How do you measure CAC payback under outcome pricing in 2027?

1. Why Traditional CAC Payback Breaks Under Outcome Pricing

The classic formula — CAC ÷ (ACV × Gross Margin) in months — assumes three things that outcome pricing violates. First, ACV is fixed at signature: in outcome pricing, the contract caps usage or sets a per-event rate, but the actual revenue depends on customer demand. Second, gross margin is stable: in outcome pricing, each Fin resolution costs Intercom inference + escalation, so margin moves with the model and the mix. Third, revenue starts on day one: in outcome models, revenue starts when the first event fires, which can be 30-90 days post-close and ramps for 6-24 months.

2. The Five Reformulations

2.1 Replace ACV With T12M

Use trailing-12-month revenue, annualized as the numerator. For a customer 8 months in, T12M = (sum of 8 months × 12 ÷ 8). Snowflake reports 125% NRR on this basis, and Datadog, MongoDB, and Stripe all publish T12M-equivalent metrics in their 10-Ks. The advantage: T12M reflects actual consumption, not the contract ceiling, and removes the perverse incentive to oversize commits.

2.2 Replace Gross Margin With Outcome-Derived Contribution Margin

Strip out the direct COGS of each billed outcome — for Intercom Fin, that is LLM inference + human-fallback cost; for Salesforce Agentforce, the Flex Credit consumption + Hyperforce compute; for Aviso, the forecast-accuracy guarantee reserve. The result is a per-outcome contribution margin, typically 55-72% in 2026 for AI-agent SKUs, versus the 75-82% seat-based SaaS norm.

2.3 Add A Consumption Ramp Curve

Replace the constant-revenue assumption with an explicit ramp curve. Public disclosures and analyst models (Bessemer, Mostly Metrics, OpenView Consumption Sales Index) suggest:

• Months 0-3: 10-25% of run-rate (pilot, integration)
• Months 4-9: 40-70% of run-rate (production rollout)
• Months 10-18: 80-100% of run-rate (steady state)
• Months 19-24+: 100-130% (expansion to new use cases, the NRR lift)

Fold this curve directly into the payback calculation so finance stops over-counting Month-3 revenue.

2.4 Split CAC By Motion

A single blended CAC hides which acquisition motion is actually paying back. The 2027 standard is a four-motion split:

Bessemer Cloud 100 companies report blended payback that masks PLG paying back in 6 months while enterprise drags at 26 — the split is the only way to allocate growth dollars rationally.

2.5 Leading-Indicator Payback

Before T12M exists, build a proxy payback from leading indicators: activation rate (first valuable outcome within 14 days), expansion velocity (second use case within 90 days), and outcome density (events per active user per week). Stripe, Twilio, and Snowflake all use a "committed consumption forecast" built from these proxies as their internal CAC-payback truth source.

3. The Math, Worked

Consider an Intercom-like AI agent vendor closing a $100K committed-spend outcome contract.

Two numbers fall out: the steady-state payback (the marketing-deck number) and the ramp-inclusive payback (the CFO number). Publish both. Hiding the ramp is how Series B companies surprise themselves at the Series C raise.

4. AI-Attributed CAC: The New Boardroom Metric

When 11x Alice, Artisan Ava, and Clay workflows source pipeline, the conventional CAC denominator quietly shifts. The 2027 instrumentation pattern:

• Stamp every opportunity with a source_agent_id in Salesforce.
• Roll those into a quarterly AI-attributed pipeline % (Bessemer Cloud 100 median crossed 22% in Q1 2027).
• Compute AI-attributed CAC = (agent platform spend + AI Agent Ops headcount + inference) ÷ (closed-won ARR sourced by agents).
• Benchmark against human-AE CAC. When AI-attributed CAC is 40-60% below human-AE CAC, OpenView and Tomasz Tunguz call it the "agent advantage line" and recommend rebalancing pipeline mix.

Clari, Gong, Salesloft, and Outreach all shipped AI-attribution columns in their 2026 releases, so the data is now native to the forecast stack — no Snowflake gymnastics required.

5. Tooling The Reformulated Payback

The analytics stack that makes outcome-pricing CAC payback computable in real time:

The Stripe / Metronome / Orb layer is the load-bearing addition for 2027: outcome contracts produce per-event invoices that legacy ARR systems cannot reconcile, and Finance needs the metering data joined to Salesforce opportunities to compute T12M-based payback correctly.

