What you'll learn: how to build and run a repeatable revenue-quality screener for public SaaS stocks, updated for October 2026 realities — consumption pricing, AI-driven monetization, and tighter capital discipline. This guide is for individual investors and portfolio managers who want a system to prefer durable recurring revenue and avoid growth that doesn't convert to cash.

Why focus on revenue quality (updated context)

Headline growth is still seductive, but through 2026 the market has punished companies whose growth relies on transitory items: large one-off services, pass-through compute, or aggressive accounting for AI credits. Revenue quality — the degree to which revenue is predictable, recurring, and cash-convertible — remains the best single differentiator between durable SaaS winners and short-term stories.

New in 2026: two structural shifts make revenue-quality screening more important:

  • Wider adoption of consumption-based pricing (usage, credits, inference billing) increases quarter-to-quarter revenue volatility unless normalized.
  • Many SaaS vendors now monetize embedded generative AI features. These streams can have very different margins (compute-heavy) and contract terms than subscription fees, so investors must separate them when assessing sustainability.

Prerequisites / Context

Before you start: pick one data stack (commercial like FactSet, Bloomberg, YCharts, Sentieo, Capital IQ or an open stack using SEC filings + Python). You will need quarterly filings (10-Q/10-K), investor decks, and access to management commentary. Budget time for manual extraction: ARR definitions, NRR cohorts and AI/consumption footnotes still require human review in many filings.

Overview: Updated 10-step process

  1. Define the universe
  2. Pull raw metrics
  3. Normalize revenue and adjust one-offs, consumption, and AI
  4. Compute core SaaS and new consumption-adjusted retention metrics
  5. Assess unit economics including compute and inference costs
  6. Measure margin & cash conversion (separate subscription vs consumption)
  7. Score and rank (add consumption/AI weight)
  8. Qualitative overlay and new red flags
  9. Portfolio construction and sizing
  10. Monitoring, rebalance and event-trigger rules

Step 1 — Define the universe

Start narrow and purposeful. Recommended filters (actionable):

  1. Industry: Application software / SaaS / Cloud platforms (SIC/NAICS primary).
  2. Listing: public companies in OECD markets for consistent disclosure.
  3. Market cap: >$1bn for core screening (expand to >$250m for opportunistic ideas).
  4. TTM revenue: >$100m to ensure scale of reported metrics (you can relax this for high-quality early-stage names with transparent unit economics).
  5. Exclude companies where >30% revenue is pass-through reseller or marketplace fees unless you can map margins cleanly.

Why: a focused universe reduces false positives from small microcaps with thin disclosures and decreases manual workload when extracting cohort charts and AI revenue disclosures.

Step 2 — Pull raw metrics

Collect the following each quarter (minimum):

  • TTM revenue and YoY growth
  • Reported ARR, subscription revenue, and percentage recurring
  • Gross margin (GAAP and adjusted), segmented by subscription vs consumption where available
  • Operating expenses: S&M, R&D, G&A in dollars and % of revenue
  • Free cash flow (FCF), FCF margin, and cash conversion metrics
  • Customer counts, average contract value, and cohort charts (if provided)
  • NRR and GRR (existing-customer retention metrics) and any cohort definitions
  • Explicit disclosures on AI-related revenue, credits, or compute pass-throughs

Why: these inputs let you separate sustainable, high-margin subscription streams from variable, often lower-margin consumption and one-time items.

Step 3 — Normalize revenue and adjust one-offs, consumption, AI

Normalization is now broader than removing professional services. Add these checks:

  1. Separate subscription ARR from consumption/usage revenue and stateful marketplace fees. If not dissected in filings, estimate using management commentary, product-based revenue disclosures, or slide deck charts.
  2. Identify AI-related revenue lines: labeled “AI credits,” “inference revenue,” or “feature monetization.” Tag them and estimate their gross margin impact (compute cost per inference can be material).
  3. Remove or flag pass-through vendor or cloud provider reimbursements (these inflate top-line but do not reflect company economics).
  4. Adjust for large non-recurring license deals or implementation projects using the company’s commentary to scale back their contribution to sustainable revenue.

Result: an adjusted recurring revenue number and a small set of segmented revenue streams you can analyze separately.

