Markdown Recommendation: Check the Math Before Approving

The line on your screen

It's the morning walk. A line pops up telling you to take a markdown on the yogurt multipacks. It names a discount and a dollar figure you'll lose if you don't. You have about ninety seconds before someone needs you at the back door.

You can tap approve and move on. Or you can spend those ninety seconds checking whether the numbers behind the call actually match your store. This post walks through one markdown recommendation line by line, so you know what each number means, where it comes from, and when it's wrong.

This isn't about ranking a whole list of calls or building a weekly schedule. It's about one decision, read properly.

Why the math deserves a second look

Every markdown recommendation, whether it comes from software, a spreadsheet or your assistant manager, rests on the same small chain of inputs. How many units you have, how fast they sell, how long until they expire. Each later number is built from the ones before it.

So if one input is stale or wrong, everything after it is wrong too, and it looks just as confident. A miscounted shelf or a sell-through rate thrown off by last week's promo gives you a discount that's too deep or too shallow. The fix isn't to distrust the call. You just check the inputs, because you know things the records don't.

The worked example

Worked example: picture a store with a yogurt six-pack on the shelf. The recommendation reads like this:

• Units on hand: 14
• Daily sell-through: 0.4 units/day
• Days left: 5 (best-by Friday)
• Units that won't clear: ~12
• Projected loss: $25.20 at cost
• Suggested discount: 30% off, starting today

Here's what to check on each line, in order.

Units on hand

This is the line that's wrong most often, and the cheapest one to check. The number comes from your stock records. Those are only as good as the last receiving entry, the last count, and whether anyone logged the two crushed packs that went in the bin on Tuesday.

For example, if the record says 14 and you walk over and count 9, the whole call shrinks. Nine units at 0.4 a day over 5 days means about 2 sell and 7 don't. You still need a markdown, just a smaller problem. Count the shelf and the backstock if there is any. If the count is off, fix the record before you approve, or the next call will be off too.

Daily sell-through

This is the rate the product actually moves at, usually averaged over recent weeks. Ask yourself whether those weeks were normal.

For example, a 0.4/day rate that includes a week when the item was out of stock for three days will understate real demand. A rate that includes a two-for-one promo will overstate it. If you know the history was unusual, adjust the rate in your head and redo the multiplication. It's the single input where your memory beats the record.

Days left

Simple, but check which date is being used. Is it the best-by on the pack, or your own pull date, which may be a day or two earlier? If you pull dairy a day ahead of the printed date, there's a day less to sell than the call assumes.

Also check that it's the right batch. If two lots are sitting on the shelf with different dates, the call should be about the one expiring first. If the dates are mixed up front to back, the rotation needs fixing before the price does. The daily expiry routine covers where that check fits in the morning.

Units that won't clear

This is the first number that's worked out, not pulled from the records, and you can check it in your head.

For example: 0.4 units/day × 5 days = 2 units expected to sell at full price. 14 on hand − 2 expected sales = 12 that won't clear. If the screen says ~12, the math holds. If it says something wildly different, one of the three inputs above isn't what you think it is. Go find which one.

Projected loss

This line tells you what doing nothing costs. Check two things: is it priced at cost (what you paid) or at retail, and is the cost current?

Worked example: 12 units × $2.10 cost each = $25.20. If your supplier raised the case price last month and the record still shows the old cost, the real exposure is higher. A loss quoted at retail looks scarier but mixes up lost margin with money actually spent. For a write-off decision, cost is the honest number.

The suggested discount

This is where you're most likely to override, because it's the call that leans hardest on a guess about how customers will behave.

For example, a 30% markdown on a $3.49 pack makes it about $2.44, still above the $2.10 cost. For the call to work, the discount has to lift sell-through from 0.4 a day to roughly 2.4 a day so that all 12 extra units go in 5 days. Ask yourself:

• Has a discount this size moved this item before, or similar items in this aisle?
• Does the marked-down price stay above cost, or are you choosing a smaller loss over a bigger one? (Sometimes that's the right choice. Just make it on purpose.)
• Would starting smaller and going deeper on day three recover more? Step-down markdown schedules lay out that approach, and dynamic markdown pricing covers how the depth gets set.

When to override

Approving isn't the default. It's a decision. Override, or at least adjust, when:

• Your shelf count disagrees with units on hand. Fix the record first.
• You know something the history doesn't: a holiday weekend, a school closing, a competitor's promo, weather that drives people to stock up.
• The item is a traffic driver you'd rather not be seen discounting, or there's a better outlet, like a donation partner or a return-to-vendor credit.
• The discount crosses below cost when a smaller cut would probably do the job.
• The product is about to be pulled anyway because of quality, not date. No discount fixes a bruised or swollen pack.

When you override, write down why in a line. A notebook or the day-old & markdown rotation log works fine. Over a few weeks those notes show you whether your gut or the math was right more often, and on which categories.

A ninety-second habit

Run every markdown recommendation through the same quick sequence:

• Count the shelf. Does it match units on hand?
• Think back. Was the sell-through period normal?
• Check the date. Right batch, right pull date?
• Multiply. Rate × days, subtract from on-hand. Does it match units that won't clear?
• Check the cost. Is the loss at current cost?
• Judge the discount. Will it move this item, and is the depth deliberate?

A notebook can't do the first pass of this across hundreds of items every morning. Nobody has time to multiply sell-through by days left for every dated SKU. What a person can do, and should keep doing, is the shelf count and the judgment call at the end.

Where ShelfSense fits

That split is how ShelfSense by ShelfLifePro is designed. It's a daily agent that sits alongside the system you already run. Once a day it reads your stock, batch, cost and expiry records. It surfaces the few markdowns worth acting on, ranked by money at risk, with the arithmetic shown on every line, like the example above. It recommends, you approve, and it never changes a price itself. Over the weeks it learns from which approved calls sold and which still went to waste.

If you want to see how each number on a recommendation gets built, here's how ShelfSense thinks. To try it on your own data, start with a CSV export of your stock and expiry dates. ShelfSense's free audit reads one CSV export, no account needed.