SMS Segmentation: Build an Audience That Knows When to Stop

By Rene Siller. Sources checked 2026-10-28.

A practical system for intent, eligibility, exclusions and overlap, with a worked audience and an editable segment launch planner.

## Quick answer

SMS segmentation is the practice of grouping subscribers by the message that is useful to them, while keeping permission, current customer state and exclusions part of the decision. A shopper can be interested in a product and still be the wrong person to text today.

Source context: Recommended SMS segments (https://help.klaviyo.com/hc/en-us/articles/360047879512); Segment conditions reference (https://help.klaviyo.com/hc/en-us/articles/115005062847).

## Start with the message the shopper needs

The common failure starts with an attractive audience: recent clickers, high-value buyers or people due for a refill. The team writes the offer, schedules the campaign and checks the number at the bottom. But that number can hide a person who has already purchased, a shopper waiting for help and someone who received another promotion a few minutes ago.

Start with one message and write the reason a person should receive it. Then write the events that should stop it. This guide turns that decision into an inspectable rule, a worked audience and a launch planner you can use with your own events. Its example windows are choices to test, not universal best practice.

[Download the segment launch planner](downloads/segment-launch-planner.html). Open the completed example first, then make a copy of the plan for your store. [Get the printable decision brief](downloads/segment-launch-planner.pdf).

## Eligibility comes before interest

A segment is an audience definition. Send eligibility is a second question: may this account send this kind of message to this person now? Treat the two as related controls, not interchangeable labels.

Klaviyo’s recommended-segment documentation requires current SMS marketing eligibility alongside the activity condition. That matters when the activity is joined with OR. The intended logic is eligible AND either recently subscribed OR recently clicked. Without the right grouping, a click can accidentally become an alternative to permission.

Table: Keep the permission gate outside the intent group
Rule | Meaning | What it cannot establish
Currently eligible for SMS marketing | The profile meets the platform’s present channel eligibility condition | That any particular offer, timing or use is appropriate
Recent relevant behaviour | The shopper performed the event in the defined window | That the shopper still needs the message
No conflicting current state | No purchase, support or contact-pressure rule takes priority | That an event feed is complete or current

Keep the platform’s permission and suppression records as the authority. A phone number, email subscription or past purchase is not a replacement. Verify the account, destination and programme rules that apply to the actual send. This guide explains an operating method, rather than certifying a programme’s compliance.

Source: Recommended SMS segments (https://help.klaviyo.com/hc/en-us/articles/360047879512).

## Build a rule you can inspect

Write the audience rule in ordinary language before building it: send a product follow-up to currently SMS-eligible people who clicked the relevant product in the last seven days, have not bought since that click, have no unresolved support issue and have not received the promotional contact excluded by this plan. The seven-day window here is a teaching input. Choose yours from the decision and evidence you actually have.

The purchase condition deserves more care than it usually gets. “No order in seven days” and “no order after the relevant click” are different rules. A shopper who bought six days ago and clicked again yesterday is handled differently by each. Decide which question matters, then check whether your platform can express it with the events available.

Klaviyo’s conditions reference distinguishes event counts, event properties and timeframes. Use the actual product identifier on the relevant event when the message concerns that product. A general site visit cannot establish interest in a specific variant. Document what happens when the product identifier, event time or person match is missing.

Do the same for time boundaries. Write whether the start and end are inclusive, which clock the event uses and when the audience is evaluated. A profile on the boundary should be a known-answer test, not an accidental decision left to the campaign builder.

Source: Segment conditions reference (https://help.klaviyo.com/hc/en-us/articles/115005062847).

## The audience count is a sequence of decisions

Consider an audience of 5,000 profiles. In this worked example, all 5,000 begin currently eligible for SMS marketing. A relevant click leaves 800. Removing people who purchased after that click leaves 620. A current product or journey check leaves 500, the recent-contact rule leaves 420, and the support exclusion leaves 350.

Figure description: Worked audience: 5,000 currently eligible profiles, 800 relevant clickers, 620 without a later purchase, 500 matching the current product or journey, 420 outside the recent-contact rule and 350 without an unresolved support issue. These are original teaching counts, not merchant observations.

Worked audience: 5,000 currently eligible profiles, 800 relevant clickers, 620 without a later purchase, 500 matching the current product or journey, 420 outside the recent-contact rule and 350 without an unresolved support issue. These are original teaching counts, not merchant observations.

Table: Sequential filters, each count starts from the previous row
Stage | Profiles remaining | Removed at this step
Current SMS eligibility | 5,000 | Starting population
Relevant click in the chosen window | 800 | 4,200
No purchase after that click | 620 | 180
Current product or journey match | 500 | 120
Outside the chosen recent-contact exclusion | 420 | 80
No unresolved support issue | 350 | 70

The send-ready group is 7% of the starting population in this example. That is a result of these inputs, not a target share. A much larger or smaller audience can be correct. The useful question is whether the people removed at each step are the people the rule intended to remove.

