Retail margin is easy to ask for and easy to get wrong, because a bed is sold as a set but recorded as separate stock lines. Ask "what is our margin on Orthopaedic Supreme Firm" and there are three defensible answers depending on what you mean. This article explains which number you get, and when to ask for a different one.
The three margin questions
"What did we earn this month?" — the period figure. Every sale in the month, minus every return in the month, whether or not the return relates to something sold this month. This is the number that ties to the income statement, and it is the one to use for "did the price increase work". Ask for retail margin by month.
"What did this month's sales ultimately deliver?" — the cohort figure. It follows this month's sales forward through their own returns. It cannot be known until the returns have finished arriving, so it keeps moving for about 90 days. Useful for judging a range; wrong for judging a month.
"What did we make on that bed?" — the set figure, described below.
These are all legitimate and they do not agree. Mixing them is the single most common way a margin conversation goes wrong, so ask for whichever you mean by name.
What counts as a bed set
A bed set is a mattress and the base sold with it on the same invoice. The pairing comes from the article master, where each mattress records the base it belongs with. Where the customer takes a different base — which happens — the system falls back to matching on size.
A storage base is not a bed base. It is furniture, it is sold as an upsell, and it earns its own margin — several points above a bed. It is therefore reported on its own and never folded into a bed set, because a storage base sitting inside the set figure quietly lifts it and makes beds look better than they are. Where a mattress went out on a storage base and nothing else, it is reported under its own heading rather than being counted as a bed set or as a mattress-only sale — it is neither.
Roughly one mattress in six is sold without a base, because the customer already has one. Those are reported separately rather than being forced into a set, and they consistently earn several points more margin than the same mattress sold inside a set. That is the intended design, not an anomaly — set prices are built to be competitive, so a mattress sold on its own carries more of the margin.
This matters when you read a set-level margin figure: it covers only the sales that formed a set. The mattress-only sales sit outside it and earn more, so a set-level number is the cautious view of a range, not the whole of it. Ask for both if you are judging the range.
Giveaways are allocated into the set
When a headboard or pillow is given away to close a bed sale, the margin on that item alone looks terrible and the bed looks better than it was. Neither is useful. Instead, the giveaway is absorbed into the set it helped close, so the set carries its true economics and you see what the deal actually earned.
Discounts work the same way. The till posts a discount against the invoice; it is then spread across the lines in proportion to what each line actually sold for, and stored per line. So an invoice with a discounted mattress and a free headboard shows the discount where it belongs.
An accessory sold on its own — someone buying a pillow and nothing else — stays a standalone sale. It was never part of a bed.
Returns
About one rand in nine of retail sales comes back, and roughly seven in ten of those are exchanges rather than refunds — the customer swaps rather than walks away. That distinction matters commercially even though both reduce margin identically, so returns are reported split by type.
Two things are worth knowing about returned cost:
- When goods come back, the cost is not credited back to stock. A mattress that has been slept on cannot be resold, so the cost is charged to a returns cost account instead. This is correct, and it means returned stock genuinely costs you money rather than being recovered.
- A return can be traced back to the specific bed it came off only when the original sale is still within the reporting window and was itself part of a set. Today that reaches about four returns in ten by value. The rest are reported at range level with the untraced portion shown explicitly — it is never quietly spread across sets to make the numbers look tidy.
Because of this, month-level and range-level return figures are complete; set-level ones are partial and labelled as such.
Judging a price increase
Ask for the effect of a price change at the level the price was set. Bed prices are decided as a set price covering mattress and base together. When that increase is loaded, one price gets split across two stock lines, and the split is a bookkeeping choice — it is not the decision.
So a per-product report can show a base "unchanged" while the set it belongs to went up 8%. Both figures are correct; only the set-level one reflects what was actually decided. If you are asking whether an increase worked, say by set and you will get the answer to the question you meant.
The other thing to check is timing. A price set against one month's costs but selling from two months later carries whatever cost movement happened in between, and the legs of a set do not always move together. Comparing delivered cost against the cost the pricing model assumed — rather than against today's standard cost — is what separates "we priced it wrong" from "costs moved after we priced it".
There is a third possibility that neither of those catches: the set you priced is not always the set you delivered. Where the customer takes a different base from the one the price was built on, that set carries a higher cost than the model assumed. The report separates this from ordinary cost drift, because the two have completely different remedies.
But be careful how you read it. A higher cost on those sets is not automatically a loss — if the upgrade is also being charged for, the two cancel out and the margin is unaffected. So the report shows the substituted sets bridged all the way through to delivered margin, not just their extra cost. Only if their delivered margin is genuinely worse is there something to fix. Ask for the shortfall split by cause and carried through to margin, never as a single cost number.
Then go one step further, because this is where the analysis is most often wrong: ask for it split by which base actually shipped. "Substituted" lumps together two opposite things — a customer who upgrades to a dearer base and pays for it, and a customer who receives a dearer base at the same price. Averaged together they can look like nothing is happening while a real cost is being absorbed on hundreds of sets. The general rule for reading any of these figures: keep splitting a number until the groups underneath it stop disagreeing with each other.
And when a split does turn up a group behaving in the opposite direction to the rest, ask a second question before you explain it: does that group belong in the figure at all? A storage base was the answer here — not a mispriced upgrade but a piece of furniture that was never a bed. Splitting a number checks the arithmetic; only someone who knows the products can tell you the boundary is wrong.
Finally, say what you are comparing against. "We are a point short" means nothing without a baseline, and the answer genuinely changes depending on whether you mean against last quarter or against where the business ran a year ago. Ask for the margin trend alongside the figure.
What the numbers refresh from
Set-level figures rebuild automatically overnight from the previous 90 days of trading, so the recent past keeps improving as more returns land against their original sales. Figures older than that window are settled and will not move.
If a number looks wrong, the fastest check is whether you are comparing a period figure to a cohort figure, or a per-product figure to a per-set one. That accounts for most surprises.