Step 4 — Compute core SaaS metrics (plus consumption-adjusted retention)

Standard metrics remain essential, but add consumption-aware variants:

  • Recurring Revenue Share (RRS): subscription revenue / total revenue. Target: >70% for pure SaaS. If a company has significant usage revenue, compute subscription RRS and subscription+stable consumption RRS.
  • Net Revenue Retention (NRR): existing-customer revenue end period / start period including upsells. Target: >100% (best-in-class >110%). For consumption-heavy names, compute NRR both including and excluding usage to see stability.
  • Gross Revenue Retention (GRR): retention before upsells. Healthy mature products: >85–90%.
  • Consumption-Adjusted NRR (cNRR): NRR normalized for volatility in usage (e.g., use 12-month trailing median consumption vs quarter). This helps spot growth driven purely by temporary spikes in usage.

Why: consumption spikes can mask churn in subscription base. Two NRR figures — raw and consumption-adjusted — reveal the true stickiness of the base.

Step 5 — Assess unit economics (updated for compute and AI)

Unit economics must reflect the full cost base:

  • CAC: S&M spend divided by new ARR (or new subscription ARR). For consumption-first sellers, calculate CAC per new subscribing customer (not per dollar of usage).
  • LTV: use subscription gross margin and stable churn to estimate LTV; for AI/usage revenue, split LTV into subscription LTV and usage LTV, accounting for variable compute costs.
  • LTV/CAC ratio: target >3 for scalable subscription models. For businesses whose revenue mix is usage-heavy, expect lower LTV/CAC and compensate with higher margin per subscription or longer payback windows.
  • Compute payback period on CAC using recurring margin (not revenue) — including incremental compute costs for AI features.

Example: a company that reports strong top-line growth driven by inference fees may show a healthy CAC payback on revenue but poor payback on contribution margin once cloud compute costs are included. Model both.

Step 6 — Measure margin & cash conversion (separate by stream)

Split margin and cash metrics by revenue stream where possible:

  • Subscription gross margin: long-term leaders often exceed ~70% after scale.
  • Consumption gross margin: frequently lower because of pass-through cloud costs; normalize separately.
  • FCF margin and conversion: FCF / revenue and FCF conversion (FCF / EBITDA or FCF / net income). Watch for companies with high accruals or large D&A adjustments — conversion rates under 50% deserve scrutiny.
  • Compute cost trend: track cloud provider bills (companies sometimes disclose % of cost increases YoY). Rising compute as % of revenue compresses contribution margins quickly.

Why: a blended gross margin masks weakening economics when lower-margin consumption grows faster than subscription ARR.

Step 7 — Score and rank (new weighting)

Update the composite revenue-quality score to reflect consumption and AI risk. Example weighting you can start with:

  • Subscription RRS (20%)
  • NRR (15%)
  • Consumption-adjusted NRR (10%)
  • FCF margin (15%)
  • Gross margin (subscription-focused) (10%)
  • LTV/CAC (15%)
  • Revenue growth (15%)

Score each metric on a 0–100 band using sector-appropriate benchmarks (use BVP/Bessemer/SaaS Capital reports for trend context). Define buckets for actions:

  • 80–100: Core — prefer larger position sizes and longer-term holds.
  • 60–79: Watchlist — buy only after qualitative checks or on valuation pullbacks.
  • 60: Avoid or deep-dive before considering.

Step 8 — Qualitative overlay and red flags (2026 additions)

Quant filters miss governance and product nuances. Add these 2026-specific checks:

  • Customer concentration: >20% from one client remains a major red flag.
  • AI revenue disclosures: lack of clear definitions or aggregation of AI with subscription revenue is a red flag.
  • Channel and marketplace economics: platform marketplaces that recognize gross revenue can materially overstate take-rates — verify net vs gross presentation.
  • Cloud-provider dependency: single-cloud reliance with rising unit costs requires scenario stress-testing.
  • Contract terms: true multi-year, dollar-denominated contracts are higher quality than month-to-month usage contracts.

Step 9 — Portfolio construction and sizing

Translate scores into tradable rules keeping risk controls tight:

  1. Allocate within your SaaS sleeve: 60% Core (score ≥80), 30% Watchlist (60–79), 10% cash/opportunistic.
  2. Position sizing: cap any single core holding at 6–8% of portfolio; watchlist names 2–4% with mandatory stop-loss or fundamental exit rules.
  3. Entry discipline: prefer entry on pullbacks where revenue-quality metrics are stable or improving; avoid buying growth-first moves that widen unit-economics deficits.
  4. Valuation overlay: use EV/Revenue for high-growth names but require stronger quality for valuation premiums. For usage-heavy firms, prefer lower EV/Revenue unless margins and cNRR justify premium.

