# Win-back campaign: add a phone call to the sequence

Source: https://dialmilo.com/hub/win-back-campaign
Updated: 2026-11-01

A win-back campaign that adds a phone call for your best lapsed customers: when to call, what to ask first, the consent rules, and a worked cost per win-back.

Almost every win-back campaign guide is about email. Email is the right default for most lapsed customers, but for the top slice by value, a phone call does something an email can't: it finds out why they left before it makes an offer. Run the emails first, call the best lapsed customers who didn't respond, ask what changed, then present one written offer that fits the reason. Check consent before you dial, because in the US and UK an AI voice is held to stricter rules than a person.

The standard win-back campaign is a sequence of three to five emails: a reminder, an incentive, a request for feedback, a "last chance" and a goodbye. It works well enough, and it's cheap. Its weakness is the feedback step. Very few lapsed customers fill in a survey, so most win-back offers are built on a guess about why people left.

A phone call fixes the guess. This article is about where to put a call in a win-back sequence, who should get one, what it should say, and what it costs.

## Who gets a call in a win-back campaign, and who doesn't

Calling everyone who lapsed is expensive and, for low-value customers, annoying. Calling no one leaves your most valuable lapsed customers with the same email as someone who bought once two years ago. The answer is to segment by value and recency, the familiar RFM split, and attach a channel to each segment:

| Segment | Example definition | Channel |
| --- | --- | --- |
| Recently lapsed, high value | Top 20% by spend, past 2–3× their normal purchase gap | Email, then a call if no response |
| Recently lapsed, average value | Middle of the spend range, same recency | Email sequence only |
| Long lapsed, high value | Top 20% by spend, inactive 6–12 months | A call first, then email |
| Long lapsed, low value | Bought once, a long time ago | One email, then suppress |
| Left after a complaint | Any value, last contact was a complaint | A call that leads with the complaint, not an offer |

The last row matters more than its size suggests. A customer who left after a bad experience and then receives "We miss you! 20% off" learns that you didn't notice. A call that opens with "you had a problem with your last order and I wanted to check it was sorted" is the only win-back message that addresses the actual reason.

Define "lapsed" from your own data. If customers typically reorder every 30 days, someone at 75 days is lapsing; if they buy once a year, 75 days means nothing. Two to three times the normal gap is a sensible starting rule.

## Where the call goes in the sequence

A sequence that works for the high-value segment:

  
    What's new since they last bought, with no discount. Some people come back on this alone, and they're the cheapest win-backs you'll get.
  
  
    Ask what changed. Present the written offer only if the reason calls for it. Record the outcome.
  
  
    Accepted the offer: send the code by text and email. Asked for a callback: schedule it. Said not interested: stop. Opted out: suppress them from phone everywhere.
  
  
    One message, then stop. Accepting on the phone and acting on it are different things.
  
  
    Leave the door open and stop contacting them.
  

If your emails come from a journey tool, the call can be a step in the same journey. Each Dial Milo campaign has a secret "start a call" webhook that any CRM or journey tool able to send a webhook can call: Customer.io, Braze, Optimove, HubSpot workflows or an in-house system. The journey POSTs the customer's number, name, ID, language and an offer code, and an optional `at` time to schedule it. Duplicates are ignored and opted-out people are rejected automatically. With Customer.io, Dial Milo sends the outcome back as events such as `milo_offer_accepted` or `milo_opted_out`, so the next step branches on what the customer said. The full wiring is in [Customer.io AI phone calls](/hub/customer-io-ai-phone-calls). Without a journey tool, upload the segment as a CSV, or filter your contact book and add the group to a campaign, as described in [managing contacts and lists](/help/managing-contacts-and-lists).

## What the win-back call should sound like

The structure is the same as any good outbound call, with the order changed to put the question before the offer:

1. **Who's calling, including that it's an AI.** Dial Milo discloses this at the start by default.
2. **Confirm the right person** before mentioning their account.
3. **The reason, tied to them.** "You used to order every month and we haven't seen you since July."
4. **Ask.** "Did something change?" Then listen.
5. **Respond to the reason.** Complaint: apologise, note it for your team. Price: present the written offer once. Moved or no longer needs it: thank them and close.
6. **Record the outcome**: accepted, declined, callback requested with a time, opted out, wrong person or no decision.

> **Milo:** Hi, this is Milo, an AI assistant calling for Greenleaf Coffee. Am I speaking with Tom?
>
> **Tom:** Yes, that's me.
>
> **Milo:** Thanks, Tom. You had a monthly delivery with us until July, and I wanted to ask whether something changed.
>
> **Tom:** The last two bags arrived late, and one was stale. I just switched to the supermarket.
>
> **Milo:** I'm sorry, that shouldn't have happened. I'll note it for the team. We've moved to a new courier since then, and for returning subscribers we're offering the first bag free with no commitment. Would you like to try it?
>
> **Tom:** What if it's late again?
>
> **Milo:** You can cancel at any time, and the free bag doesn't tie you in.
>
> **Tom:** Alright, go on.
>
> **Milo:** Great. I've noted that. The code will come to you by text and email shortly.

(Illustrative call; Greenleaf Coffee is a fictional business.) The order did the work. The apology and the courier change addressed the real reason; the offer only made it easy to test the fix. The "accepted" outcome is what triggers the text and email from your own marketing tool.