6. Benchmarks To Hold Yourself To

• Bessemer Cloud 100: 12-18 months healthy CAC payback at Series B, <12 at scale.
• OpenView Consumption Sales Index: PLG payback <9 months, enterprise <24 months.
• Snowflake / Datadog / Stripe public filings: 125-135% NRR, <15 month payback on T12M basis.
• Pavilion 2026 Benchmark Survey: median outcome-pricing payback 17.4 months end-to-end (ramp-inclusive).
• Rule of 40 still applies — but the growth half must be computed on T12M consumption, not booked ACV, or the rule lies.

7. FAQ

7.1 Why is ACV the wrong numerator under outcome pricing?

ACV reflects the contract ceiling, not the realized consumption. Outcome contracts routinely under-consume the cap in year one and over-consume by year three. T12M annualized is the only honest numerator.

7.2 What is a healthy CAC payback in 2027?

12-18 months at Series B, under 12 at scale, per Bessemer Cloud 100. Outcome-pricing companies should publish two numbers: steady-state payback and ramp-inclusive payback.

7.3 How long does the consumption ramp take?

18-24 months to run-rate is the modal answer across Snowflake, Datadog, MongoDB, and Stripe public disclosures. Pilots produce 10-25% of run-rate revenue in months 0-3.

7.4 How do I compute AI-attributed CAC?

Stamp every opportunity with a source_agent_id in Salesforce, then divide (agent platform spend + AI Agent Ops headcount + inference cost) by agent-sourced closed-won ARR. Clari, Gong, Salesloft, and Outreach ship native AI-attribution columns as of 2026.

7.5 Should I report a single blended CAC payback to the board?

No. Report the four-motion split — PLG / inbound AE / outbound AE / enterprise — plus the AI-attributed cut. A blended number hides which motion is actually paying back and which is destroying cash.

7.6 What new tools do I need for outcome-pricing CAC math?

A metering and billing layer — Stripe Billing, Metronome, or Orb — joined to Salesforce opportunities, with Snowflake + dbt + Hightouch + Looker computing T12M and contribution margin downstream. Without metering, T12M and outcome-derived margin are uncomputable.

How Consumption Ramp Curves Change Payback Calculation

Outcome pricing introduces a consumption ramp curve that fundamentally alters payback timing. Unlike fixed subscriptions where revenue starts month one, outcome-based models often see customers begin with low initial consumption (testing phases) that grows over 6–18 months to steady-state usage. For example, a customer paying per AI resolution might consume only 50 resolutions in month one, growing to 500 by month six. This means the effective CAC payback period must model the cumulative revenue trajectory, not just divide CAC by a static monthly figure. To account for this, apply a ramp factor — typically 0.3–0.5x of eventual run-rate in the first quarter, scaling to 0.8–1.0x by months 9–12. Use trailing-3-month average revenue (T3M) as a proxy until you have 12 months of data. Companies like Snowflake and Databricks have shown that ignoring this ramp can overstate payback by 40–60% in early periods, leading to premature scaling decisions.

Segmentation by Pricing Motion for Accurate Payback

Not all outcome-priced customers behave the same, so segment CAC payback by pricing motion to avoid misleading averages. The three primary motions in 2027 are: (1) per-resolution/per-action (e.g., Intercom Fin, Salesforce Agentforce) — revenue is highly variable but predictable once usage patterns stabilize; (2) per-accuracy/per-outcome (e.g., Aviso, Gong) — revenue tied to measurable business results, often with higher margins but longer sales cycles; and (3) per-token/per-compute (e.g., OpenAI, Anthropic) — consumption can spike or dip dramatically based on customer workflows. For each motion, calculate separate CAC payback using motion-specific trailing revenue and contribution margins. A healthy payback for per-resolution models might be 9–15 months due to faster ramp, while per-accuracy models may require 15–24 months because of longer proof-of-value periods. Bessemer’s 2027 benchmarks suggest that companies with multiple pricing motions should report a blended payback only alongside motion-level breakdowns to investors.

