What Your Sales Data Is Trying to Tell You About Your Customers

Your historical sales data contains the answers to the vital questions about your customers: who comes back, what makes them come back, and when they’re likely to buy again.

You don’t need a data team to get value out of this. You just need to know which questions to ask and what to do with the answers.

Who comes back, and why

Repeat purchase rate. What share of your customers buy a second time within a set window. This number tells you whether you’re building repeat business or running a series of one off transactions. If it’s low, your marketing spend to acquire new customers is doing all the work every single month, and that’s an expensive way to run a store.

Which products bring people back. Not every sale is equal. Some products or categories are gateway purchases: buy one, and you’re highly likely to buy again. Others are one-and-done, no matter how good the product is. Once you know which is which, you can stop treating all first-time customers the same. Someone who bought a gateway product deserves a different follow-up than someone who bought a one-off.

Category or genre stickiness. If you sell across multiple categories, some will have devoted repeat buyers and others won’t. This tells you where loyalty programs and subscriptions will actually work, and where they’re a waste of effort.

When they come back

Mean and median time between purchases. These two numbers together tell a story the average alone can’t. If the mean is much higher than the median, you’ve got a small group of infrequent big spenders pulling the number up, while most repeat customers actually buy again quite quickly. That distinction changes your email marketing cadence. A customer who typically buys again within six weeks and hasn’t come back in ten needs a different message than one who typically takes eight months.

A useful rule of thumb: mean > median usually points to a long tail of infrequent buyers. The larger that gap, the more valuable customer segmentation becomes.

If you see the reverse, median higher than mean, check your data before you trust it: a short observation window can cut off customers who simply haven’t repurchased yet, and split orders or returns booked as new purchases can drag the mean down artificially. But if neither of those explains it, look at whether you’re blending genuinely different repurchase cadences, consumables and durables, subscriptions and one-off buys, into a single number. That’s not really a data problem, just a sign the metric needs to be split by product type before it means anything.

Seasonality. When do repeat purchases actually happen? Gift giving periods, start of school year, EOFY, whatever’s relevant to your category. Knowing this lets you time your win-back campaigns to when people are already in a buying mindset, rather than fighting an uphill battle in the wrong month.

Who’s worth the most, and who’s about to leave

RFM segmentation (Recency, Frequency, Monetary value). This is the classic way to sort your customer base into groups based on three things: how recently they bought, how often they buy, and how much they spend. Score customers on each of the three, then combine the scores. Someone who scores high on all three is a genuine champion. Someone who used to score high but has gone quiet on recency is at risk of lapsing. It sounds technical but the output is simple: a short list of segments, each needing a different message. Your best customers don’t need a discount to keep buying, but your lapsing customers might. We show you how to create RFM segmentation later in this post.

New versus returning revenue split, tracked over time. If your growth is coming entirely from new customer acquisition and returning revenue is flat or shrinking, you’ve got a retention problem hiding behind a growth number. This is worth checking quarterly at minimum, because it’s exactly the kind of thing that looks fine until it suddenly isn’t.

Customer lifetime value by acquisition channel. Not all channels bring in equally loyal customers. A channel with a cheap cost per acquisition but low lifetime value might be worse than a more expensive one that brings in customers who stick around. This should directly inform where your next marketing dollar goes.

What they buy together, and why

Basket analysis. What products actually get bought together, in the real data, not in your assumptions about your customers? This tells you what to bundle, what to recommend at checkout, and what to put next to each other in your navigation.

Discount sensitivity. Do your promotions build genuine repeat behaviour, or do they just pull forward purchases people were going to make anyway? A customer who reliably buys during sales isn’t necessarily disloyal, they might just be price-conscious and predictable. The distinction that actually matters is whether the relationship deepens over time, more categories, higher spend, referring others, or whether it stays purely transactional, they show up for the discount and nothing else changes about how they engage with you. The first is real repeat business worth investing in. The second tells you the promotion is doing the work the relationship should be doing. This is worth checking before you plan next year’s promotional calendar.

Where they buy from, and what they buy in

Bulk buyers versus individual retail customers. If you sell to both, lump them together in your analysis and you’ll draw the wrong conclusions from both groups. Separate them first, then run every other question on this list against each group individually.

Format preference, if you sell in more than one, print versus digital, paperback versus hardback. Format often predicts repeat behaviour better than category does.

Channel and geography. Regional purchase patterns and channel level repeat rates round out the picture, useful if you’ve got a physical distribution footprint or you’re deciding where to focus regional marketing.

What to actually do with this

Don’t try to answer every question on this list at once. Start with repeat purchase rate and mean/median time between purchases, because those two numbers alone will tell you whether you have a retention problem worth solving, and roughly when to reach out to a customer who’s gone quiet.

Once you’ve got a handle on those, RFM segmentation is the next highest value step. It turns a wall of data into a short list of customer groups you can actually act on this month.

A common approach is to score each dimension from 1 to 5, where 5 is best.

Example RFM scoring table

Customer	Recency (days since last purchase)	R Score	Orders in last 12 months	F Score	Total Spend ($)	M Score	RFM Score
Alice	7	5	12	5	$2,400	5	555
Ben	21	4	8	4	$1,100	4	444
Cara	45	3	5	3	$650	3	333
Dan	120	2	3	2	$300	2	222
Emma	240	1	10	5	$2,100	5	155

How the scores are assigned

Recency

Days Since Last Purchase	Score
0-14	5
15-30	4
31-60	3
61-120	2
121+	1

Frequency

Purchases in Last Year	Score
10+	5
7-9	4
4-6	3
2-3	2
1	1

Monetary

Total Spend	Score
$2,000+	5
$1,000-$1,999	4
$500-$999	3
$250-$499	2
Under $250	1

Resulting segments

RFM Pattern	Segment	Typical Message
555, 554, 545	Champions	Early access, VIP treatment, referrals
455, 445, 444	Loyal Customers	New products, cross-sell recommendations
355, 345	Potential Loyalists	Encourage the next purchase
155, 255	At Risk	Win-back campaign, special offer
111, 112	Lost Customers	Reactivation campaign or suppress from frequent marketing

Notice that Emma (155) is a great example of an at-risk customer. She spends a lot and buys often when active, but she hasn’t purchased in a long time. That’s usually a much higher priority target than someone with a score like 222, who was never especially valuable to begin with.

Next week’s post walks through how to pull these numbers out of your own sales data, along with the queries to get there. If you’ve read this far and you’re already itching to know your own repeat purchase rate, that one’s for you.