## The consent question, answered properly

Win-back campaigns feel safe because these are your customers. The rules are more specific than that. As of October 2026:

**United States.** The FTC's [Telemarketing Sales Rule guidance](https://www.ftc.gov/business-guidance/resources/complying-telemarketing-sales-rule) allows an "established business relationship" exemption from the National Do Not Call Registry for 18 months from the last payment, transaction or shipment, and three months after an enquiry. That's useful for live calls by a person. But the same guidance says the relationship "does not permit a seller or telemarketer to place a prerecorded message call", and on 8 February 2024 the [FCC ruled](https://docs.fcc.gov/public/attachments/DOC-400393A1.pdf) that AI-generated voices are "artificial" under the TCPA. So an AI win-back call that makes an offer needs prior express written consent from the customer, not just a past purchase.

**United Kingdom.** The ICO's [telephone marketing guidance](https://ico.org.uk/for-organisations/direct-marketing-and-privacy-and-electronic-communications/guide-to-pecr/electronic-and-telephone-marketing/telephone-marketing/) says consent for automated marketing calls "must specifically cover automated calls". A tick-box for marketing emails doesn't cover it.

This isn't legal advice. It's a summary of published US and UK guidance as of October 2026. The practical step: check whether your checkout or sign-up form asks customers to agree to calls, including automated or AI calls, and only put customers who did into the calling segment. Everyone else stays in the email sequence. The [AI phone call laws](/hub/ai-phone-call-laws) article goes into more depth.

## What a call-assisted win-back costs

Dial Milo is pay as you go with no subscription. An outbound call is the phone line, priced for the exact number dialled and your caller ID and billed per started minute (US numbers from 4¢, a UK mobile from a UK number 8.5¢), plus AI at 9.57¢ per talk minute. That's roughly 14–20¢ a minute all in for US and UK numbers, and calls that don't connect have no minute charge. Price any number on the [rates page](/rates).

**Worked example (illustrative figures).** An online retailer has 2,000 lapsed US customers. 400 are in the high-value segment and consented to calls; 120 of those respond to the first email. The other 280 get a call.

- Assume 35% answer within two attempts: 98 conversations.
- Average talk time 2.5 minutes, billed as 3 line minutes: 98 × 3 × 4¢ = $11.76 line, 98 × 2.5 × 9.57¢ = $23.45 AI.
- Unanswered attempts and voicemails: allow $8. Dial Milo detects voicemail and hangs up without leaving your brand on it, then retries.
- Total: about **$43**.

If 15 of the 98 accept and 10 actually place an order worth $60, that's $600 of revenue for $43 of calls, plus the cost of the offer itself. More useful than the revenue: 98 recorded answers to "why did you stop?". The AI campaign report groups them and labels each customer with a next step, which tells you whether you have a price problem, a delivery problem or a product problem.

Measure against a holdout. Keep a random 10–20% of the high-value segment on emails only. The difference in return rate between the called group and the holdout is your real win-back lift; without it, you'll credit the call with customers who would have come back anyway.

## Testing the call

Two variables are worth testing early, one at a time. First, the reason line: "we haven't seen you since July" against "we've changed a few things since you last ordered". Second, the offer: a free item against a percentage off. Dial Milo can run up to five scripts side by side, each customer keeps the script they first heard, and a winner is only declared at 95% confidence; see [A/B testing scripts](/help/ab-testing-scripts).

Don't test offer size before you've read the reasons. If most lapsed customers left over delivery, a bigger discount is the wrong lever however well it tests.

Two operational details make a bigger difference than either test. Call in the customer's own time zone, inside sensible hours; Dial Milo does this from the phone number or a timezone column, between 10:00 and 20:00 local time by default. And expect callbacks: people ring back a number that just called them, sometimes hours later. Milo recognises a returning customer, thanks them for calling back, confirms who they are and picks the win-back conversation up where it stopped, so a missed call isn't a lost one.

## Where this fits

A win-back campaign is the last line of retention. The cheaper save happens earlier, when the warning signs show and the customer hasn't left yet; that's covered in [churn prevention calls](/hub/churn-prevention-calls). The offer mechanics, consent and pricing for any outbound pitch are in [AI sales calls](/hub/ai-sales-calls). And if your lapsed customers are players on a gaming platform, with safer-gambling checks and self-exclusion rules on top, start with [iGaming player reactivation](/hub/igaming-player-reactivation).

## Frequently asked

**What is a win-back campaign?**

A win-back campaign is a planned sequence of messages to customers who have stopped buying, cancelled or gone inactive, aimed at bringing them back. Most are email-only. Adding a phone call for the most valuable lapsed customers turns the sequence into a conversation: you find out why they left, and you can make an offer that fits the reason.

**When should you start a win-back campaign?**

Define lapsed against your own buying cycle rather than a fixed number of months. A useful rule of thumb is two to three times your typical gap between purchases: if customers normally reorder every 30 days, someone who hasn't ordered in 75 is lapsing. Start the sequence then, while they still remember you, rather than a year later.

**Should a win-back call include an offer?**

Have one ready, but ask why they stopped before you make it. If the reason was a bad experience, an apology and a fix matter more than a discount. If the reason was price, present the written offer once, with its exact terms. A win-back offer that ignores the reason tends to bring customers back for one order and lose them again.

**Is it legal to call lapsed customers with an AI?**

It depends on where they are and what they agreed to. In the US, an established business relationship lets a seller make live calls to people on the Do Not Call Registry for 18 months after their last purchase, but the FTC says it does not permit prerecorded message calls, and the FCC treats AI-generated voices as artificial voices under the TCPA, so AI marketing calls need prior express written consent. In the UK, ICO guidance says automated marketing calls need consent that specifically covers automated calls. This isn't legal advice.

**How many lapsed customers come back from a win-back call?**

It varies too much by business to quote a reliable benchmark, and any single figure you see online is someone else's list. Run the call on a few hundred of your best lapsed customers, compare them with a held-out group who only got the emails, and use the difference as your number.