What inputs does the AI Feature Attribution need?

It takes 6 inputs: totalArr, totalUsers, aiSpecificUsersFraction, aiUsersArpuUpliftPercent, aiInfraMonthlyCost, churnDifferenceWithAi. Outputs returned: arrAttributableToAi, netAiAttributableArr, grossMarginPercentAiCohort.

Can I verify the AI Feature Attribution with a worked example?

Yes. With $2M ARR, 5000 users, 30% on AI, 25% uplift, $8k/mo infra, −2% churn. the tool returns baseArpu ≈ $372, attributable ≈ $139k, netAttributable ≈ $43k after $96k infra.

Where does the AI Feature Attribution get its benchmark data?

Reference data is sourced from: Bessemer State of the Cloud (as of 2024); Open Source SaaS Benchmarks (as of 2024).

What can the AI Feature Attribution not tell me?

Known limitations: Doesn't model self-cannibalization (AI users may have come from the same base). Retention lift is reported as a percentage, not converted to dollars — pair with the CLV calculator.

Methodology: AI Feature Attribution

1. Scope

Estimates the ARR uplift attributable to AI features given total ARR, AI cohort size, ARPU uplift, infra cost, and retention delta.

2. Inputs and outputs

Inputs

totalArr number ($)
totalUsers number
aiSpecificUsersFraction number (0–1)
aiUsersArpuUpliftPercent number
aiInfraMonthlyCost number ($)
churnDifferenceWithAi number (-1..1)

Outputs

arrAttributableToAi

(ai_arpu − base_arpu) × ai_users.
netAiAttributableArr

Attributable minus annualized AI infra cost.
grossMarginPercentAiCohort

Cohort revenue × 80% baseline minus AI infra, ÷ revenue.

Engine source: src/lib/ai-feature-attribution/engine.ts

3. Formula / scoring logic

ai_users = total_users × ai_fraction
base_arpu = total_arr / (non_ai_users + (1 + uplift) × ai_users)
ai_arpu = base_arpu × (1 + uplift)
attributable = (ai_arpu − base_arpu) × ai_users

4. Assumptions

80% gross margin baseline on non-AI revenue (typical SaaS blended).
All AI infra cost is allocated to the AI cohort, not amortized to non-users.
ARPU uplift is the only difference between cohorts other than infra cost.

5. Data sources

Bessemer State of the Cloud as of 2024
Open Source SaaS Benchmarks as of 2024

6. Known limitations

Doesn't model self-cannibalization (AI users may have come from the same base).
Retention lift is reported as a percentage, not converted to dollars — pair with the CLV calculator.

7. Reproducibility

Input
$2M ARR, 5000 users, 30% on AI, 25% uplift, $8k/mo infra, −2% churn.

Expected output
baseArpu ≈ $372, attributable ≈ $139k, netAttributable ≈ $43k after $96k infra.

8. Change log

2026-05-08 methodology first published.

Worked example

Run live against the same engine this site ships (/engines/ai-feature-attribution.js). The inputs and outputs below are recomputed on every build and independently re-verified in CI — they are never hand-authored.

Input

tool: ai_feature_attribution
total_arr: 2000000
ai_specific_users_pct: 30
ai_users_arpu_uplift_percent: 25
ai_infra_monthly_cost: 8000
churn_difference_pct: -2
total_users: 5000

Output

aiUsers: 1500
nonAiUsers: 3500
baseArpu: 372.09
aiArpu: 465.12
arrAttributableToAi: 139534.88
annualAiInfraCost: 96000
netAiAttributableArr: 43534.88
grossMarginPercentAiCohort: 66.24
grossMarginPercentNonAiCohort: 80
effectiveChurnReductionPercent: 2

Frequently asked questions

What does the AI Feature Attribution calculate?: Estimates the ARR uplift attributable to AI features given total ARR, AI cohort size, ARPU uplift, infra cost, and retention delta.
What inputs does the AI Feature Attribution need?: It takes 6 inputs: totalArr, totalUsers, aiSpecificUsersFraction, aiUsersArpuUpliftPercent, aiInfraMonthlyCost, churnDifferenceWithAi. Outputs returned: arrAttributableToAi, netAiAttributableArr, grossMarginPercentAiCohort.
What formula does the AI Feature Attribution use?: The exact computation is: ai_users = total_users × ai_fraction; base_arpu = total_arr / (non_ai_users + (1 + uplift) × ai_users); ai_arpu = base_arpu × (1 + uplift); attributable = (ai_arpu − base_arpu) × ai_users
Can I verify the AI Feature Attribution with a worked example?: Yes. With $2M ARR, 5000 users, 30% on AI, 25% uplift, $8k/mo infra, −2% churn. the tool returns baseArpu ≈ $372, attributable ≈ $139k, netAttributable ≈ $43k after $96k infra.
Where does the AI Feature Attribution get its benchmark data?: Reference data is sourced from: Bessemer State of the Cloud (as of 2024); Open Source SaaS Benchmarks (as of 2024).
What can the AI Feature Attribution not tell me?: Known limitations: Doesn't model self-cannibalization (AI users may have come from the same base). Retention lift is reported as a percentage, not converted to dollars — pair with the CLV calculator.