T27

The MRR walk became so central to SaaS analysis (ARR, NRR, GRR, CAC payback all derive from it) that practitioners began applying it to non-subscription models where it does not hold. The claim that the walk is 'the subscription case of a family' is contested because many non-SaaS companies have tried to adapt ARR-style metrics to their businesses and found that investors accept it — suggesting the community has not fully internalized this limitation. The practical risk: applying subscription-model GTM logic (maximize NRR, minimize churn) to a usage-based or transactional model produces structurally wrong decisions.

The MRR walk identity (MRR(t+1) = MRR(t) + New + Expansion - Contraction - Churn) assumes that revenue is a stock — it persists from period to period unless explicitly churned. This is only true for contractual recurring revenue (subscription SaaS). For other revenue models, the appropriate stock variable is different: (a) Usage-based (Snowflake, Twilio): revenue stock is consumption capacity and data volume; the walk is a utilization walk, and 'churn' is replaced by 'usage decline'; (b) Transactional (Stripe, Toast POS): no stock — each transaction is independent; the correct model is a repeat purchase probability (BG/NBD), not a retention rate; (c) Marketplace (Airbnb, Upwork): two-sided stock — host supply and guest demand; the walk requires separate equations for each side; (d) Consumption (AWS, Azure): revenue = stock of committed capacity * utilization rate; the optimization target is utilization, not churn. Applying subscription metrics to these models systematically understates churn (because partial usage reduction looks like expansion in ARR terms), misvalues customers, and generates wrong GTM investment priorities.

• Snowflake's investor day (2022): explicitly guided away from NRR as a primary metric for usage-based cloud, introducing net revenue expansion rate with a different formula — the company recognized the subscription-model limitation
• Twilio S-1: used 'dollar-based net expansion rate' rather than NRR, acknowledging that the usage-based walk is structurally different from subscription ARR
• Stripe and other payment processors: do not report ARR or NRR; report transaction volume, TPV growth, and net take rate — the subscription walk is inapplicable to the transactional model
• Fader-Hardie BG/NBD model: the correct mathematical framework for non-contractual (transactional) customer bases produces fundamentally different retention curves than the sBG model designed for contractual SaaS

• Many usage-based companies have hybridized their business models with subscription floors (AWS Reserved Instances, Snowflake committed capacity) that partially restore the MRR walk applicability
• The conceptual framework (grow new customers, retain existing, expand) applies qualitatively to all models even when the exact walk equation does not
• Investors have accepted adapted MRR-walk metrics for non-subscription models (annualized run rate, GMV equivalent) — the claim may be more precise than the market requires

For each revenue model in your company, identify the correct stock variable before building your GTM dashboard. If subscription: use the MRR walk directly. If usage-based: use a utilization/capacity expansion walk. If transactional: use repeat purchase rate and BG/NBD survival curves. If marketplace: use separate supply and demand metrics. The wrong stock metric not only misleads; it misaligns GTM investment — a usage-based company that optimizes for 'churn reduction' (subscription logic) when it should be optimizing for 'consumption expansion' (usage-based logic) will under-invest in customer activation and use case expansion.