Customer Segmentation When the Customer Is a Storefront

Customer segmentation, when you are the one selling in, means dividing a market of companies you have never spoken to into groups worth approaching differently. It is not the retention exercise the term usually describes. You have no purchase history for these businesses, no support tickets and no account manager — only what is visible from outside.

That distinction matters because nearly everything written on the subject assumes the opposite. Most of it is addressed to a store segmenting its own shoppers. This is addressed to the people who sell to those stores: performance and Meta agencies, retention and Klaviyo specialists, shop developers, creative and video studios, and software companies selling into ecommerce.

The disclosure belongs here rather than at the bottom. We build Keaz Signals, which sells ecommerce leads with dated buying signals attached. We are a competitor to every list and segmentation product in this space, and we benefit directly if you finish this page believing that timing belongs in a segment definition. There is a section below on where a general segmentation tool is the better purchase. It is not a courtesy — it is where we lose.

They do not transfer because every one of them reads a history you do not have. RFM ranks the recency, frequency and monetary value of past orders. Cohort analysis groups customers by when they first bought. Lifetime-value tiers extrapolate from spend. All three require the company to have already been your customer.

Persona and firmographic segmentation get closer, because they describe a business rather than a purchase record. Headcount, revenue band, industry, technology in use — all of these can be established for a company that has never met you. They are also the most heavily worked axes in B2B, which produces the second problem.

An attribute is shared. Everyone filtering for the same conditions gets the same rows, and the file is worn out the day it is exported. A segment built only on attributes can be entirely correct and still hand you a list that three competitors wrote to in the same week, with the same opener. The list problem itself is worked through in outbound marketing when the list is the strategy.

There is an honest historical reason the discipline grew up this way, and it is not a failure of imagination. For most of its existence, segmentation ran over a database that changed only when somebody bought something. Condition was the only thing there was to segment on.

Five, for an ecommerce store, and they are the ones the public segment preview exposes: platform, country, vertical, audience size and current advertising activity. Each is observable without the store telling you anything, and each maps to a real difference in what you would sell them.

Platform comes first, because Shopify or WooCommerce decides whether an app, a theme or an integration is installable at all. Country is next: a retention offer priced in euros and a US-only payment integration are different products. Vertical — nineteen of them in the public preview, from fashion and beauty to supplements and tools — decides the wording more than it decides the offer.

The remaining two are qualifiers rather than definitions. Audience size arrives as follower bands rather than a revenue estimate you would have to invent, a proxy for budget that does not pretend to be an accounting figure. Advertising activity — whether Meta ads run now, and whether the active count is climbing — already shades into the next section, because it is halfway to being an event.

For scale, as of September 2026: 4.1M stores under watch, re-scraped weekly — 1.7M in Europe, 2.4M in North America, 2.5M Shopify against 1.6M WooCommerce. The live builder carries dozens more filters than the preview. The pool is described on the ecommerce leads database page.

Recency of change — not recency of purchase, which is the R in RFM, but how recently the business itself did something. It is the one axis that cannot be shared, because it expires. A store that switched its ads on last Tuesday is in a different state from the same store a month later, and its attributes are identical in both.

Retrieved 14 September 2026, the live catalogue reads: 370,000 stores with new Meta ads running, 128,000 with a product launch, 96,000 with storefront or site changes, 80,600 with a changed newsletter rhythm, 34,800 with a jump in social growth, 31,000 with a rising active ad count. Each event carries the date it was last seen, which is what lets you tell a queue from a population.

Segmenting on one of those returns a far smaller list than any attribute filter, and that is the point. An attribute segment is a population and it sits still. An event segment is a queue, and it empties. Across the whole pool, about 41,200 stores carry a fresh signal in any seven-day window as of September 2026 — roughly one percent of what is under watch.

Which categories of data can see change at all, rather than condition, is worked through in what each category of intent data can actually see.

Narrow the axes, then read the count before planning anything around it. The count updates as each condition is added, so you learn the size of the addressable market before committing budget, headcount or a quarter of pipeline assumptions to it.

