Customer segmentation is rarely a product of its own. For most companies it is a feature of the email platform, a job for the customer data platform, or a SQL query an analyst runs in the warehouse — and the buyers describe the same problem from all three directions: marketers often need a data team's help for more complex segments, and by the time a segment reaches the channel it may already be out of date.
What is customer segmentation software?
Customer segmentation software groups customers by who they are and what they do — how recently and how much they buy, what they browse, how likely they are to churn — so each group can be messaged or targeted differently.
In practice it is a capability inside another tool: the email or marketing platform, a customer data platform, or the data warehouse. As one buyer without a CDP puts it: “we'll end up doing audience creation within our email service provider tool.”
What the interviews add is how segmentation actually works inside companies:
Among companies that name a tool they segment with, about six in ten name their email or marketing platform and about a third a CDP. Companies segmenting in a CDP are mostly large; email-platform segmentation spans every size.
Across these companies, more complex segmentation often depends on analysts or data teams, even where marketers can build simpler groups in their marketing platform; segments often go stale or fail on the way to channels, and few companies measure whether a segment worked.
What is the difference between a segment, an audience and a CDP?
A segment is a group of customers identified through shared characteristics, behaviors or predictive models; an audience is a group selected for targeting in a marketing channel, sometimes created from an existing segment and sometimes defined within the channel itself; a CDP is one system that can build, manage and distribute those groups. Buyers use the words loosely, and the handoffs between them are where segmentation breaks.
Owned messaging is where most segments go — email, SMS and push — and paid media is second, as audiences synced to Meta, Google or an onboarding service such as LiveRamp. Among buyers in market for segmentation tools, paid media is the job that comes up most often.
Whether a company needs a CDP to do this is its own question; see does your ESP replace a CDP.
Segment
A group of customers defined by shared characteristics, behaviors or predictions — lapsed buyers, high-value customers, likely churners — built in the marketing platform, CDP or data environment.
Audience
A group of people targeted through a marketing channel, such as an email platform or an ad network. Audiences may come from existing segments or be built within the channel; for uploaded audiences, identity matching and sync timing decide how much arrives.
Customer data platform
A system that unifies customer data so segments can be built once and synced to many channels — one option alongside the email platform and the warehouse.
Who builds customer segments?
For anything beyond simple segments, analysts and data teams more often than the marketers who use them. Among companies that say who builds their segments, those that depend on analysts, data science, IT or an agency outnumber those where marketers build their own by more than three to one. One buyer describes a segment request as “a three team exercise that takes several days, weeks to get that information”; another wanted a tool to avoid “dumping it on an analyst who then has to put it in his backlog and five weeks later tells you the answer.”
Self-serve is possible — “it has taken some of the reliance off of our data engineering team,” one buyer says of a warehouse-based tool — and the claims cluster on tools such as Klaviyo, Hightouch and Segment. Buyers in market lean toward the warehouse: about half of them are building segments in it or want to, against fewer than one in ten of the other companies in these interviews. One describes the current state plainly: “we are using Snowflake as a data warehouse to manually define the segmentation and attribution.”
What customer segments do companies build?
Five main types, led by ad audiences and predictive segments. Among companies that describe their segments, these appear in this order:
- Lookalike and ad-platform audiences
- Predictive segments: propensity, churn and lifetime value
- Value tiers built from recency, frequency and spend
- Behavioral segments: browsed, abandoned, bought
- Lifecycle stages: new, active, lapsed
Demographic and persona segments, loyalty tiers and suppression lists are smaller groups.
Lookalike and ad-platform audiences
Customer lists and value groups sent to ad platforms to find similar people or exclude existing buyers.
One buyer: “we can easily build lookalike audiences based on these groups and their value buckets.”
Predictive segments
Propensity, churn and lifetime-value scores, from the email platform's built-in models or a data-science team's own.
One company keeps the models in its warehouse: “We build the predictive models within Snowflake, and those tools are connected to Tealium.”
Value tiers
Customers ranked by recency, frequency and spend into tiers that set offers and treatment — a job some email platforms now make easy: “the CDP for RFM, so like recency frequency stuff, is really, really powerful, and it's very, very simple to use.”
Behavioral segments
What customers browsed, abandoned or bought, often combined with a prediction: one company runs “a product introduction sequence for customers that view a product they've never bought before but we think they have a high propensity score.”
Lifecycle stages
New, active, lapsed and at-risk customers, each with its own flow. One buyer's version: “Who are the active customers? Not so active? Dead? Ghost?”
Segments break most often in transit. Companies describe syncs that run once a day, integrations that time out, match rates nobody can confirm and lookalike seeds that are badly out of date — “lookalike audiences based on consumer data that's, I want to say, like a year old at this point,” as one buyer admits. AI is the newest route in: a small number build segments by describing them — “I could just type in plain English what I wanted and get reasonably close to the segmentation I wanted” — and several times as many say they want to.
Which customer segmentation tools do buyers rate highest?