These losses are sequential and disjoint in the table. Do not subtract separately exported exclusion lists as if their members never overlap. If the same shopper purchased and has an open support issue, they must leave the final audience once. Keep the final unique-person count and the first exclusion that applied.

Membership can change between the preview and the send. A purchase arriving late in the event feed is still a purchase. Check the actual platform’s last evaluation point and pause or cancellation behaviour. Record the preview time, send time, event delay and final skip reasons so that a count change can be explained.

## Give overlapping journeys a priority

Figure description: Proposed promotional-message priority: a current suppression stops the send; an unresolved support issue defers the promotion to its owner; a purchase ends the obsolete recovery message; only then evaluate the relevant audience and contact policy. Adapt the priority to the actual programme.

Proposed promotional-message priority: a current suppression stops the send; an unresolved support issue defers the promotion to its owner; a purchase ends the obsolete recovery message; only then evaluate the relevant audience and contact policy. Adapt the priority to the actual programme.

An overlap policy answers which journey owns the next contact when a person qualifies for several. Someone may be a VIP, a recent browser and a cart abandoner at the same time. A more specific need can deserve the next action even if another segment looks more valuable in a campaign report.

Write the order explicitly. In a promotional flow, a current suppression wins. An unresolved service problem can route to a person instead of another offer. A completed purchase can stop the recovery message. Only after these checks should the team decide which eligible campaign or conversation comes next.

Klaviyo documents separate Smart Sending windows for email and SMS. So an SMS frequency check cannot, by itself, tell you that the shopper has not just received an email. Decide whether your programme also needs a shared contact history, and which system and owner can evaluate it. Do not assume an integration supplies this merely because two tools are connected.

A deliberate defer also needs an end condition. Who clears the support flag? When does the message become obsolete? What happens if the next event never arrives? A queue with no owner or expiry can silently turn a temporary exclusion into a permanent one.

Source: Understanding Smart Sending (https://help.klaviyo.com/hc/en-us/articles/115002779311).

## Replenishment needs product-specific timing

For a health or supplement store, a repeat-purchase audience should start with the item and purchase history, rather than an assumed condition or treatment need. Buying a product does not tell you a person’s diagnosis, usage or reason for buying it. Keep those inferences out of the rule.

Suppose the merchant wants to evaluate a refill reminder. The plan needs the product or variant, pack size, relevant purchase date and any current subscription or repeat order that makes the reminder unnecessary. A thirty-day rule across the whole catalogue can be wrong when different pack sizes or customer habits imply different timing.

Treat the first window as a hypothesis. Look at the store’s actual repeat-order intervals by product, the quality of the person match and the share already covered by another programme. If you cannot observe these inputs, use a modest information-led message or defer the automation while the data is established.

The content also needs a boundary. A relevant question about ordering or pack size is different from personalized medical advice. Keep approved product information, permitted claims and the handoff route clear. Segmentation should make the message more relevant without inventing knowledge about the shopper.

## Use the reply without guessing the person

A reply can explain why the click did not turn into an order. “Will this arrive before Friday?” and “Does this come in a smaller pack?” point to different next actions. Preserve the actual question and its context instead of collapsing both into a generic engaged-subscriber label.

Use a stated preference where the shopper has actually provided one, and retain when and how it was recorded. An inferred preference based on a single click should remain an inference. Missing behaviour is especially ambiguous: no click may mean no interest, a missed message, a technical limit or a different way of buying.

If you use reply themes to propose new segments, have a reviewer inspect the categories and a sample of the underlying cases. Keep sensitive material out of marketing profiles unless there is an approved reason and handling process. A useful operating category is the action the team needs to take, such as delivery clarification or human help.

## Test the people who should stay out

Table: Test profiles before the campaign
Profile case | Expected result | Check to retain
Eligible, relevant click, no exclusion | Include | Exact event, product and evaluation time
Relevant click, no current SMS eligibility | Exclude | Permission remains required despite the click
Eligible, purchase after the click | Exclude | Later purchase stops the obsolete follow-up
Eligible, unresolved service case | Defer | Named owner and release condition
Qualifies for two journeys | One chosen action | Priority rule and unique-person identity
Click exactly on the cutoff | Matches the written boundary | Timezone and inclusive or exclusive rule
Missing product or event time | Review or exclude under the plan | No guessed product interest
Permission revoked after preview | Do not send | Final eligibility or cancellation observation

Use controlled test profiles and the platform’s preview tools. Keep expected and observed results side by side. A segment passing the positive case does not prove the negative cases work, and one clean profile cannot establish how late events or duplicate identities behave.

