Skip to content

SIGNAL

What the world tells you

The segments are already in your orders table.

Most brands buy an audience tool before using the data they already have. This groups your customers by how they actually buy, and writes a portrait of each group that the creative work is briefed from.

Buys often Spends Champions recent, often, high Going quiet was all three Bought recently Groups fall out of the corners, not out of a survey.
Groups from your own orders, on two axes. Illustrative, not a screenshot.

What Audience Intelligence is

Audience Intelligence reads your customer data and groups it by how recently, how often and how much each person buys, producing eight segments from your best customers through to lapsed. Each segment carries its own demographics, its behaviour, and a written portrait of who is in it. It also builds programmatic audience lists with estimated sizes and match rates. The point is not the chart. The point is that the portrait becomes the brief: the creative made for people who are about to lapse is different from the creative made for people who buy every month, and the platform knows which is which.

The problem

One message, sent to everybody, at the same time.

Almost every growth-stage brand has the data to segment and does not use it, because the work of turning an orders table into something a copywriter can act on is nobody's job. So the newsletter goes to the whole list, the discount reaches people who would have paid full price, and the customers quietly drifting away get the same email as the ones who bought last week. The waste is invisible, which is why it survives.

What you get

Eight groups, and what each one needs

The groups, from best to gone

Your customers grouped by how recently, how often and how much they buy, from your strongest through to the ones who have lapsed, each with a written portrait of who they are and what stops them.

From your own orders

What each group is worth

Value and spend per group, so the argument about where the budget goes starts from a number rather than a preference.

Value, not headcount

A health flag per group

Which groups are healthy and which are draining, so the one quietly leaking revenue does not need somebody to notice it first.

At a glance

Straight out to the ad platforms

Push a group to your paid and programmatic channels as an audience, with the match rate shown, instead of rebuilding it by hand in each one.

Synced

In practice

Eight groups, and what each one actually needs

18,400 customers, grouped by how recently, how often and how much they buy Your best customers 2,940 First access, not discounts Regulars 4,120 Basket size, not frequency Could become regulars 2,610 The second purchase Just arrived 1,880 Tell them what else you make Going quiet 2,240 Worth the most to win back Cannot afford to lose 890 Call, do not email Dormant 2,470 Cheap reactivation only Gone 1,250 Stop paying to reach them Figures are illustrative. Every published example is aggregated and synthetic.
The chart is not the point. The line under each group is: two of these should never see the same ad.

Each group carries a written portrait of who is in it, and the creative made next is briefed from that portrait rather than from a hunch. That is the difference between a segmentation chart and a segmentation anyone acts on.

Coverage

Where the data comes from

Route What it needs
Your commerce platformOrders with a customer identifier, a date and a value
Your warehouseA table you already model, however you model it
A scheduled fileA counter export, a distributor file, or a spreadsheet
Anything elseIf you hold it, there is a route in. Nothing here assumes you sell online

Three fields are enough to start: who, when, how much. Everything else makes the portraits better rather than making the grouping possible.

Where it connects

What the segments feed

  • Content Studio briefs from the segment portrait, so the ad for lapsing customers is not the ad for regulars.
  • Campaigns targets by segment, and reports back by segment, which is what makes the loop close.
  • Ads Manager pushes audiences to the ad platforms rather than rebuilding them by hand in each one.
  • Analytics joins performance back to the segment it came from, so a number always has a person attached.

Questions

Audience Intelligence, answered

What data do you need to start?

An order history with a customer identifier, a date and a value. That is enough for the segmentation. Everything else makes the portraits better.

We sell offline, or through marketplaces. Does this still work?

Yes, and it is a common case. The segmentation runs on rows, not on a storefront: a billing counter export, a distributor sales file, a WhatsApp order log or a spreadsheet of phone numbers and dates all work, through direct CSV upload, scheduled file drops or BigQuery. What it cannot do is invent customers you never identified, so a brand selling purely through a marketplace that masks contact details will have less to work with than one that captures a phone number at the counter.

Is this the same as RFM?

It is, without the initials. Recency, frequency and monetary value are the three axes. We say it in words because that is how the people using it talk.

Do you store our customers' personal data?

The segmentation works on aggregates. We ask you to bring the minimum needed, and nothing identifiable is ever used in our own examples or demos.

How often does it refresh?

On a schedule you set. Segments move slowly enough that daily is usually unnecessary and weekly is plenty.

See your own customers grouped by how they buy.

Bring an export to the call, or connect a source live. Either way you leave with the segmentation rather than a slide about it.

Book a Demo