By Brendan Clay, Director of Venture Debt & Innovation Banking Group, and Kenneth Sechler, Managing Director of Specialty Lending
For the better part of two decades, lending into software businesses has rested on a comfortable assumption: that recurring revenue is, in fact, recurring. A subscription contract is signed, a seat or a license is provisioned, an invoice is issued, and a customer either renews or doesn’t. The diligence framework that grew up around this model (testing ARR build, calculating churn, reviewing contracts, reconciling deferred revenue) works because the underlying business is built around predictable and contractually committed revenue.
That world, however, is steadily shrinking. Usage-based billing, consumption pricing, and outcome-based commercial models are no longer fringe business models. They are becoming a default commercial structure for software companies, and overwhelmingly so for recent or AI-native applications. The reason for this shift is obvious: AI businesses cannot afford to disconnect their pricing from their internal delivery costs because their cost of goods is largely usage-driven.
For lenders in this space, adapting to this shift can’t be accomplished by changing go-to-market or sales & marketing strategies. It is now decidedly a credit and underwriting problem. The diligence playbook that has historically governed the industry was built for a revenue model that is now being displaced.
What is Actually Changing in Software Business
Software billing models exist on a spectrum. At one end sits pure subscription: a fixed annual fee for a defined number of seats or licenses, billed in advance, ratably recognized. At the other end sits pure consumption: a customer pays only for what they use, billed in arrears, with no contractual commitment beyond perhaps an enterprise master agreement. Between those structures lie a number of hybrid structures: subscription with overage, tiered platform fees with consumption-based add-ons, minimum commitments with true-up mechanics, outcome-based pricing tied to specific customer results, and credit-pool models where customers prepay against future consumption.
What is actually changing in software sales is that the center of gravity is moving rightward on this spectrum. New software businesses, particularly those built around AI primitives, are launching with consumption pricing as the default and adding subscription layers after the fact. Established SaaS companies are, themselves, introducing usage-based components to compete. For lenders and investors into the SaaS space, the result is now a population of subject companies whose revenue, while still recurring in the loose sense, is no longer recurring in the contractually committed and predictable sense that the historical underwriting framework was built around.
Where the Standard Due Diligence Toolkit Strains
Traditional due diligence on recurring revenue business models tests several things in service of one underlying set of questions: is the revenue real, is it durable, and is it predictable enough to support the proposed credit facility or investment? The standard toolkit answers that question well in a pure subscription world. However, it does not answer it well in a consumption-based billing world. In fact, in some cases, it can produce actively misleading conclusions. Below is an evaluation of the impact of changing software revenue models and a summary of the diligence toolkit strain:
- Normalization of Revenue – The most obvious challenge for a lender or investor in AI-driven change to a usage-based billing model is that usage-based revenue can and will fluctuate significantly from month to month based on seasonality, experimentation with AI features, temporary pullbacks, cost-optimization cycles, one-off spikes, and other factors. This volatility not only creates difficulty in annualizing MRR, but it can also skew churn or retention analysis if unaccounted for: two key metrics investors and lenders have historically relied upon as KPI’s in the space. For a lender sizing a borrowing base or an investor pricing a deal, the methodology behind that figure is now as material as the figure itself.
- Retention Rate Masking Material Churn – In a seat-based business, net revenue retention (“NRR”) above 100% is a clear positive signal: existing customers are expanding faster than they are churning. The metric is useful because seat expansions and contractions are discrete, contractual events distributed roughly evenly across the customer base. In a consumption-based business, the same metric can mask deterioration. Consider a software company reporting 110% NRR and 95% logo retention over twelve months. This is typically evidence of a healthy, expanding customer base. However, now let’s assume 70% of that revenue expansion came from three customers tripling their consumption, while 60% of the remaining customer base consumed less than it did a year ago. The headline metrics are unchanged, but now the credit profile of the revenue base is not. As revenue models shift in Software, our retention analysis will need to be evaluated at a much more granular level with a stronger emphasis on where the growth and decline are actually concentrated.
- AI Churn Impact – Historically, a dissatisfied SaaS customer might stay on the platform for 12 to 24 months past peak engagement, but AI tools that replace incumbent software can accelerate this abandonment dramatically. A company displaced by an AI-native competitor may see usage collapse months before formal cancellation with minimal warning. To catch this, a Lender or Investor’s diligence process should include consumption decay, as a customer-level analysis of usage trends over time may be the leading indicator of future churn.
- Deferred Revenue in the Age of AI – In an annual-prepay subscription business, deferred revenue is one of the most useful tells in our diligence toolkit. A growing deferred revenue balance signaled healthy bookings and some forward visibility, while a declining balance signaled trouble. However, with revenue models increasingly shifting to usage, invoicing typically happens monthly in arrears and so the construct of deferred revenue is effectively eliminated. This materially impacts the historical diligence framework and eliminates a KPI that was historically significant but is now increasingly rare.
What the Next Generation of Diligence Should Look Like
The implication in today’s world of recurring revenue is certainly not that this form of collateral is broken or evaporating. These businesses are still very much ubiquitous: they exist, the cash flows are very real, and the asset class remains attractive to lenders and investors alike for how quickly they can scale (e.g. the potential for unicorns). However, the implication is that the analytical apparatus we’ve historically used to diligence these businesses must evolve.
For example, a modernized recurring revenue diligence process should treat consumption decay as an analytical pillar alongside traditional churn analysis. It should also test minimum commitments by stratifying customers into cohorts and should track migration across cohorts over the review period. It should require explicit disclosure of ARR annualization methodology and stress-test the revenue base under alternative methodologies. It should treat retention as a multi-dimensional exercise, separately tracking logo retention and (increasingly) usage intensity per retained customer. Finally, it should approach contract review with a sharper eye on the line between contractually committed revenue and merely that which involves “expected consumption”.
Each of these refinements is a reasonable extension of existing methods, and the work is in adapting them to the diligence context while consistently building them into engagement scopes.
The Investment Implication
For lenders and investors, the operative question is whether the diligence work supporting current and prospective businesses is built for the business model actually being financed in 2026. The rapid shift away from subscription billing makes due diligence considerably more important, given the uncertainty and unpredictability it injects into the revenue data. The diligence process that emerges from this transition will certainly look different, and investors and lenders alike should be more skeptical than ever of headline ARR and MRR metrics. They should also be more rigorous about requesting and reviewing consumption-level data, more explicit about the underlying methodology, and more willing to push back on the subject company’s own framing of its revenue. Those who recognize this early will underwrite better deals, and those who do not may discover the gap the hard way.
Brendan Clay leads the Venture Debt & Innovation Banking Group at LCG Advisors, where his team performs due diligence for banks and private credit funds financing software businesses. Kenneth Sechler is the Managing Director of LCG Advisors’ Specialty Lending practice, overseeing the Consumer Finance, Commercial Finance, Healthcare Advisory Services, and Venture Debt & Innovation Banking verticals. The views expressed are their own. For more information, please contact Kenneth Sechler at [email protected].