A launch decision needs the rule version, source events, the expected audience, the actual sample checks and a responsible person. The planner keeps these together. It is an operational review, not a universal guarantee that the system will never fail.

## Compare within the same eligible audience

A high-intent segment will often contain people who were already closer to purchasing. Comparing its conversion rate with the whole list does not isolate the effect of the new message. The audience selection itself changed the starting conditions.

For a bounded test, define eligibility first and compare messaging policies within that same eligible population. Keep the person assigned to one policy, use the same outcome window and count purchases by assigned group rather than only the orders attributed to a click. Include variable order costs, discounts and message costs in the decision.

Also record the operating outcomes the segment was meant to improve: obsolete messages prevented, unresolved questions, support transfers and opt-outs. A smaller audience can be the correct outcome if the removed messages were unnecessary. Choose the primary decision before reading the results so a convenient secondary metric cannot replace it.

Use the [segment launch planner](index.html#planner) to keep the audience definition, expected exclusions and actual observations beside the decision. Keep any external benchmark’s population and measurement definition attached to its number.

## Use the completed segment planner

The segment launch planner starts with the 5,000-to-350 example filled in, followed by editable rule and review fields. The example is visibly separate from your store’s inputs. Use it to identify the source event, the condition that ends the message and the person responsible for each exception.

Replace the worked counts with your own snapshot. The calculator rejects rising stage counts and missing or non-integer values. That keeps a malformed input from looking like a healthy audience. It cannot determine whether the platform captured the correct people, so keep the profile checks alongside the count.

Save the completed plan locally as a JSON file and keep it with the campaign record. The tool also supports importing that plan and printing it. Saving a file is explicit; editing the page alone is not a backup. The downloaded tool contains its own text and logic and does not need an account.

[Download the segment launch planner](downloads/segment-launch-planner.html). Open the completed example first, then make a copy of the plan for your store. [Get the printable decision brief](downloads/segment-launch-planner.pdf).

## Where Zobi fits the next conversation

When a segment exists because shoppers have a buying question, the next useful improvement may be the answer rather than a finer audience label. A checkout follow-up that invites a reply needs a way to understand that reply, use the current approved facts and move an exception to the right person.

Zobi’s EQ models are designed to read a shopper’s question in context and choose a relevant reply from merchant-approved information, within the programme parameters the merchant approves.

That gives the merchant a concrete evaluation: bring one checkout or repeat-purchase question family, the product and policy sources, and the audience’s stop conditions. Discuss how Zobi would handle that conversation within those boundaries. Keep the segment test and the reply test connected, so a useful answer does not arrive after the shopper has already bought or asked to stop.

Discuss your audience and reply flow with Zobi (https://zobi.com/dtc/).

## Questions operators need to settle

### How many SMS segments should we start with?

Start with the decisions you can operate and inspect. One audience with a clear purpose, stop condition and owner is more useful than ten overlapping labels without a contact policy. Add a segment when it changes a message, timing, offer or responsible team.

### Can a recent click replace an eligibility check?

No. In the cited Klaviyo setup, current SMS marketing eligibility remains required alongside the intent conditions. A click describes behaviour, rather than permission.

### Should we send to every VIP?

A value label alone cannot establish the current need for a message. Check the relevant product, purchase and support state, the person’s expressed preferences and the actual programme permissions.

### What if the segment count is lower than expected?

Inspect the first filter where the count changes, then sample the people included and excluded. Do not loosen a permission or suppression condition just to reach the planned volume.

## Sources and method

Primary Klaviyo documentation was read on 28 October 2026 for the limited platform behaviours described here. The audience counts, seven-day window, operating priority and test profiles are original teaching inputs and proposed methods. They are not a Zobi merchant study, a provider performance comparison or evidence of conversion lift.

The segment planner checks the arithmetic and order of supplied counts. It does not connect to a merchant account, collect permission or verify source events. Review the actual programme and platform configuration before using the plan operationally.

Source: Recommended SMS segments (https://help.klaviyo.com/hc/en-us/articles/360047879512).

Source: Segment conditions reference (https://help.klaviyo.com/hc/en-us/articles/115005062847).

Source: Understanding Smart Sending (https://help.klaviyo.com/hc/en-us/articles/115002779311).

Source: Understanding SMS consent collection (https://help.klaviyo.com/hc/en-us/articles/360035056972).

## Publisher disclosure

Zobi publishes this guide and offers conversational commerce services. Source documentation, original worked calculations and proposed operating methods are identified in the article.

Zobi · zobi.com