Save Rate

Save rate measures cancel flows, the real-time system that intercepts a cancellation attempt and presents save offers. Dunning recovery uses Recovery Rate: the Formula instead.

Every outcome counted.

The complete cohort rule is a dunning rule and doesn't apply here. Cancel flows are real time and session based, so there's no cohort waiting period. Results count as soon as they happen, and a later session in the same 24 hour window can still change one. See Two Measurement Frames.

Save rate = Saved / (Saved + Cancelled), counted on definitive sessions. Every definitive session is in the denominator. Nothing is set aside.

The unit: the definitive session

A session is a single instance of a customer entering the cancel flow. A customer clicks cancel, moves through the flow, and either cancels or doesn't, almost always within a single session, usually within minutes.

Customers sometimes come back, though. So save rate doesn't count sessions one by one. It counts definitive sessions.

A definitive session is the single result that combines every session one subscription has within a 24 hour window. The window starts when a cancel session begins. It is the unit of measurement for cancel flows, analogous to a campaign in dunning.

The three outcomes

Each definitive session gets one result. Cancel beats offer, and offer beats abandon.

• Cancelled. The customer completed the cancellation through the flow in any session in the 24 hour window. Counts as a loss.
• Accepted offer. The customer accepted a save offer (such as a skip, pause, swap, discount, or content offer) and didn't cancel through the flow in the 24 hour window. Counts as Saved.
• Abandoned. The customer left the flow, or the session expired, without accepting a save offer, and didn't cancel through the flow in the 24 hour window. It counts as Saved, in the numerator and the denominator.

The formula

Save Rate = Saved / (Saved + Cancelled)

• Saved = definitive sessions that didn't end in cancellation: accepted offers plus abandons
• Cancelled = definitive sessions where the customer cancelled through the flow

A save counts as soon as it happens. A recent result can still change until its 24 hour window closes, and a later session in the same window can replace a result already counted. A customer who skips a box, takes a discount, and cancels a few hours later is one loss, not two saves and a loss.

Why an abandon counts as a save

An abandon can mean "I'll think about it," or it can mean a UX or technical issue pushed the customer out of the flow. Either way the customer didn't cancel through the flow in the window, so it counts as Saved.

What counts as a save

A save is a definitive session that didn't end in cancellation: the customer accepted a save offer or abandoned the flow. Only a cancellation inside the flow counts as a loss, so a customer who later cancels through your support team or your platform's portal still counts as a save.

Tools count saves differently. Some credit a save at each step of the flow, so a customer the flow deflects at one step, who cancels seconds later, still shows up as a save. Before you compare two save rates, check how each tool counts a save: per step or per window, and whether abandons count.

Worked example

Illustrative numbers, not a benchmark.

A cancel flow has 200 definitive sessions counted in the dashboard.

• 25 accepted an offer (Saved)
• 10 abandoned (Saved)
• 165 cancelled (losses)

Save rate = 35 / (35 + 165) = 17.5%.

The same data produces three wrong numbers.

• Set abandons aside: 25 / 190 = 13.2%.
• Count abandons as losses: 25 / 200 = 12.5%.
• Count each entry into the flow instead of definitive sessions. Say 10 of the 165 cancellations came from customers who accepted an offer in an earlier session in the same window. Counting entry by entry turns those into 10 extra saves: 45 / 210 = 21.4%.

Tracking save rate over time

In cancel flows, rolling windows smooth volume variance only.

Watch cancel volume, the number of definitive sessions, next to save rate. A flat save rate with rising cancel volume means more customers are trying to leave while the flow holds its share. A falling save rate on flat volume means the flow is saving fewer of the same attempts. Together the two numbers answer the cancel flow side of "is churn going up?"

"Cancellation" means two things

In dunning, a cancellation is a customer cancelling while a recovery campaign is running, and it counts as Lost in the recovery rate. Here it's a definitive session that ended in cancellation, counted as a loss in the save rate. See Two Measurement Frames.

Testing copy

A flat A/B result means the same thing on cancel flow copy as on dunning copy. When a test runs long enough on enough volume and shows no separation, the copy is already inside the narrow band of what good looks like, and wording isn't the lever. The general principle is on What Moves a Recovery Rate.

Don't measure it this way

• Setting abandons aside. An abandon with no cancellation in the window counts as Saved, in the numerator and the denominator.
• Counting abandons as losses. The customer didn't cancel through the flow, so each counts as Saved.
• Counting each entry into the flow as its own result. One subscription's sessions within 24 hours are one result.
• Applying the dunning cohort wait to cancel flow sessions. A definitive session counts immediately and is final once its 24 hour window closes.
• Comparing a save rate to a recovery rate. One counts definitive sessions, the other counts campaigns.
• Comparing two tools' save rates without checking how each counts a save: per step or per window, abandons in or out.

---

Prerequisite: Two Measurement Frames.

Next: Realistic Recovery Rate Range, in Section 5, Benchmarks.