Churnkey vs Underheard: an offer or a conversation?

Underheard vs Churnkey at a glance

How we compared: public documentation, pricing pages, press coverage, and reviews, all checked on July 20, 2026. Performance figures from either vendor are marked as vendor reported, and we link sources throughout.

What is Underheard?

Underheard helps businesses better understand their customers by turning thousands of conversations into actionable insights, through intelligent voice led interviews at scale. The conversations happen inside your product at the moments that decide retention. A customer who clicks cancel completes the legally free cancellation path first, then sees a voluntary invitation to a short conversation, 2 to 4 minutes, voice or chat.

The agent listens for the real reason and solves what is honestly solvable while the customer is still in the room: a pause, a plan switch, a missing feature routed to the right person, a real complaint escalated to a human. Everything else lands in a recurring root cause report with quantified reasons, real quotes, and recommendations. The cancel flow outcome is a measured save rate uplift, A/B tested against the flow you run today, and the same conversations serve the rest of the lifecycle, onboarding, power user wish lists, accounts going quiet, plus focused one off studies when a specific question needs an answer.

What is Churnkey?

Churnkey calls itself growth infrastructure for subscription companies, and it is the incumbent of the offer engine category. Founded in 2020 and backed by a $1.5M growth round from CreativeCo in 2024, it serves hundreds of customers including VEED, Jasper, and Superhuman, and reports having recovered close to $300 million in revenue across its customer base.

The product is broad and mature: segmented in app cancel flows with pauses, discounts, and plan switches, exit surveys with follow up questions, A/B testing, session recordings, churn metrics, and since October 2025 an AI line with Adaptive Offers that picks per customer discounts, automatic translation into 50 languages, and Feedback AI that turns typed cancellation feedback into revenue weighted themes. On top sits a serious failed payment recovery stack. Integration takes about 35 minutes against Stripe, Paddle, Braintree, or Chargebee, and the entry tier is a published $250 per month with a 14 day free trial.

The core difference: the same doorway, two philosophies

Underheard and Churnkey meet the same customer at the same moment, which makes this the closest comparison in our series. The difference is what each believes that moment is for. Underheard treats it as the one chance to hear the truth: open a conversation, listen for the reason, fix what can be fixed honestly, and let the rest inform the roadmap.

Churnkey treats it as a conversion problem: segment the customer, show the right survey, and compute the offer most likely to keep the subscription alive. Their own Adaptive Offers material describes discounts tuned between 5 and 95 percent, with vendor reported results of 34 percent higher save rates than static offers. It is a disciplined, well engineered version of its philosophy.

Put simply: Churnkey optimises what people pick. Underheard captures why they leave, and acts on it. Which one your cancel flow needs depends on whether your bigger problem is this month's save rate or the reasons that keep refilling the cancellation queue.

What does the cancelling customer actually experience?

With Underheard, the customer cancels freely first, then chooses whether to talk. The conversation runs 2 to 4 minutes, sounds like a person who genuinely wants to understand, and ends either with a solved problem, a fair offer taken voluntarily, or a thank you. Nobody is gated, nobody is argued with, and declining costs one click.

With Churnkey, the customer moves through a configured flow: an exit survey, sometimes follow up questions, then a pause or discount picked for their segment. Done well, it is quick and can feel considerate, and an accepted pause is a real save. Done lazily, it is the discount popup everyone has learned to click through. One G2 reviewer notes the offers cannot always adapt to the use case, which their newer AI tier is built to address.

The experiences differ most for the customer who leaves anyway. In an offer flow, that person clicked past a discount and is gone. In a conversation, that person just told you exactly what would have kept them, and often something you can still fix for everyone else.

Where does the why come from?

Underheard gets the why from dialogue. The first answer in any cancellation is polite, and the reason behind it appears one good follow up later. Because the agent can react, probe gently, and offer a fix, the why arrives with context: what stalled, when, and what would have changed the outcome. The root cause report ranks those reasons by frequency and revenue, with quotes attached.

Churnkey gets the why from forms and telemetry. Exit surveys capture a category, follow up questions can capture a typed sentence, session recordings show behaviour, and Feedback AI aggregates it into revenue weighted themes. That is honest, useful analysis. Its input stays whatever a leaving customer is willing to type into a box, and most type little.

Churnkey's session recordings deserve their own mention, because they answer a question conversations do well to leave alone: how people behave inside the flow itself, where they hesitate, and which screen loses them. For flow optimisation that is genuinely useful telemetry. It shows what happened on the screen, while the conversation explains what happened in the company.

The distinction is depth per answer. A treemap of checkbox categories tells you pricing leads. A hundred conversations tell you the price was fine and setup never finished, which is a different roadmap entirely.

What an Underheard conversation sounds like

A condensed exchange in the shape of a real cancel flow conversation, this time from the reason column no offer engine can fix:

No discount was offered and none was needed. The feature existed, the customer never found it, and a survey checkbox reading missing features would have filed this account under roadmap gaps forever. Multiply that across a quarter and you learn how much of your churn is actually a discoverability problem wearing a feature request costume.

What about failed payments?

Here Churnkey wins outright, and this page says so without hedging. Involuntary churn from failed payments can account for a large share of total churn, and Churnkey ships a complete recovery stack: machine timed retries by decline code, unlimited email and SMS dunning campaigns, in app recovery walls, and white labeled recovery pages. Their vendor reported numbers reach as high as 89 percent of failed payments recovered, and their VEED case study cites over 14,000 recovered payments.