Hightouch rates 7.9, Klaviyo 7.7, Braze and LiveRamp 7.5, Amperity 7.3, Segment 7.0, and Salesforce Data Cloud and Salesforce Marketing Cloud 6.8. These are overall product ratings for the tools companies segment with, not scores for segmentation alone.
| Tool | Rating | What it means here |
|---|---|---|
| Hightouch | 7.9 | Builds and syncs segments from the data warehouse; some buyers credit it with less reliance on data engineering. |
| Klaviyo | 7.7 | The email platform most companies here segment in, and the one most named for building segments in plain English. |
| Braze | 7.5 | A messaging platform where many companies build segments for owned channels. |
| LiveRamp | 7.5 | Onboards first-party segments to ad platforms, used mostly by large companies. |
| Amperity | 7.3 | A CDP some companies treat as the source of truth for their segments. |
| Segment | 7.0 | A CDP some companies use mainly to build and sync segments. |
| Salesforce Data Cloud | 6.8 | Salesforce's CDP, used for audience building beside Marketing Cloud. |
| Salesforce Marketing Cloud | 6.8 | Segments built in its audience tools; several buyers say they need a specialist to build them. |
Email and marketing platforms
Klaviyo · Salesforce Marketing Cloud · Braze · Bluecore · Iterable · Attentive
CDPs, warehouse and onboarding
Segment · Salesforce Data Cloud · Amperity · Tealium · Hightouch · Census · Snowflake · LiveRamp
How do companies measure whether segmentation works?
Mostly they do not. Only a small number of companies report observed results, and even fewer show whether segmentation produces incremental revenue.
The results buyers report are limited and measure different things. One direct-mail program tracks purchases through unique codes assigned to each segment — “I do unique coupon codes on those postcards, so I can see exact revenue” — which attributes revenue to the segment without proving it would not have come anyway. Another company says of its own audiences, “we don't see the improvement in performance.” Several companies describe holdout or control groups, the test that shows incremental lift, but almost none report what those tests found. The cost complaints are about features held back for higher tiers and data costs that grow with volume; the quality complaints are about counts that do not reconcile — “you question the output of information, like how real is it?”
Due for a stack physical?
Book mine →Should you buy customer segmentation software?
Start from where your segments need to go and who needs to build them, not from a new tool. If your segments mostly feed email and SMS, the email platform may already do the job; if they need to reach ad platforms, the website and the warehouse at once, a CDP or a warehouse-based tool is the more likely fit.
Three things follow from the interviews; the recommendations are ours, not practices buyers describe. First, test whether the marketers who need segments can build them in the trial, without an analyst in the room. Second, check sync timing and match rates for each channel, since that is where segments go stale. Third, attach a holdout to the first segments you launch, so you know whether segmentation is paying for itself.
For the platforms these segments run on, see what is a CDP, does your ESP replace a CDP and what is zero-party data.
Common questions
What is customer segmentation software?
Customer segmentation software groups customers by who they are and what they do, such as how recently and how much they buy, what they browse or how likely they are to churn, so each group can be messaged or targeted differently. In practice it is usually a capability inside another tool: the email platform, a customer data platform or the data warehouse.
Where do companies build customer segments?
Mostly in the email or marketing platform. Among companies that name a tool they segment with, about six in ten name their email or marketing platform, such as Klaviyo, Salesforce Marketing Cloud or Braze, and about a third a customer data platform; a smaller group builds segments in the data warehouse. Companies segmenting in a CDP are mostly large, while email-platform segmentation spans every size.
What is the difference between a segment and an audience?
A segment is a group of customers identified by shared characteristics, behaviors or predictive models, such as lapsed buyers or high-value customers. An audience is a group selected for targeting in a marketing channel, sometimes created from an existing segment and sometimes built within the channel itself, such as an ad platform. The difference matters because segments uploaded to other channels, often through an onboarding service such as LiveRamp, can break or go stale on the way.
Can marketers build segments without a data team?
Simple ones, usually; more complex ones, often not. Among companies that say who builds their segments, those that depend on analysts, data teams, IT or an agency outnumber those where marketers build their own by more than three to one. Self-serve claims cluster on tools such as Klaviyo, Hightouch and Segment, and several buyers in market name that dependence as their reason for shopping.
What types of customer segments do companies use?
The most common in these interviews are lookalike and ad-platform audiences and predictive segments such as propensity, churn and lifetime value, followed by value tiers built from recency, frequency and monetary value, behavioral segments such as browsed or abandoned, and lifecycle stages such as new, active and lapsed. Demographic and persona segments are less common.
Can AI build customer segments?
Some companies already use it. A small number build segments by typing what they want in plain English, most of them in Klaviyo, and several times as many say they want to, including buyers evaluating new tools for that reason. Some are skeptical about trusting segments when they cannot see why the AI chose them.
How do you measure whether customer segmentation works?
Few companies in these interviews do. Only a small number report an observed result, positive or negative, and even fewer show whether segmentation produces incremental revenue. Several describe holdout or control groups, but almost none report what the tests found. Unique codes per segment show which revenue a segment is associated with; only a holdout or control group shows whether the segmentation added anything.
A physical for your tech stack
Peer-powered: built from hundreds of verified buyer interviews, and fitted to the tools you already run. You get a shortlist for your stack and your next steps.
Book my stack physicalFree · no password · your answers are anonymized before they inform research like this.