Leading-Indicator Payback for Pre-Ramp Customers

For customers in their first 6–12 months under outcome pricing, trailing revenue data is insufficient. Use leading-indicator payback — a predictive model based on activation milestones and expansion proxies. Key leading indicators include: time-to-first-outcome (e.g., first AI resolution delivered), activation rate (percentage of seats/users consuming within 30 days), and early consumption velocity (growth rate of monthly outcome volume). These feed into a projected T12M revenue formula: multiply month-3 consumption by a ramp factor (typically 2.5–4.0x based on cohort data) and annualize. For example, if a customer consumes $500 in month three and your cohort ramp factor is 3.0x, projected T12M = $500 × 3.0 × 12 = $18,000. Divide CAC by this projected monthly average ($1,500) to get a leading payback estimate. Update this monthly as real data replaces projections. This approach lets you flag payback risks 6–9 months earlier than waiting for actual T12M data, critical for board reporting in 2027 where outcome pricing is the norm.

FAQ

What is the biggest challenge with measuring CAC payback under outcome pricing? The core difficulty is that revenue becomes variable, lagged, and consumption-shaped instead of a fixed recurring subscription. You can’t simply divide acquisition cost by a stable ACV and gross margin—you must wait for trailing-12-month data to emerge, which can take 18–24 months for enterprise customers.

How do you handle the consumption ramp in the payback calculation? You add a consumption-ramp curve to the model, acknowledging that customers often take 18–24 months to reach run-rate usage (similar to Snowflake’s observed patterns). Until then, you use leading indicators like activation milestones and early expansion proxies to estimate payback trajectory.

Do you still use gross margin in the formula? Yes, but you replace traditional gross margin with outcome-derived contribution margin. This nets the direct cost of every billed outcome—such as compute, API calls, or agent actions—so you’re only measuring the profit from the variable revenue stream.

Should CAC be calculated differently for different customer types? Absolutely. You split CAC by motion: PLG self-serve, inbound AE, outbound AE, and enterprise. Each has a distinct cost structure and ramp timeline, so a single blended payback number can be misleading. Bessemer’s Cloud 100 benchmarks healthy payback at 12–18 months for Series B and under 12 months at scale.

What is AI-attributed CAC and why does it matter? AI-attributed CAC is the share of acquisition cost sourced by autonomous agents (like 11x, Artisan, or Clay). As these tools handle more prospecting and qualification, you need to isolate their cost and efficiency separately from human-led motions to accurately measure payback in 2027.

Can you use leading indicators before you have 12 months of data? Yes. You introduce a leading-indicator payback built from activation signals and early expansion proxies—such as first outcome delivered, time to first repeat usage, or initial upsell triggers. This gives you a directional view until T12M data is available, avoiding a blind wait.

Bottom Line

Outcome pricing did not kill CAC payback — it killed the lazy version of CAC payback. Swap ACV for T12M, swap gross margin for outcome-contribution margin, model the 18-24 month ramp explicitly, split CAC by four motions, and add AI-attributed CAC as the metric the 2027 board actually cares about. Companies that publish both steady-state and ramp-inclusive payback raise their next round on the front foot.

Related on PULSE

• [How should Snowflake price Cortex agents — per query or per outcome?](/knowledge/q1578)
• [How do you calculate CAC payback period correctly for a hybrid PLG-plus-sales motion in 2027?](/knowledge/q16185)
• [Should I Hire a Fractional CRO If My CAC Payback Exceeds 24 Months?](/knowledge/q15886)
• [How do you calculate the CAC payback period in 2027?](/knowledge/q12869)
• [What is CAC payback period and what is a healthy benchmark in 2027?](/knowledge/q12693)
• [What is CAC payback — and how do you calculate it?](/knowledge/q10814)

Sources

• Bessemer Venture Partners — Cloud 100 Benchmarks Report 2026
• OpenView Partners — 2026 Consumption Sales Index and SaaS Benchmarks
• Snowflake — Q4 FY2026 10-K and consumption-revenue commentary
• Datadog — FY2026 investor disclosures on customer ramp
• Stripe — 2026 outcome and usage-billing benchmarks
• Intercom — Fin AI Agent outcomes and per-resolution pricing
• Salesforce — Agentforce pricing models and FY2026 results
• Pavilion — 2026 Outcome Pricing Benchmark Survey
• Tomasz Tunguz — AI-Attributed CAC and the Agent Advantage Line (2026)
• Mostly Metrics — How Snowflake Forecasts Consumption-Based Revenue
• Gartner — 2026 Forecast: Outcome-Based Pricing in Enterprise SaaS

People also search for: measure cac payback under outcome pricing · how to measure cac payback under outcome pricing · measure cac payback under outcome pricing guide