Case 008SaaS metrics and diagnosticsCore

Read Vantrel Analytics' cohort table: three yearly cohorts, each measured today. What is each cohort's annual net revenue retention, and what does the trend mean for the growth plan?

The situation

Vantrel Analytics sells demand forecasting software to consumer goods distributors. Its data room groups customers by the year they signed. The 2022 cohort started at Rs 4 crore of ARR and is worth Rs 5.6 crore today, three years later. The 2023 cohort started at Rs 7 crore and stands at Rs 7.7 crore after two years. The 2024 cohort started at Rs 12 crore and stands at Rs 11.0 crore after one year.

Management's plan takes ARR from today's Rs 24.3 crore to Rs 34 crore next year, assuming 110% net revenue retention on the existing base and filling the rest with new customers.

Your task

Compute the annualised net revenue retention of each cohort, say what the trend tells you, and test the plan against it.

Quick check

The 2022 cohort grew from Rs 4 crore to Rs 5.6 crore in three years. What is its annual NRR?

Worked solution

Try it on paper, then open one step at a time.

30-second answerThe answer to give first

Annual NRR has fallen cohort by cohort: about 112% for 2022, 105% for 2023 and 92% for 2024. Weighted by today's ARR, the base would grow about 0.5% next year, not the 10% the plan assumes, so new sales must bring Rs 9.6 crore instead of Rs 7.3 crore. Newer customers look worse, and the plan has not priced it in.

Step 1How do you turn cohort totals into a yearly retention rate?

A savings account that grows from Rs 4,000 to Rs 5,600 over three years did not earn 13.3% a year; it earned the rate that, compounded three times, turns 4,000 into 5,600. Net revenue retentionRevenue from a fixed group of customers now, as a percentage of what the same group paid a year earlier. Expansion lifts it; downgrades and churn lower it. compounds the same way, so annualise with a root, not a division. The 2022 cohort: 5.6 over 4 is 1.4, and the cube root of 1.4 is 1.1187, about 112% a year. The 2023 cohort: the square root of 1.1 is 1.0488, about 105%. The 2024 cohort: 11 over 12 is 0.9167, about 92%.

The relationship
NRRannual=(ARR todayARR at start)1/years(5.64)1/3=1.119\text{NRR}_{\text{annual}} = \left(\frac{\text{ARR today}}{\text{ARR at start}}\right)^{1/\text{years}} \qquad \left(\tfrac{5.6}{4}\right)^{1/3} = 1.119
ARR todaywhat the cohort pays now
ARR at startwhat it paid in the year it signed
yearsthe cohort's age
What it says in wordsDivide where the cohort is by where it began, then take the root for its age to get the yearly rate.
Each cohort's ARR, indexed to 100 when it was signed801001201402022: 140, 112% a year2023: 110, 105% a year2024: 92, 92% a yearYear 0Year 1Year 2Year 3Years since the cohort was signed
Indexed to 100 at signing, the 2022 cohort compounds at about 112% a year to 140, the 2023 cohort at about 105% to 110, and the 2024 cohort shrinks to about 92 in its first year.
Step 2Is the falling trend about worse customers or just younger ones?

Be careful here, because cohorts of different ages are not directly comparable. Many software customers trim seats in their first year and expand later. The fair test is to compare every cohort at the same age, so ask for each cohort's first-year NRR. You can already bound the answer: if the 2022 cohort had also fallen to 92% in its first year, it would have needed about 123% a year in years two and three to reach 140. That is possible but rare, so the more likely reading is that the 2024 customers are genuinely weaker, perhaps smaller distributors won by discounting.

Step 3What does it do to next year's plan?

Apply each cohort's own rate for one more year: Rs 5.6 crore grows to 6.26, Rs 7.7 crore to 8.08, and Rs 11.0 crore shrinks to 10.08. The base becomes Rs 24.42 crore, a blended NRR of about 100.5%, because the weakest cohort is also the largest. The plan's 110% assumption borrows the oldest cohort's behaviour and applies it to a base dominated by the newest. To reach Rs 34 crore, new sales must bring Rs 9.58 crore, not Rs 7.27 crore.

The Rs 34 crore plan with 110% retention, and with the cohorts' ownPlan: 110% NRRExisting base 26.73New 7.27Cohort NRRExisting base 24.42New 9.58Target: Rs 34 croreNew sales must be 32% larger than the plan assumes, because the base barely grows.
At the plan's 110% retention the existing base supplies Rs 26.73 crore of the Rs 34 crore target, but at each cohort's own retention it supplies Rs 24.42 crore, so new sales must rise from Rs 7.27 crore to Rs 9.58 crore.

Say the limitation plainly: three cohorts and one data point each is thin evidence, and applying last year's rate forward is an assumption, not a forecast. But the direction is clear enough to change the conversation. The questions to ask are which customers are in the 2024 cohort, what discounts they were signed on, and how much of their decline is seats removed versus whole customers lost. A company whose newest customers shrink needs to fix who it sells to before it spends more on selling.

Where candidates lose it

The trap is dividing total growth by years. Calling the 2022 cohort 113% instead of 112% looks harmless, but the same habit turns a two-year 110% into 105% by luck rather than method, and interviewers watch for the root.

The second loss is averaging the three rates into one number of about 103% and calling retention fine. The cohorts are different sizes and moving in one direction; the trend and the weighting are the whole answer.

What the interviewer asks next

  • The 2024 cohort's decline is all from three customers that churned. Does your view change?
  • How would gross revenue retention help separate churn from downgrades here?
  • What would you want to see in the 2025 cohort's first six months before investing?

Company names and figures are illustrative.