I started calculating 'revenue protected per support interaction.' It changed how I think about the team.
For the first two years, I thought about support cost in one direction:
Cost per ticket. Salary per agent. Total support spend as a percentage of revenue.
All inputs. All costs.
About six months ago I started calculating in the other direction.
Revenue protected per support interaction.
Here's how I built the model:
Step 1: Calculate true customer lifetime value
For us: average ACV × average retention in years + estimated expansion revenue + estimated referral value.
Our number came out to approximately $12,400 per customer over their lifetime with us.
Step 2: Estimate what percentage of support interactions involve a customer who might churn
We tagged support tickets for 60 days. About 18% of tickets came from customers showing at least one churn signal (declining usage, billing questions, "how do I export" questions, second complaint in 30 days).
Step 3: Estimate what percentage of those customers were retained by a good support interaction
This is the hard one to measure. We used a combination of:
- Post-resolution survey asking if the interaction changed their view of the product
- 90-day retention comparison between customers who had churn-signal tickets resolved well vs. those who had them resolved poorly (measured by response time and follow-up)
Our estimate: good support retained approximately 22% of customers who would otherwise have churned.
Step 4: Do the math
In a month with 400 tickets:
- 72 from customers with churn signals (18%)
- ~16 retained by good support (22% of 72)
- Revenue protected: 16 × $12,400 = $198,400
Our total support cost that month: $6,800.
Revenue protected: $198,400.
I'm not claiming this model is perfectly precise. There are assumptions in it.
But even if it's 50% wrong, the ROI calculation still makes the investment in support quality obvious.
The framing shifted everything for me. Once you see support as revenue protection rather than just cost, every investment decision looks different.
Has anyone else built a model like this? I'm curious how others think about the return side of support investment.
