α
AlphaSteve
← Brain index

SaaS / Subscription Economics

Adopted 2026-06-10 to close the gap identified in kit-debrief-001-PLTR (no SaaS-economics module). See changelog-2026-06-10-hard-changes.

The problem this fixes: the kit's first real test (PLTR) was a software business, and the debrief found the kit reaching for SaaS metrics ad hoc with no codified standard. Software is a large share of the modern opportunity set; analyzing it from a generic industrial template produces exactly the normalization errors the quality-of-earnings file warns about. This module gives software names a proper lens — and, critically, keeps them inside the deep-value discipline rather than letting "recurring revenue" become a license for narrative.

The core translation: subscription revenue → owner earnings

The deep-value question is unchanged — what can the owner pull out, sustainably, with no growth? SaaS just hides the answer behind growth spend and capitalized costs. The work is to separate the cost of running the existing book from the cost of growing it.

  1. Strip growth S&M. A no-growth SaaS business still spends to retain the base, but not to acquire net-new logos. Estimate maintenance S&M (the spend tied to renewals/retention) vs. growth S&M (new-logo acquisition). EPV uses maintenance S&M only; growth S&M is a discretionary investment evaluated on its return, not a cost of current earnings.
  2. Capitalized commissions (ASC 606) and capitalized internal-use software — add back the capitalization games or, better, expense them at the cash rate. These are the two most common places SaaS earnings get flattered.
  3. Stock-based comp is a real expense, in full. No "adjusted" add-back. Dilution is the cash cost arriving later. This is non-negotiable per quality-of-earnings.
  4. Result: normalized owner earnings for a SaaS book at steady state, which feeds EPV exactly as any other business does.

The metrics that matter (and what each is really testing)

Metric What it is What it actually tests Deep-value read
Net revenue retention (NRR) Revenue from existing cohort a year later, incl. expansion, net of churn Whether the installed base is an appreciating or depreciating asset NRR > 110% = base compounds without new sales; NRR < 100% = a melting asset wearing a growth costume
Gross revenue retention (GRR) Same, excluding expansion The true floor — how sticky is the product really GRR > 90% = genuine switching cost; < 85% = churn problem masked by upsell
Gross margin After cost of hosting/support Whether it is software economics or reselling < 70% = question the "software" label
CAC payback (months) Months of gross profit to recover acquisition cost Whether growth spend earns its keep > 24–30 months = growth that may destroy value (ties to the ROIIC > WACC gating test)
Rule of 40 Growth % + FCF margin % Quality-of-growth sanity check < 40 consistently = growth bought at a loss
Magic number Net-new ARR ÷ prior-period S&M Sales efficiency < 0.75 = each growth dollar returns little

How it plugs into the existing gates

  • The three Greenwald gating tests (see earnings-power-value-greenwald) get SaaS-specific evidence:
    • ROIIC > WACC → tested via CAC payback and magic number (is incremental customer acquisition value-creating?).
    • Durable moat → tested via GRR and switching-cost mechanism (is retention structural or bought with discounts?).
    • Multi-year runway → tested via NRR trajectory and cohort behavior, not TAM slides.
  • The ±20% flip test (banded-valuation-standard) is applied to NRR and CAC payback — the two inputs that carry SaaS growth value. A SaaS buy that flips when NRR is shocked 110%→88% was a narrative, not a thesis.
  • Quality-of-earnings red flags specific to SaaS feed the data-integrity-gate: capitalized commissions, capitalized software, "adjusted EBITDA" that adds back SBC, billings/revenue divergence, RPO that isn't converting.

The anti-narrative guardrail

SaaS is where deep value most often gets talked out of its discipline ("you can't value it on earnings, it's all about growth"). The rule: if the name only works on credited growth value, and the growth value fails the flip test or the gating tests, it is EPV-only — and most high-multiple SaaS is a pass on EPV-only. That is the correct, uncomfortable answer the kit exists to give. The module makes software analyzable, not exempt.

Linked