Underheard has no dunning product. A customer whose card expired has no reason to be interviewed, and a conversation cannot fix a decline code. If failed payments are where your revenue is bleeding, start with a dunning tool, whether Churnkey, Paddle Retain, or Churn Buster, before thinking about conversations at all.

The clean way to see it: dunning addresses the customers who never chose to leave. Underheard addresses the ones who did, and asks them why.

Do discounts fix reasons?

An accepted offer is a real save, and Churnkey's evidence base here deserves respect: vendor reported save rates around 54 percent of flow completers, over 10 million cancel sessions processed, and case studies like VEED's roughly 35 percent save rate lift. For this month's retention number, a tuned pause or discount demonstrably works.

What an offer does with the underlying reason is a separate question. The customer who cancelled over a setup problem and accepted 40 percent off still has a setup problem, and the renewal conversation arrives with the same reason and a smaller margin. Churnkey's own Adaptive Offers pitch acknowledges the cost side by promising fewer and smaller discounts for the same saves.

Whether an honest conversation saves more customers than a tuned offer is an open question, and we refuse to answer it with marketing. It is exactly what Underheard's A/B measures per client, against whatever flow you run today, offer engine included. The reasons in the report, meanwhile, compound in value either way.

Pricing: what each actually costs

Underheard prices on conversation minutes. The mid tier lands around $2,000 per month for about 650 conversation minutes, roughly 200 conversations of 2 to 4 minutes, sized on the minutes you need. Design partners currently get 50 to 60 percent off for the first six months. Entry is a demo and a signed pilot with the A/B built in.

Churnkey publishes an entry price and hides the rest. Starter costs $250 per month billed yearly, aimed at teams with under $5,000 in monthly churn volume, with a 14 day free trial and no credit card. Core, Intelligence, and Enterprise are custom quoted for larger churn volumes, and reviewers grumble about exactly that: one G2 reviewer calls it expensive and wishes prices were public, and a Trustpilot reviewer reports a real world minimum around $600 per month. The model is flat fees by volume tier, with no revenue share.

The honest summary: Churnkey's entry point is far cheaper and self serve, and at small churn volumes it is the accessible choice. Underheard sits in a different weight class because a conversation costs more to run than a popup, and it justifies the difference only through the uplift number and the report. That is a bet the pilot exists to test.

What does integration actually look like?

Churnkey is the faster start and earns the point. The claimed 35 minute integration against Stripe, Paddle, Braintree, or Chargebee is realistic for a standard setup, the 14 day trial needs no credit card, and a competent developer ships a first cancel flow the same day. Flows, offers, and surveys are then configured in their dashboard without further engineering.

Underheard's setup is one embedded invitation in your cancel flow, placed after the legally free cancellation path, plus the trigger wiring for who gets invited. The heavier lift sits in the words: the conversation guide is tuned to your product and your customers' vocabulary during the pilot, together with the design partner team. Expect days to a first live conversation rather than an afternoon, with the A/B running from the first day.

Ownership also differs. A Churnkey flow typically lives with whoever owns billing or growth tooling. An Underheard loop lives with the team that owns the roadmap, because the report is where its value lands.

Which should you pick?

Can you run Underheard and Churnkey together?

Yes, and for many teams that is the strongest setup, with one rule: keep the jobs separate. Churnkey owns what it is best at, failed payment recovery, where Underheard does not compete at all. For the voluntary cancel moment, pick one owner of the experience, because a customer who meets a survey, an offer, and an interview invitation in one flow meets a committee.

The practical patterns we see: either Underheard takes over the conversation moment entirely while Churnkey keeps dunning, or the two run as a straight A/B, offer flow against conversation flow, and the uplift number decides. Both patterns leave you with the root cause report either way, which is the part no offer engine produces.

Starting with Underheard next to, or instead of, Churnkey

There is no migration in the technical sense, and nothing about starting requires ripping anything out. The path looks like this:

1. Decide the split. Underheard owns the voluntary cancel conversation; Churnkey, if you run it, keeps failed payment recovery. Or set up a straight A/B between offer flow and conversation flow.
2. Embed the invitation. One prompt after the legally free cancellation path, voice or chat, declinable in one click.
3. Tune the conversation. During the pilot the guide learns your product's vocabulary and your customers' real objections.
4. Read the loop. Live dashboard between reports, and the recurring root cause report with quantified reasons, quotes, and recommendations.
5. Judge it by the number. Save rate uplift against your current flow, offer engine included. Design partner terms make the first six months cheap to prove.

Alternatives to both

The cancel flow and retention space is crowded, and depending on your bleeding point, a different tool fits better:

• ProsperStack and Chargebee Retention are the closest offer engine rivals to Churnkey, with similar flow builders and published entry pricing.
• Raaft plays the lightweight end of cancel flows for smaller teams.
• Paddle Retain and Churn Buster focus on the dunning side of the problem.
• Listen Labs sits on the research side: recruited interviews for market questions rather than a save motion in the product. We compared it in detail in Listen Labs vs Underheard.
• User Intuition runs AI churn interviews as research, without acting inside the cancel flow.

Key takeaways

Frequently asked questions

Henri Winkeler

Co-Founder of Underheard. Spends his days turning customer conversations into decisions, and writing down what the tools in this space actually do. Every comparison on this site is researched from public sources and updated when the facts change.

Related comparisons