Churn looks simple on paper. In practice, it rarely is.

In a clean SaaS model with consistent monthly revenue, churn is straightforward: lose a customer, lose a predictable stream of revenue.

Put once you look at a transactional or payments business, the picture becomes more complex. Revenue becomes uneven with seasonality, usage-based billing, or fluctuating customer behaviour. Measuring churn in these cases becomes far less obvious.

The Core Problem

The difficulty is not the mechanics of calculating churn, it’s the methodology of defining what “lost revenue” actually means.

Depending on the methodology, the same customer can produce very different churn outcomes. You can measure churn based on:

  • Revenue in the final month before churn
  • Revenue 12 months prior (annualised)
  • Total revenue over a period (usually 12 months)

Each approach produces a different answer, and each with its own flaws.

The problem is not that the calculation is wrong. The problem is that the definition is unclear.

ARR Methodology vs Reality

Traditional ARR-based approaches assume consistency. If a customer generates $40 every month, then losing that customer means losing $40 × 12.

That logic works in a true SaaS environment. It breaks down when revenue fluctuates. Side note: In my view, ARR and transactional revenue are fundamentally different metrics, and trying to apply ARR logic to transactional revenue is a category error. From a commercial perspective, it’s often useful to annualise revenue for valuation purposes and depending on your selected annualised month, could result in a higher equity valuation.

If one month is $16 and another is $40, what exactly have you lost?

The answer depends entirely on which month you choose which introduces subjectivity into what should be an objective metric.

Churn vs Downsell

A second issue is classification.

Many methodologies split revenue movements into:

  • Churn (lost customers)
  • Downsell (reduced spend)

But in practice, these are often part of the same story.

A customer may gradually reduce spend over several months before fully churning. The question becomes:

Where does downsell end and churn begin?

Different methodologies answer that question differently, which leads to materially different results.

The Impact of Seasonality

Seasonality adds another layer of complexity.

Adjustments can help smooth data at an aggregate level, but they often fail at the individual customer level.

In cases where ARR is used, it creates a risk:

  • Churn may be overstated if a high month is used
  • Churn may be understated if a low month is used

In both cases, the metric becomes sensitive to timing rather than underlying performance.

New Revenue vs Upsell

Another common distortion appears in Net Revenue Retention (NRR).

When customers ramp up over time, increases in revenue can be misclassified as upsell, when in reality they are simply the natural progression of a new customer.

Separating true upsell from new customer ramp can materially impact the accuracy of retention metrics.

No Perfect Method

The uncomfortable conclusion in some limited analysis that I have done is that there is no single “correct” way to measure churn in these situations.

Each method has trade-offs:

  • ARR-based methods can overstate churn
  • Period-based methods can understate it
  • Seasonality adjustments can distort it

In many cases, the best approach is not to choose one metric, but to understand the story behind them.

A single churn number can be misleading. A well-explained set of metrics is far more useful.

Closing Thought

Metrics like churn and NRR are often treated as objective truths.

In reality, they are models and all models are simplifications.

The role of finance is not just to calculate these numbers, but to interpret them, explain their limitations, and ensure they reflect the underlying business as closely as possible.