Retail analytics: understanding customer behavior in 2026
A practical guide for small and mid-size retailers: which customer behavior metrics matter, how to segment customers with RFM, how to read cohorts, price changes and promotions correctly, and how to pull store, web and email data together.
Retail analytics used to mean counting what sold. Understanding customer behavior means asking who bought, whether they came back, what brought them back and what it cost to get them. A small retailer can answer most of that from data it already has: order exports, payment records, web analytics and email results.
This guide covers the measures worth tracking, how to calculate each with a small example, and the mistakes that lead to wrong conclusions.
The customer metrics that matter
Repeat purchase rate
Repeat purchase rate = customers with 2 or more orders / customers with at least 1 order, over the same period (usually 12 months).
If 4,000 people bought from you last year and 1,000 of them bought more than once, the repeat rate is 25%. It is the clearest single signal of whether customers like what they got.
Purchase frequency and average order value
- Purchase frequency = orders / unique customers in the period
- Average order value (AOV) = revenue / orders
Watch them together. A promotion that raises order count while AOV falls may be moving the same spending into more, smaller baskets.
Customer lifetime value
A simple, honest version for retail:
CLV = AOV × orders per year × gross margin × expected years as a customer
With a $60 AOV, 3 orders a year, a 40% gross margin and 2 expected years: 60 × 3 × 0.4 × 2 = $144. Use gross margin, not revenue: CLV is what you can afford to spend to win and keep a customer. Compare it with what you pay to acquire one.
Lapse rate
Retail has no cancellation button, so define lapse: no purchase in a period that is long for your category, such as 180 days for apparel or 60 days for consumables. Track the share of last year's customers who have lapsed.
RFM segmentation
RFM scores each customer on three behaviors:
- Recency: days since the last order
- Frequency: number of orders in the last 12 months
- Monetary: total spend in the last 12 months
Sort customers on each measure and split them into five equal groups, scoring 5 for the best fifth and 1 for the worst. A customer who ordered 10 days ago, 8 times this year, for $900 in total scores 5-5-5. One who ordered once, 300 days ago, for $40 scores 1-1-1.
Then group the scores into segments you can act on:
| Segment | Typical scores | What to do |
|---|---|---|
| Champions | R 5, F 4 to 5, M 4 to 5 | Early access, referrals, no discounts needed |
| Loyal | R 3 to 4, F 4 to 5 | Recommendations, loyalty rewards |
| New | R 5, F 1 | A good second-purchase offer within weeks |
| At risk | R 1 to 2, F and M 3 to 5 | Personal win-back message before they lapse |
| Hibernating | R 1, F 1, M 1 to 2 | Low-cost reminders, or leave them alone |
Keep the thresholds fixed for a year so you can see customers move between segments. Our guide to customer segmentation with RFM and behavioral methods goes deeper.
Cohorts: comparing newer and older customers
Group customers by the month of their first order and track what share buy again in month 1, 2, 3 and so on. If the March cohort repeats at 18% by month 3 and the September cohort at 11%, something changed: the product mix, the acquisition channel or the offer that brought them in. Totals hide this; cohorts show it. See cohort analysis for a worked layout.
Price and promotion analysis
Price elasticity
Elasticity = % change in quantity sold / % change in price
You raise a product from $20 to $22 (+10%) and weekly units fall from 1,000 to 850 (−15%). Elasticity is −1.5, and revenue drops from $20,000 to $18,700. When elasticity is between 0 and −1, a price rise increases revenue; beyond −1 it reduces it. Remember margin: at a $12 unit cost, gross profit goes from $8,000 to $8,500 even though revenue fell.
Test on a few products at a time and compare against similar products whose price did not change, so seasonality does not get the credit.
Promotion lift
Incremental sales = sales during the promotion − expected sales without it
Estimate the expected level from the same weeks last year or the average of the four weeks before, adjusted for trend. Then check the weeks after: if sales dip below normal, customers bought early rather than more. Also count the discount cost on sales that would have happened anyway.
Forecasting demand
For most small retailers, a forecast of weekly units or revenue per category is enough to plan stock and spend. Good practice:
- use at least a year of weekly history so seasonality shows up, two years if you can
- mark promotions and stockouts, because both distort the history
- judge a forecast by its error on past weeks it did not see, not by how smooth it looks
The omnichannel problem: one customer, three records
The same person may buy in store with a card, online with an email address and respond to an SMS campaign. Unless something ties those together, you count three customers and understate repeat rate. Practical steps: ask for an email at the till for receipts, use the same customer ID across systems where your tools allow it, and report store and online repeat rates separately until you can match them.
Where clariBI fits
clariBI connects the sources a small retailer typically has: payments from Stripe, PayPal or Square (Square includes orders and locations), Google Analytics including its e-commerce report, Google Ads and Meta Ads, email from Klaviyo, Mailchimp or Brevo, support from Gorgias or Zendesk, and order exports as CSV or Excel. Each source gets a dashboard automatically, and you can ask questions across them, such as "What was revenue by week next to ad spend?" or "What share of customers who first bought in March have ordered again?"
Two limits matter for retail. Shopify connects only through the Shopify Storefront app, which reads the product catalog, not orders, customers or revenue, so export orders regularly for sales analysis. And there are no inventory or warehouse system connectors; stock levels come from an export or a sheet.
The e-commerce templates cover repeat purchase and retention, AOV trends, cart abandonment, CAC by channel and inventory turnover with stockout risk. From Starter up, forecasting projects weekly revenue or units with prediction bands and checks your other metrics, such as ad spend, as possible drivers. The e-commerce and retail use case page lists what connects, and the article on inventory, CLV and store metrics covers the operational side.
A first step
Export 12 months of orders with order date, customer ID or email, and order total. Upload it and ask: "What share of customers placed more than one order, by month of first order?" That single chart shows if newer customers come back as often as older ones.
