Measuring the Sales You Never Made

A missing sale is not an observed transaction

When an item is unavailable, low sales can mean constrained supply rather than weak demand. The counterfactual is what might have sold if the customer had found the product. That quantity must be estimated; a till record alone cannot prove it.

Atlas provides the scoped sales and merchandise context for the investigation. Core Plus supports size and location stock decisions. This article describes an independent estimation method, not a promise that a particular automatic lost-sales model or causal attribution is available in every configuration.

Choose comparable available trading before extrapolating

Use a period with relevant price, exposure, store and product scope. A promoted Saturday is not automatically a fair comparison for an ordinary weekday. Check whether the item was genuinely unavailable or merely absent from the selling shelf while usable stock remained elsewhere.

The worked example assumes five comparable selling days, with the item available on three and unavailable on two. Thirty units sell on the three available days. That gives an observed ten-per-available-day rate for the chosen evidence window.

Keep estimated demand separate from recorded sales

Applying the ten-unit rate to the two unavailable days gives twenty estimated unserved units. Combining them with the thirty observed sales suggests fifty units of unconstrained demand over the five-day window. The twenty are not transactions, receivables or recognised revenue.

At an assumed $60 price, their possible sales value is $1,200. With a $35 unit cost, possible product gross profit would be $500 before other expenses. Both financial amounts depend on those units actually being sold at the stated price; they are not a confirmed loss or recovery.

Test whether substitution changes the category effect

Some shoppers may choose another available product. In a separate sensitivity, assume eight of the twenty potential missed purchases are instead served by a substitute. The net category-unit effect would then be twelve, not twenty, under that attribution assumption.

The article does not claim eight substitutions were observed or caused by the gap. A style-level availability estimate and a category-level incremental-sales estimate answer different questions. Keep them separate before using the result to justify a larger category buy.

An available-day rate can also be sensitive to which days are chosen. A slower comparable rate of eight units per day would imply sixteen unserved units rather than twenty; twelve per day would imply 24. Treat that range as a sensitivity, not a confidence interval inferred from three observations.

Inspect the size and timing of the proposed response

An estimate of twenty unserved medium units does not justify buying twenty assorted units or moving stock after the window closes. Core Plus supports reviewing the matching size, usable source, destination and arrival. The source’s retained cover matters alongside the destination’s opportunity.

If the period has already ended, the operational lesson may concern the next buy or allocation rather than a current transfer. An historical estimate should not be turned into an urgent live action without checking the current product, date and available stock basis.

Avoid learning the next forecast from constrained sales alone

Using only thirty recorded sales could repeat the original shortage if the ten-per-day estimate is credible. Using fifty without qualification could overstate demand if the available days were unusually strong. Retain both the observed value and the estimated adjustment with their reasons.

Atlas supports the analysis and Core Plus the stock response, with the configured data and permissions determining the actual workflow. The forecast owner needs the uncertainty and evidence window, not only a single adjusted total that looks more complete than the observations allow.

Do not assess the forecast adjustment on the same data that supplied its assumption and call that independent validation. Compare later, genuinely available trading with the accepted estimate where possible. Retaining the original observed sales and adjustment separately lets the next reviewer identify whether the method improved or simply increased the number.

Use the estimate to improve the next decision

For this scope, the independent estimate is twenty unserved units on two missing days, giving fifty total potential demand rather than thirty recorded sales. The substitution sensitivity lowers the category effect to twelve, but only under its separate assumption. No actual recovery is represented.

Merchmix connects the availability question with the stock decision it can inform. The value is a better supported range, size or allocation assumption, not a bigger reported sales number. An honest estimate shows what is known, what is inferred and which evidence would change the conclusion.

Common questions