The order that works is constraints first, qualifiers second. Platform and country decide whether your offer can be delivered at all, so they go in before anything else. Vertical follows, because it governs the wording. Follower band and ad activity come last — they trim a segment rather than define one. Only then add a single signal and watch what the number does.

That drop is the useful part, and it is where a segment stops being a spreadsheet exercise. What remains is the portion of your market you could defend writing to today. In our own campaigns, as of August 2026, sending in the week an event happened replied at about 8.5 percent against about 3.9 percent otherwise, with the message unchanged. That is our figure, for our outreach, in our market. It is not a forecast for yours, and we are not making a reply-rate promise — we would have no way to keep one.

Sizing a market you can count rather than estimate is the subject of customer acquisition strategy when your market is countable.

Three, and the omissions are deliberate. There is no conversion rate broken out by vertical or by segment in this post, because we have not measured one in a way that would survive being quoted. A figure averaged across nineteen verticals and two continents tells you almost nothing about the slice you care about, and the number worth having is the one for your own segment.

There is no comparison of our coverage or freshness against any named vendor's. We have not measured their data. A percentage about somebody else's database that we did not take is not a fact we are entitled to publish, however easily it would slot into a paragraph. If we ever run that measurement, the post will say who ran it and when.

And there is no cost-per-record table set against other products, because what each product meters is not the same unit. A table pretending otherwise would be dishonest arithmetic dressed as a buyer's aid.

On data protection the position stays narrow on purpose. We are GDPR-conscious by design and will not claim more than that; responsibility for what you send stays with you. A segment you assemble yourself from public attributes and a segment that arrives with contacts already attached carry the same obligation in different packaging. That is a structural observation about where the duty sits, not a claim about anyone's compliance, and it is not legal advice.

Often, and in more cases than we would like. What we sell covers one kind of business, on two platforms, in two regions. Outside that, a general segmentation or analytics tool is the correct answer and we are the wrong supplier.

If you are segmenting customers you already have, RFM, cohorts and lifetime-value tiers are the right instruments, and nothing above is an argument against them. If your buyers are not ecommerce stores, the pool is Shopify and WooCommerce and nothing else. If your market sits outside the EU and North America, we do not cover it, and we would rather say that than describe a thin pool as a pool.

If you need the records outside the system, we do not export — leads move into campaigns and stay there, and for some teams that alone settles it. If LinkedIn is your channel, we have no LinkedIn signals and no LinkedIn sending, and no plans to add either.

And if the axis you want to segment on is firmographic, the honest answer is that it is not here yet. Registry and firmographic triggers, hiring signals, app installs and uninstalls, funding and ownership changes, cross-border expansion and review momentum are all marked planned on the signal catalogue as of 14 September 2026. Planned means they do not exist today, and a segment cannot be built on one.

Build one segment and read the count. Pick the platform and market you can actually serve, add the signal that matches your offer, and look at what survives. It takes minutes, and it replaces a quarter of arguing about market size with a number you can check.

The full set of events is in the buying signal catalogue, and once a segment exists, the outreach written per store works from the signal and your own knowledge base rather than from a template with a merge field.

Signing up gives 1,000 credits to spend inside segments you build yourself, as of September 2026, and what each tier includes is on the pricing page. Access is capped per market, so the honest first step is the waitlist rather than a purchase.

Segment-builder axes (platform, country, vertical, Instagram follower band, active Meta ads), the nineteen verticals, the pool figures — 4.1M stores, 1.7M Europe, 2.4M North America, 2.5M Shopify, 1.6M WooCommerce, 41,200 with a fresh signal in the last seven days — the 1,000 signup credits and the no-export position. Page last updated 22 August 2026, retrieved 14 September 2026: /ecommerce-leads-database. These figures move, so the date is part of the number.

Per-signal counts and which events are live versus planned, retrieved 14 September 2026: the buying signal catalogue. The reply-rate observation — about 3.9 against about 8.5 percent, our own campaigns, as of August 2026, message unchanged — is stated on both pages above.

Measurement disclosure: no conversion rate by segment or vertical is quoted, no competitor coverage or freshness figure appears, no cost-per-record comparison appears, and no guarantee is made about inbox placement or reply volume. We did not take those measurements.