Step 10 — Monitoring, rebalance and exit rules

Set clear, repeatable triggers:

  • Quarterly refresh: update core metrics after 10-Q/earnings and re-run scores.
  • Rebalance schedule: formal review semi-annually; rebalance toward core names and trim those whose score drops >15 points.
  • Fundamentals-based exits: cut positions when FCF margin turns negative and the company lacks a credible path to positive conversion within 12–18 months.
  • Event rules: re-evaluate after material M&A (especially acquisitions that add low-margin consumption streams), management change, or disclosures that change ARR/NRR definitions.

Practical implementation tips (updated)

  • Data sources: combine SEC filings and investor decks with BVP/Bessemer/Bain SaaS reports and SaaS Capital benchmarks. For compute-cost trends use vendor pricing pages (AWS, GCP, Azure) and cloud-cost monitoring reports from Cloudability or CloudHealth.
  • Extraction: many companies still publish NRR/GRR in slide decks only. Build a simple pipeline (Python + pandas + manual review) that stores cohort images and extracted numbers for reproducibility.
  • Backtesting: simulate the scoring logic over at least a 4–5 year window, include transaction costs and slippage, and run sensitivity tests where consumption revenue grows faster than subscription revenue.
  • Real examples to practice on: established names that disclose ARR and cohort charts (e.g., Salesforce, ServiceNow, Atlassian, HubSpot, Snowflake) are useful training cases — specifically compare raw NRR with consumption-adjusted NRR where applicable.

Common mistakes and how to avoid them

  • Treating consumption revenue like subscription: normalize usage revenue and compute margins separately.
  • Relying on headline NRR without definitions: confirm cohort composition and whether NRR includes new product lines or acquired revenue.
  • Ignoring cloud cost trends: rising inference or data-transfer costs can materially change profitability within quarters.
  • Overfitting to short timeframes: use rolling TTM or 12-month medians to smooth seasonality and project-based revenue.

Pro tips

  • Build a dual-scorecard: subscription-quality score and mixed-revenue risk score. Use the former to size positions and the latter to set monitoring intensity.
  • Model three scenarios for each company: Base (current mix), Upside (subscription regain or margin improvements), and Downside (consumption growth outpaces subscription and compresses margins). Compare portfolio-level downside across scenario sets.
  • Automate red-flag alerts: customer concentration changes, a 10-15 point decline in NRR, or a >200 bps QoQ increase in cloud costs should trigger an analyst review.

FAQ

How should I treat AI-related revenue in my model?

Separate AI-feature revenue from subscription ARR. Tag it as consumption if billed by inference or credits. Estimate incremental compute costs and model a lower contribution margin for that stream unless the company discloses efficient inferencing or value-based pricing with higher margins.

What if a company reports NRR but doesn't show cohort definitions?

Treat reported NRR as directional, not definitive. Request clarification from investor relations if possible, and reconstruct retention from ARR bridges or cohort charts in past investor decks. If you cannot infer cohorts, downgrade the confidence weight in your score.

Can I apply this screener to smaller-cap SaaS names?

Yes, but expect more manual work and higher disclosure risk. Lower market-cap firms often omit cohort charts and detailed ARR breakdowns. Consider raising your market-cap threshold or applying a larger margin-of-safety to position sizing for small-cap picks.

How often should I rebalance based on screener scores?

Quarterly metric refreshes and semi-annual portfolio rebalances are practical. Use rolling triggers for score deterioration (e.g., drop >15 points) to avoid overtrading while ensuring you react to material fundamental changes.

Where should I look for benchmarks to set score thresholds?

Use industry benchmark reports from Bessemer, BVP, and SaaS Capital for cohort, NRR and margin medians. Supplement with sector-specific data from company filings and repay your model with peer medians in the same revenue growth/maturity bucket.

Wrapping up

Revenue quality remains the most enduring signal for long-term SaaS performance. In October 2026 the incremental work is separating subscription ARR from consumption and AI-derived revenue, and explicitly modeling cloud/compute costs. Implement this updated 10-step screener, run backtests across scenarios, and operationalize monitoring rules. Over time the discipline of scoring, qualitative overlay, and scenario stress-testing will help you own fewer structural losers and more durable SaaS winners.