store technology should have handled, many of them from a shopper who couldn’t find what the app or the shelf label promised.
:format(webp))
With agentic commerce, your brand and your products are going to show up for customers you’ve never been able to capture before, and you’ve got one shot. If the store data feeding that agent says the product they need for tomorrow’s trip is in stock, and they drive to your store and can’t get it, you haven’t only created a negative experience for a would-be customer. You also risk that disappointment being taken out on your already maxed-out store employees.
Anne Mezzenga, CEO and Founder, Retail Field Report
“Let me check in the back”: the interruption tax
Times associates are asked to check stock per week
| Answer | Share |
|---|---|
| > 10 | 22 |
| 6 - 10 | 18 |
| 3 - 5 | 35 |
| 1 - 2 | 19 |
| None | 5 |
| Sector | Checks |
|---|---|
| Drugstore | 6.7 |
| DIY | 6.3 |
| Grocery supermarket | 5.9 |
| Hypermarket | 5.8 |
| Department store | 5.7 |
| Convenience | 5.3 |
Survey question: "In a typical week, how many times does a customer or colleague ask you to go and physically check whether something is in stock?" Left: share of all associates. Right: average times a week by sector
Every physical stock check is a question the system was supposed to answer, and a task it could have prioritized instead.
Whether the screen said one thing and the shelf looked like another, or nobody checked the screen at all, someone walked to find out.
Associates are asked to go and physically check whether something is in stock around six times a week. More than one in five are asked more than ten times.
Over a year that’s roughly 300 trips per associate and a lot of wasted steps to confirm whether the record was right.
What Scandit has seen in stores
In one pilot, associates worked from a daily list of shelf gaps. About half were on products the system showed as having no stock anywhere in the store, backroom included. Walking to those gaps changed nothing: the fix was an order, not a restock.
Once they were sent to ordering and the list held only gaps with stock waiting in the backroom, the estimated time spent working it each day fell from 3.8 hours to under one. Same associates, same store, a list of only the work they could actually do.
It’s not a training problem
:format(webp))
It would be convenient if the lost hours were a training problem. It isn’t. Associates who had structured training, were consulted on new devices, apps, or software rolled out in their stores, and had the business impact of their work explained to them with numbers still put the recoverable time at around five hours a week. Those with none of the three advantages put it at around four hours.
The best-prepared associates aren’t losing less time to bad data; if anything, they report losing more, perhaps because people who know what accurate data should look like notice more of the gaps, and count them. Readiness, training, explaining why it matters, and consulting with the frontline, all lift tool use and whether the associates trust the data, as chapter 5 shows.
What this means for AI in store
An AI tool acting on the record doesn’t remove the hours associates spend making up for it. Every error the floor hasn’t caught becomes an order, a forecast, or a task that was wrong from the start, and another job for the same associates to put right. More automation in this scenario doesn’t shrink the bill, it compounds it until the record is built from what is actually in the store.