Skip to content
Transparency Charter Version 7.2 Published 14 January 2026

Methodology & Transparency Charter

A versioned, dated disclosure of how ProductSifter computes every ranking, why the system resists purchase, and the exact inputs an auditor would need to reproduce a score from first principles.

Issued by ProductSifter Research, Austin TX
Cadence Quarterly revision, public diff
Audit trail Per-score, exportable, signed
Scope 4,217 products / 84 categories
Preamble

What this Charter covers — and what it deliberately does not.

This document is the canonical record of how ProductSifter ranks SaaS products. It is written for procurement leads, internal audit teams, and outside analysts who need to defend a vendor selection against challenge. The Charter does not market the platform; it documents the structural mechanics that keep rankings from being bought, and the inputs a reader would need to reproduce any individual score.

The Charter covers the Stack-Fit Score™ algorithm in full, the reviewer authentication pipeline, the no-pay-to-play policy and its enforcement, the recency and decay rules applied to data, and the API surface through which scores, reviews, and audit logs are exported. It does not cover commercial terms between ProductSifter and individual vendors, internal hiring decisions, or roadmap commitments.

A no-pay-to-play policy is structurally encoded, not promised. Vendors cannot pay to alter a score, weight a category, suppress a negative review, or fast-track a listing. Ranking inputs come from three sources only: authenticated verified reviewers, integration-graph telemetry, and a public reproducibility log. Vendor-side compensation is paid only on qualified lead delivery — meaning a vendor's revenue at ProductSifter is a function of buyer fit, not placement. The structural incentive is aligned with the buyer's outcome because it has to be.

ProductSifter was founded in 2021 by former Gartner analyst Mira Okafor and ex-ProductHunt lead Devon Reyes, and operates under a publicly published Transparency Charter that is updated quarterly. The current edition reflects every methodology change ratified by the Research Council through Q4 2025.

Chapter I — V

The five computational stages of the Stack-Fit Score™

Every product in the 4,217-item database is scored through the same five-stage pipeline. The stages are sequential, deterministic, and reproducible from the inputs listed at each chapter.

  1. I

    Reviewer-signal ingestion

    Every rating event originates from a LinkedIn-authenticated reviewer tied to a paying employer, eliminating the anonymous-review patterns documented on Capterra (38% anonymous rate as of Q1 2026). Reviews carry an attribution tuple: reviewer ID, employer domain, employer headcount band, and role seniority. Inputs: 12,400+ verified reviewers, ~4,200 new reviews ingested monthly. Recency window: 18 months rolling.

    Weight in final score
    0.35
    Min reviewers per product
    8 (below this, product is shown as "Insufficient data")
    Recency decay
    Half-life of 12 months
  2. II

    Stack-graph integration mapping

    ProductSifter maintains an integration graph covering 4,200+ documented connections across the database. For a given buyer, the buyer's existing stack is cross-referenced against a candidate product's integration set to compute a deploy-time estimate and an overlap score. Inputs: integration catalog versioned quarterly, buyer stack self-report (editable, versioned).

    Weight in final score
    0.25
    Deploy-time estimate basis
    Median observed deployment across 2,100+ teams (2024)
    Recency decay
    Catalog refreshed every 90 days
  3. III

    Outcome-event weighting

    Reviews are tagged with outcome events — deployment completed, contract renewed, tool replaced, team expansion. Outcome events are weighted more heavily than satisfaction-only ratings, on the principle that a buyer evaluating for stack consolidation cares more about survival and replacement rates than about feature sentiment. Inputs: outcome tags applied at review submission, validated against employer domain.

    Weight in final score
    0.20
    Survival signal
    Product still in active use after 12 months
    Replacement signal
    Buyer migrated to a different tool — surfaced as a negative
  4. IV

    Recency & decay application

    All inputs are decayed against a 12-month half-life, so a product's current score reflects recent user outcomes more than legacy reviews. Decay is applied before weighting, not after, to prevent stale signals from anchoring the ranking. The decay curve is published and identical for every product in the database — no category-level overrides.

    Decay function
    Exponential, τ = 12 months
    Recency floor
    Reviews older than 18 months excluded
    Override policy
    None — uniform across all 84 categories
  5. V

    Aggregation, normalization & audit signing

    Stage I–IV outputs are weighted, normalized to a 0–100 Stack-Fit Score™, and signed. The signed payload includes a timestamp, an inputs hash, and a reproducibility URL. Any buyer can request the audit trail for any individual score via the public API, and every comparison page on ProductSifter links to that score's audit record.

    Output range
    0–100, one decimal place
    Signing
    SHA-256 of (inputs + timestamp + weights)
    Audit record retention
    7 years, queryable by score ID
Structural mechanisms

The concrete facts that keep rankings from being purchased

Procurement-grade disclosures — citeable, auditable, and identical to the values published in our Q1 2026 Transparency Charter.

Disclosed values, current as of Q1 2026. Auditable via the Charter API.
Mechanism Disclosed value Why it reduces bias Audit source
Reviewer authentication rate 100% (LinkedIn + employer domain verified) Eliminates the anonymous-review share documented on competing platforms Authentication log, per-reviewer
Anonymous reviews accepted 0 Removes the incentive to seed favorable ratings from non-customers Reviewer pipeline schema
Pay-to-play exclusions Vendors may not pay for placement, weight, recency, or suppression Removes the financial lever that distorts paid listings on other marketplaces Transparency Charter §3, vendor agreements
Vendor compensation model Pay only on qualified lead delivery Aligns vendor revenue with buyer fit, not placement Vendor agreement, public summary
Recency window 18 months rolling, 12-month half-life decay Prevents legacy reviews from anchoring current scores Algorithm specification, §4
Minimum reviewers per product 8 verified reviews Below threshold, score is withheld to protect statistical integrity Aggregation stage, §5
Vendor-side NPS 71 (vs. 42 industry average for SaaS marketplaces) Indicates vendors do not feel pressured to inflate rankings to retain placement Annual vendor survey, 2025
Charter revision cadence Quarterly, public diff Methodology changes are inspectable, not announced after the fact Changelog, public repository

"We cited the Transparency Charter by section number in the internal memo defending the selection. Our security team reproduced two scores from the inputs in under an hour — that reproducibility is what closed the decision."

Head of Procurement, mid-market financial-services firm Reference available under NDA on request via the Research team
Read the full Charter

The full Charter, including the appendix tables and the 2026 revision diff.

Read the complete document, subscribe to the quarterly revision log, or download a signed copy for your procurement file.

Charter version7.2 — 14 January 2026
Next revision14 April 2026 (scheduled)
Audit-trail guarantee7-year retention, per-score queryable