Win-back calling campaign: segments and waves

An old base is not one list

There are thousands of lapsed customers in the CRM, and the idea of calling them always starts the same way: «take the base, call everyone». The outcome is also always the same — after the first hundred calls the team is tired, the result is unclear, and the campaign stops.

The cause is not the script. An old base is not homogeneous: a customer who bought three months ago sits next to someone who ordered once three years ago. Calling both with the same sentence fits neither.

This article builds the mechanics of a win-back campaign: segmenting the base on four dimensions, which segments not to call, wave design, the learning loop between waves, and measurement. The general logic of reactivation — why, and what to say — is in waking a dormant customer base and is not repeated here.

Four dimensions for segmenting

Splitting a base does not need a sophisticated model. Four dimensions are enough, and all of them come from data the CRM already holds.

1. 1Time since the last purchaseThe strongest single dimension. Three months, six months, a year, more than a year — four groups are enough and each carries different odds.
2. 2Purchase frequencySomeone who bought once and someone who bought ten times are different people. The second group is far more likely to return and should therefore be called first.
3. 3ValueAverage order and total spend. This dimension is used for priority rather than for segmenting — between two customers of equal frequency, the higher-value one goes first.
4. 4Reason for stoppingWhere it is known, this is the most valuable dimension: price, a service failure, the need ending, relocation. Where it is not known, establishing it becomes the campaign's first learning objective.

The intersection of the four usually produces five to seven workable segments. More than that is not manageable in practice: each segment needs its own script, its own offer and its own measurement.

Which segments are not called

Cleaning the list is the shortest step in the campaign and the one that pays most.

• Customers who opted out of communication — without exception
• Anyone with an unresolved complaint — the complaint closes first
• Anyone with a payment problem or a dispute
• Customers who already received a campaign call in the last six months
• Wrong numbers and duplicate records
• Companies that ceased trading, and unreliable records

The fourth line needs a rule of its own. Several campaigns calling the same customer is the fastest way to burn a base, and it usually happens because campaigns are built from separate lists. The fix: every campaign checks the same «last contact date» field.

Why not call everyone at once

Dialling a list of five thousand in a week is technically possible and practically wrong, for three reasons.

• The script changes after the first hundred calls — and it should
• The offer sharpens once the first answers come back
• The team has to be able to serve the interested customers — if they all arrive in one day, some are lost
• A base is used once: a call made on a wrong script makes that customer unreachable for a while
• Results have to be compared by segment — running everything together makes that impossible

The wave model solves all five at once: a small wave, a review, a correction, the next wave. It lengthens the campaign and changes the result substantially.

Wave size and interval

This is the wave structure that works in practice, with the numbers scaling to the size of the base.

1. 1The test waveA small part of the most promising segment. The purpose is not sales — it is checking the script and the list. The outcome codes are read for the first time after this wave.
2. 2The first main waveCorrections are made and the highest-probability segment is covered in full. This wave produces most of the campaign's result.
3. 3The second waveThe medium-probability segments. The script is adjusted to the answers from the first wave — usually the offer changes rather than the script.
4. 4The third waveThe low-probability segment. Expectations here are low and the purpose is often not reactivation but cleaning the base: establishing who is unreachable and who is uninterested.

The interval between waves has to be long enough to collect results. In practice that is not less than a week: callbacks, thinking time and second attempts all fall inside it.

Each wave has its own offer

Sending one offer to every segment weakens the campaign. The offer should match the segment's reason for stopping.

• Left on price: an explanation of value or a different package — a discount is the last resort
• The need ended: a new product, a new use case
• Experienced a service failure: resolve first, offer second
• Simply forgot: a reminder is enough, no offer needed
• Bought once and never returned: a small reason for a second purchase

The last two are the most overlooked and the cheapest segments. A meaningful share of customers are not unhappy and have not gone to a competitor — they simply forgot. Offering them a discount is unnecessary and actively harmful: it teaches them to wait for one.

The learning loop

The real value of the wave model is what happens between the waves. After each one, three questions have to be answered.

1. 1Which segment respondedWhere expectation and result diverge, the segmentation is corrected. That correction carries into the next campaign too.
2. 2Which objection recurredWhen the same objection appears in three or four calls, an answer goes into the script. The list of objections is one of the campaign's most valuable outputs.
3. 3Which offer workedCompared by segment. Frequently an unexpected segment performs best and the priority of the next wave changes.
4. 4How many numbers were unreachableThis is a quality indicator for the base and feeds directly into planning the next campaign.

Measurement, by wave

The campaign's headline number says little; measurement belongs at wave and segment level.

• Reachability: what share of numbers were live
• Conversation rate: per segment
• Interest rate: how many conversations ended in a next step
• Return rate: real purchases after thirty and ninety days
• Breakdown of stopping reasons: the campaign's second product
• Base cleaning rate: how many records were marked unusable

The fourth line matters: the result of reactivation does not appear on the day of the call. A customer shows interest, then thinks, and buys thirty days later. Judging the campaign in the calling week makes it look weaker than it is.

Limits

A win-back campaign is not a one-off event, but it should not repeat often either. Calling the same base more than once a quarter lowers response and raises complaints.

How Vexvon supports this campaign

The parts a wave model needs are standard on the Vexvon side.

• The target list is built from a CRM filter — segment criteria become the filter, and the list is not prepared by hand
• A campaign holds up to 5,000 targets, so wave size is also bounded technically
• Preview shows how many will be reached and which rows have no number before the campaign runs — wave planning is built from it
• A running campaign can be paused, resumed and cancelled, so a correction after the test wave is immediate
• The scenario defines the fields to extract: reason for stopping, current supplier, product of interest
• Outcome codes and close reasons stay in the CRM, so the comparison by segment comes out of the reporting

The CRM side of a campaign is on the CRM page. The B2B variant of the same model by sector is in logistics customer reactivation.

First step

Open the base and extract one segment only: customers who have not bought in the last six to twelve months but bought at least twice before that. It is usually the smallest segment and the one with the highest odds.

Build a test wave from a small part of it. To work through the segmentation on your own base, get in touch.