Retail data analysis: 7 practical use cases

Brazilian retail lost R$ 42.1 billion (about US$ 8 billion) in 2025, the equivalent of 1.65% of everything it sold, according to research by Abrappe and Protiviti. For every R$ 61 in sales, R$ 1 vanished into shrinkage, theft, inventory errors, and expired products on the shelf. The uncomfortable part: much of that loss was already visible in the operation's own data. Someone just had to look.

That is what separates retailers that grow from retailers that merely survive. It is not access to data: every grocery store, pharmacy, or clothing shop generates receipt-level sales, SKU turnover, and campaign history. It is the habit of turning those numbers into buying, pricing, and media decisions every single week.

Retail data analysis is the systematic use of sales, inventory, customer, and campaign information to decide what to buy, what to charge, where to advertise, and when to run promotions. In practice, it answers operational questions: which product is about to run out, which store sells more per category, which local campaign pays off, and which customer is about to churn.

Why is retail the sector with the most to gain (and lose) from data?

Because margins are thin and the volume of decisions is enormous. A manufacturer sets prices a few times a year; a supermarket decides price, placement, and replenishment for thousands of items every week. Each wrong call costs little on its own, but the total is what Abrappe measures in billions.

The gap between theory and practice is wide. A Microsoft study cited by Varejo S.A. (CNDL) found that 74% of SMB managers consult data frequently, yet only 33% manage to turn it into strategic decisions. The bottleneck is not collecting, it is analyzing and acting. That is where the seven use cases below come in.

1. Inventory: predicting stockouts before the shelf goes empty

A stockout is a sale you already paid to make and that never happened. Brazilian brick-and-mortar stores run stockout rates of 10% to 12%, according to Neogrid data compiled by AGP Pesquisas; in e-commerce the rate reaches 36%. Worse: product unavailability accounts for 42% of lost sales, according to ABRAS, the Brazilian supermarket association.

The analysis is simple to describe and tedious to do by hand: cross average turnover per SKU with current stock and the supplier's lead time. If an item sells 8 units a day, you have 20 in stock, and the supplier takes 5 days to deliver, the stockout is already locked in: you will be short about 20 units before the next delivery. Whoever sees this on Monday places an extra order; whoever doesn't finds out on Saturday, with the customer standing in the store.

2. Seasonality: buying on your customer's calendar, not your supplier's

Every retailer knows Christmas and Mother's Day exist. Fewer know, with actual numbers, what seasonality does in their own store: which category doubles in winter, which week of the month concentrates sales, what an early cold spell changes in turnover. That curve comes out of your own sales history: two years of weekly data per category is usually enough to reveal the pattern.

A common example: a regional pharmacy chain notices, looking at three years of history, that vitamin sales rise 40% during the winter months. So it moves its supplier negotiation up to early autumn, when purchase prices are still low. The extra margin didn't come from charging more, it came from buying better, with a head start only data provides.

3. Pricing and margin: knowing where you can (and can't) move

Retail prices are usually set with a standard markup over cost. It works as a general rule and fails at the extremes: destination items (the ones that bring customers into the store) tolerate thinner margins because they pull the rest of the basket, while convenience items accept higher margins without hurting volume. Without analysis, everything gets the same treatment and margin is left on the table.

The data that solves this is observed elasticity: what happened to volume the last few times the price changed. You don't need a sophisticated model at first. Comparing weekly sales before and after each price change, item by item, already shows where customers are sensitive and where they are not.

4. Local campaigns: media budgets per store, not per chain

Anyone running more than one location knows the problem: the campaign that packs the downtown store does nothing for the neighborhood one. Treating media as a single chain-wide expense hides this. Per-store analysis crosses local ad spend (radius-targeted boosts, regionally segmented Google Ads, printed flyers) with that store's incremental sales over the period.

A typical mini-case: a pet shop with three locations spends R$ 1,500 a month per store on local ads. Crossing campaigns with per-store sales, it finds store A generates R$ 9 of incremental sales per R$ 1 invested, store B generates R$ 4, and store C just R$ 1.20. Reallocating C's budget to A makes the same money work harder, without spending one extra cent. That kind of reading depends on having media and sales data in the same place, which rarely happens when each platform lives in its own silo.

5. Product mix and ABC analysis: what deserves shelf space?

ABC analysis ranks products by contribution to results: A items (typically 20% of SKUs) concentrate around 80% of sales or margin. The value lies less in the classification itself and more in the decisions it unlocks: A items must never run out; a C item sitting still for 90 days goes on promotion or leaves the mix; shelf space migrates from C to A.

