Cohort definition
Only B2B SaaS companies with annual contract value (ACV) between $24K and $480K, selling to a North American buyer, and active on the platform for at least one full quarter before measurement began.
Audit-grade · Reproducible · Vendor-independent
The 38% lift figure your CMO quoted is unfalsifiable — here is the methodology behind ours.
Vendor-reported attribution lift is the most quoted and least examined number in B2B SaaS marketing. Most published figures are calculated by the same team that sold the software, on cohorts the vendor selected post hoc, against controls that never existed. This page is a line-by-line disassembly of how our 38% number was produced, what it does and does not claim, and exactly how to reproduce it against your own funnel.
01 / The Four Chapters
Four chapters. One reproducible number.
Every benchmark we publish is structured the same way. Each chapter can be read independently, but the figure only holds when all four are accepted together. Skip one and the lift becomes a marketing claim instead of a measurement.
-
01
Cohort selection
Which 412 deployments qualified, which 2,148 did not, and the seven inclusion criteria each one had to clear before it counted toward the figure.
-
02
Model comparison
What the nine-signal model is being benchmarked against — last-click, W-shaped, and a Markov chain baseline — and why each was chosen.
-
03
Measurement window
Why 90 days, why post-onboarding only, and why we exclude weeks one and twelve from the headline number.
-
04
Statistical guardrails
Confidence intervals, holdout validation, sensitivity tests, and the four scenarios under which the lift number would be retracted.
02 / The Numbers
The numbers behind the headline.
No narrative, no framing. The four figures below are the ones we agree to defend in front of a RevOps skeptic, a marketing-attributed CFO, or a procurement reviewer who has seen this movie before.
Sample size
412
qualifying deployments, Jan 2022 – Dec 2024
Average lift in marketing-sourced pipeline
38%
95% CI [33.1%, 42.9%], first 90 days post-onboarding
Customer renewal rate
94%
annual contracts, measured at month 12
Events processed in 2024
2.1B
across 41 native connectors
Source: SS9SS Digital Benchmark Study v2024.4 · Independent replication: Forrester TEI Q4 2024
03 / The Six Decisions
Six decisions a finance team can audit line by line.
Every benchmark rests on a handful of choices that are easy to skip and impossible to defend later. Here are ours, with the tradeoff each one resolves.
Control construction
Each treatment deployment was matched to a near-identical control deployment on eleven covariates: ACV band, industry, prior quarter pipeline, sales team size, and seven more — never on the outcome variable.
Nine-signal weighting
Treatment deployments run the full Nine-Signal Attribution Engine. Product-usage events, third-party intent spikes, partner-sourced meetings, and pricing-page revisits are weighted alongside paid clicks — none treated as last-click.
Holdout validation
A randomly assigned 14% of treatment deployments were held out from the headline calculation. Their lift was reported separately and reconciled to the full sample at publication time.
Measurement window
Pipeline lift is measured between day 14 and day 90 post-onboarding. Week one is excluded to remove onboarding noise; day 91+ is excluded to keep seasonality out of the headline.
Reproducibility artifacts
Every input dataset, model weight, and SQL query is published to the SS9SS public trust portal. Any qualified buyer can re-run the calculation in under three hours.
Pillar-by-pillar acceptance criteria
A finance reviewer can scan this in 30 seconds.
| Pillar | Decision | Tradeoff resolved | Rejection threshold |
|---|---|---|---|
| P-01 | ACV band $24K – $480K | Excludes enterprise outliers and self-serve PLG | If ACV band > 2× listed range, retract |
| P-02 | Caliper-matched controls, 11 covariates | Removes outcome-variable bias in selection | If SMD > 0.15 on any covariate, retract |
| P-03 | Nine-signal weighting engine | Replaces last-click with multi-touch truth | If signal input changes, re-baseline required |
| P-04 | 14% stratified holdout | Independent check on the headline | If holdout < 26% lift, re-audit full sample |
| P-05 | Days 14–90, YoY seasonality | Removes onboarding noise and quarter drift | If cohort spills past Q4 2024, re-window |
| P-06 | Public replication artifacts | Lets any RevOps team reproduce locally | If portal access revoked, figure is suspended |
04 / The Skeptic's FAQ
Questions a RevOps skeptic will ask before believing the number.
Every methodology deck skips these. We do not. They are answered in the order a CFO or RevOps lead will raise them, with the receipts attached.
01 — Survivorship bias.Selection
Cohorts only include deployments still active at month 12. Of the original 562 onboarded in 2022–2024, 150 churned before month 12 and were excluded. Their pre-churn lift was, on average, 11% — lower than the 38% headline. If they were re-included with inverse-probability weighting, the headline moves from 38% to 34.2%. We publish both numbers.
02 — Selection effects.Self-selection
Companies that buy attribution software are not random. We control for this by matching treatment and control deployments on industry, ACV band, prior-quarter pipeline, sales team size, and seven other covariates — never on the lift outcome. Standardized mean differences across all eleven covariates were below 0.10, the conventional balance threshold.
03 — Seasonality.Time-of-year
SaaS pipeline is heavily seasonal (Q1 and Q4 are different beasts). Each deployment's lift is measured against the same deployment's prior-year pipeline, quarter-matched, so a Q3 2024 onboarding is benchmarked to Q3 2023 — not to a flat average. The 33.1% to 42.9% CI already absorbs the residual quarterly variance.
04 — Attribution-model drift.Versioning
The nine-signal weighting has shipped six minor revisions since 2022. To keep the headline stable, all 412 deployments are scored against the model version that was live at their onboarding date — not a back-fitted v2024.4. This prevents the lift from inflating simply because the engine got smarter.
05 / Reproduction Offer
Run the same benchmark against your own funnel in 30 minutes.
A senior strategist will re-execute the four-chapter analysis on your funnel data: cohort definition, control construction, nine-signal weighting, and a 90-day windowed measurement. You leave the call with a number your CFO can defend — or the honest reason your funnel cannot be measured credibly yet.
- 30-minute diagnostic, no slide deck.
- Outputs a 1-page methodology brief on your stack.
- Cancel any time — no procurement form to start.