How to measure the labour hours electronic shelf labels save in your own store

To know how many hours electronic shelf labels save in your store, you have to measure twice with the same method: before installation, how much of the working day goes on preparing, fitting and checking prices; and afterwards, what is left. The subtraction is easy. The hard part is that the first measurement is usually an impression rather than data.

This page is about that first measurement. The step-by-step breakdown of a price change and its cost in pounds is in the article on electronic versus paper shelf labels; here we deal with something else: getting timings that survive an argument with your finance director.

Why estimating from memory fails

Ask three store managers how long they spend each week on shelf labels and you will get three answers that differ by a factor of two or three. Nobody is lying. What people remember is the promotion changeover, when the job swallows the morning, not the thirty times it was done in two minutes between other things.

That is why two questions that tend to get blurred need to be kept apart:

• What one specific change costs (minutes per label, from the moment the price is decided to the moment it is verified). You measure this with a stopwatch.
• What share of the working day the whole pricing task takes, including gaps, re-checks, waiting and phone calls. You measure this by sampling.

If you only do the first, you multiply an ideal time by a number of changes and overstate the saving. If you only do the second, you do not know which step the system actually removes. You need both, and each answers a different part of the sum.

Measuring the whole task: work sampling

Work sampling is a classic work-study technique. Instead of following someone with a stopwatch, you observe them at randomly chosen moments and note what they were doing. The share of observations in each category estimates the share of time the activity takes. The condition, as the method is usually described, is that the moments are genuinely random and cover the normal range of tasks in the role.

In a store, the practical version looks like this:

1. Define five or six categories, no more: preparing price changes (printing, cutting, sorting), fitting or removing labels on the shelf, checking prices, replenishment, customer service and till, and other. The first three are what you want to isolate.
2. Generate the moments with a spreadsheet, not by eye. If you look when it suits you, you always land on the quiet moments or the busiest ones.
3. Record only what you see at that instant, without asking and without interpreting. If it fits no category, it is “other”.
4. Run the study over a period that includes a promotion week and an ordinary one. Two weeks is enough to see the difference.

How many observations

The error of a proportion estimated from n observations is the square root of p(1-p)/n, where p is the share of time the category takes. Solving for 95% confidence (the value 1.96), the number of observations is:

n = 1.96² × p × (1-p) / E²

where E is the absolute error you accept, expressed as a proportion. Applied to a task you suppose takes 5% of the day:

The useful reading is that a small task needs far more observations than it seems to before you can pin it down. If you suspect labels take 5% of the day and measure with 100 observations, your interval runs from roughly 1% to 9%, which cannot tell 2% from 8%. With a spreadsheet and a couple of people noting down for a week, 450 observations is manageable; 1,800 calls for automating the prompt, for example with a phone alarm.

Measuring the individual change: timing by step

Sampling tells you how much the task weighs; the stopwatch tells you what it is made of. Four decisions change the result.

Time the steps, not the total. If walking to the aisle weighs twice as much as fitting the label, that is the information that tells you how much a system removes and how much stays. With electronic labels the printing and fitting work disappears, but checking and replenishing do not.

Separate the situations. A single change mid-morning and forty changes in one aisle the evening before a promotion do not cost the same per label. One average gives a figure that matches no real situation.

Use the median. A change that went wrong (the item is missing, the label was stuck on badly) drags the mean up and does not represent the usual case. Log those separately too: how often they happen is a data point in itself.

Several people, several days. Someone who knows the aisle takes half as long as someone seeing it for the first time, and someone who knows they are being timed works differently. That last effect is known as the Hawthorne effect, and its size in the literature is uncertain; the 2014 systematic review by McCambridge and colleagues finds signs of an effect in most studies but concludes that better concepts are needed to study it. In practice: make sure the measurement does not change the routine, and discard the first ten minutes.

Recording template

One sheet per week of measurement, one row per batch of changes. These are the fields to capture, chosen so you can repeat the study later.

Three formatting habits that save arguments later: keep the raw data as well as the results, date every measurement, and write down what was and was not measured. In six months nobody will remember whether “fitting” included taking off the old label.

If staff are being observed, tell them what the study is for and that it measures the task, not the person. Use codes rather than names in the sheet; if you hold identifiable records of individuals, UK data protection rules apply to them.

What not to count

The measurement produces minutes; turning them into a saving is another matter. Five items that creep in often:

• Time that already overlapped. If an employee changes labels while restocking the same aisle, the time “saved” on that task does not free a whole hour, only the part that did not overlap.
• Slivers nobody can cut. Twenty minutes a day spread across four people does not become a reduction in headcount. It is capacity, useful if there is something to spend it on.
• Installation. Fitting the labels for the first time costs hours, but it belongs to the project, not to running costs. It goes in the up-front cost, not in the recurring saving.
• Pricing errors avoided. That is a separate benefit, worked out with a different measurement, for example a shelf-against-till audit like the one described in the article on a till price that differs from the shelf price.
• A national wage figure as if it were yours. The statutory floor is 12.71 pounds an hour from April 2026 for workers aged 21 and over, according to GOV.UK, and your real loaded cost is higher. Your contracts and on-costs rule; use a national figure only for a first approximation.

What to do with the numbers

With the baseline in hand there are three honest uses. First, put your minutes and your weekly changes into the savings calculator in place of the example values. Second, compare suppliers by the same yardstick: ask each which steps on your sheet they remove and which they do not. Third, repeat the study in a pilot project, with the same categories and the same people, so that the before and the after are measured identically.

And if the result is that the task takes 1.5% of the day, that is also a useful answer: the project will have to be justified on other grounds. To put it in context with the other costs, the electronic shelf label pricing guide covers the remaining items.