The classic mistake is running the curve on revenue and ignoring margin. A SKU can be a revenue champion and a profit mediocrity. Running both curves side by side (revenue and margin) changes the conversation with suppliers and the decision of what gets the endcap.

6. Customers and repurchase: the CRM retail forgets to read

Retailers with customer records (loyalty programs, tax-ID on receipts, e-commerce) sit on an underused mine: repurchase frequency. If a customer who used to buy every 30 days has gone 75 without showing up, they have probably moved to a competitor. A well-timed message with the right offer costs pennies compared to acquiring a new customer.

The rule is objective: compute each customer's average purchase interval and trigger an action when the silence exceeds twice that interval. The same data feeds the LTV calculation, which tells you how much each recovered customer is worth and, therefore, how much you can spend to win them back.

7. Demand forecasting: the step beyond history

The six previous use cases look backward. Demand forecasting looks forward: it combines history, seasonality, trend, and events (a local holiday, a planned promotion, the weather) to estimate the next weeks' sales per category. It is what turns reactive buying into planned buying. Today it is also the use case where AI delivers most in retail, as we detail in sales forecasting with AI.

Honesty is due here: forecasts miss. They don't eliminate the buyer's judgment; they narrow the range of the guess. Missing demand by 10% with a forecast is very different from missing by 40% on gut feeling; in inventory, that difference is either cash sitting idle or an empty shelf.

How these use cases compare in practice

Use case	Core data	Decision it changes	Effort to start
Stockout prevention	Turnover per SKU + coverage	When and how much to reorder	Low
Seasonality	Weekly sales per category (2+ years)	When to buy and negotiate	Low
Pricing	Volume before/after price changes	Margin per item	Medium
Local campaigns	Media + sales per store	Budget per location	Medium
ABC analysis	Revenue and margin per SKU	Mix and shelf space	Low
Repurchase	Purchase interval per customer	Retention actions	Medium
Demand forecasting	History + events + weather	Planned buying	High

Where to start without disrupting the operation?

Start with the money already leaking: stockout prevention and ABC analysis use data any POS or ERP exports today, and they pay back the effort within weeks. Spreadsheets work at first, until they don't, which happens sooner than expected once SKUs multiply (we wrote about the limits of spreadsheets). The next step is consolidating sales, inventory, and media into a single view, with the indicators a well-built sales dashboard provides.

Frequently asked questions

What is retail data analysis?

It is the use of information the operation itself generates (receipt-level sales, inventory turnover, customer records, campaign results) to guide decisions on buying, pricing, product mix, and marketing. The goal is not producing reports: it is reducing losses, preventing stockouts, and directing investment to where returns are proven.

Which data should a retailer analyze first?

Sales and inventory per SKU. Those two crossings support the highest-return use cases: stock coverage (prevents stockouts) and ABC analysis (organizes mix and shelf space). Customer and media data come next, once the operational base is under control.

Do I need a BI system to analyze my store's data?

Not at first. An ERP or POS that exports reports, plus a well-structured spreadsheet, covers the first months. The sign it is time to evolve is operational: when consolidating the data takes more hours than analyzing it, the tool has become the bottleneck.

How does data analysis reduce stockouts?

By computing coverage (current stock divided by daily turnover) for each item and comparing it with the supplier's lead time. When coverage drops below lead time, the system flags the item before it runs out. It is a simple calculation that, done per SKU and per store, directly attacks the 10-12% average stockout rate of physical retail.

Does data analysis work for a small, single-store retailer?

It does, and proportionally it pays off even more, because every unit of loss weighs more on cash flow. A single store has less data but also less complexity: ABC analysis, stock coverage, and seasonality fit into a one-hour weekly routine, as long as the reports come ready from the system.

Retailers that decide with data buy better

The difference between the 33% who turn data into decisions and everyone else is not budget or a team of analysts, it is routine. The seven use cases in this article share one trait: they use data your operation already generates and convert it into decisions that repeat every week. The retailer who installs that routine buys better, loses less, and advertises where returns show up. Whatever stays outside that routine remains margin given away for free.

If the obstacle is the work of joining POS exports, spreadsheets, Meta Ads, and Google Ads before the analysis even starts, Sherlok shortens the path: you connect your sources once and simply ask in plain language ("which items will run out of stock this week?", "which store had the best campaign return?") and get the analysis ready, with no spreadsheets and no technical team required.