AI Asset Score · Methodology v1.0

One methodology. One score. Applied the same way every time.

Every asset we evaluate gets two outputs: an AI Asset Score (0–100 across 5 fixed dimensions) and a Deal Memo (10-section structured write-up covering Thesis, Traction, Risks, Model Dependency, and Handoff SOP). No seller overrides. Same rubric on every asset.

5Dimensions · fixed rubric every time
0–100Score scale · Toy → Approved Candidate
0%Sale commission · we don't touch the deal

Two layers, one asset

Every asset that clears the initial filter gets two deliverables. Not one, not three. Two.

Layer 1 — Numeric

AI Asset Score · 0–100

Five fixed dimensions, fixed weights. A number a buyer can compare across assets and a seller can defend. Version-stamped: every score carries the methodology version used. This is the summary.

Layer 2 — Narrative

Deal Memo · 10 sections

A structured evaluation memo — Thesis, Market, Traction, Team, Moat, Risks & Mitigations, Model Dependency, Deal Terms, Handoff SOP, Post-Deal Monitoring. This is the story behind the number.

The number alone hides too much. The memo alone doesn't scale. We ship both, always paired, timestamped together.

AI Asset Score v1.0 — 5 dimensions

Five fixed dimensions, fixed weights. Total = 0–100. Every score is version-stamped so buyers know exactly which rubric was applied. Nothing subjective — every point is tied to concrete, verifiable evidence.

1. Traction

Max 25

Evidence of real demand and user adoption — not the creator's opinion. Active users, organic traffic, retention curves, payment history.

  • 25/25: 500+ MAU with organic growth, verified analytics, 3+ months of consistent usage.
  • 12/25: Small but engaged early base (50–100 users), some organic signals.
  • 3/25: Just the creator and their beta group, no independent demand evidence.

2. Revenue

Max 25

Quality and durability of current revenue. Recurring beats one-time. Verified screenshots (Stripe, Gumroad, LemonSqueezy) beat self-reported numbers.

  • 25/25: $500+ MRR with 3+ months of receipts, low churn, diversified customer base.
  • 12/25: Intermittent revenue or strong intent-to-pay signals without consistent receipts.
  • 2/25: No revenue, no verifiable monetization path.

3. Transferability

Max 20

Can a new owner run this without the creator on call? SOPs, credentials vault, no personal account lock-in, documented edge cases. This is where most AI assets fail.

  • 20/20: Full SOP + Loom walkthrough + credentials vault + zero personal account dependencies.
  • 10/20: Basic docs exist; buyer needs creator available for the first few weeks.
  • 2/20: Only works because the founder tweaks it manually — non-transferable in practice.

4. Automation

Max 20

Level of operating leverage. How much does the asset run without active human input? Higher automation = lower ongoing cost for the acquirer.

  • 20/20: Fully automated: payments, delivery, support all handled without daily intervention.
  • 10/20: Some manual tasks remain (weekly maintenance, occasional support).
  • 2/20: Requires daily operator effort to function — effectively a freelance contract, not an asset.

5. Risk

Max 10

Deductive — we score what could impair or kill this asset within 12–24 months. Lower risk = higher score. Platform lock-in, model API dependency, single-customer concentration, legal grey zones.

  • 10/10: Diversified revenue, own infrastructure, no platform risk, no compliance concerns.
  • 5/10: Depends on one AI model API — technically swappable but adds transition risk.
  • 0/10: One customer represents 60%+ of revenue, or active grey-area TOS situation.

Score bands

Total score maps to a public band. This is what buyers see next to the listing.

0–29 · Toy
30–49 · Demo
50–69 · Potential Asset
70–84 · Asset Candidate
85–100 · Approved Candidate

Toy / Demo (below 50): Not sellable as-is. We tell you exactly what would need to change. Potential Asset (50–69): Fixable. Usually a documentation / proof gap. Asset Candidate (70–84): Ready to list with minor packaging. Approved (85+): Featured in the next Weekly Drop.

The Full Memo — 10 sections

Every accepted asset gets a 1-page memo built the way venture firms write theirs. Buyers get to read the same document the IC voted on.

1. Thesis

Two sentences: why this asset is worth someone's time and money. If we can't write it in two, we reject.

2. Market

Realistic addressable audience — not TAM slides. Where the buyers actually are and how they're reachable today.

3. Product

What was built, in one paragraph and one screenshot. Stack, dependencies, hosted-vs-owned components.

4. Traction

Verified revenue (Stripe / Gumroad screenshots), verified usage (analytics), retention shape. Nothing self-reported without proof.

5. Team fit

Who built it, why they're selling, and whether the buyer needs any of them post-sale. Founder-dependent assets get flagged.

6. Moat & Model Dependency

What stops a competitor cloning this in a weekend. Foundation-model exposure gets its own subsection — see below.

7. Risks & Mitigations

Three to five candid failure modes with likelihood and mitigation. The section VC memos always have and marketplace listings never do.

8. Deal Terms

Two valuations (operator / strategic), the delta explained, and what price would move the deal from Pass to Buy.

9. Handoff SOP

Concrete transfer checklist: domains, credentials, analytics, payments, API keys, docs. Empire-Flippers-style migration plan promised as a deliverable.

10. Post-Deal Monitoring

Optional 90-day check-ins on the same metrics that drove the grade. Because "sold and gone" is how brokerages lose the next customer.

Model Dependency — the AI-specific risk layer

The one section that separates an AI-asset memo from a generic SaaS memo. Every listing gets scored on four sub-factors inside Risk.

Sub-factorBest caseWorst case
Provider concentrationModel-agnostic — swappable between 3+ providers with a config changeHard-coded to one API, prompt-tuned to that model's quirks
Switch cost< 1 day of engineering + re-benchmarkWeeks of prompt re-engineering + regression on user-facing quality
Policy exposureUse case allowed under all major providers' TOSGrey-area use case (scraping, PII, generation of restricted content) at risk of overnight ban
Feature absorption riskValue is in distribution / data / workflow — not the model callValue is a thin wrapper around a capability the foundation model could ship natively next quarter

A worst-case profile across all four caps operator value at the low end of the 1–2× band regardless of current revenue. Two absorptions of adjacent wrapper categories have already happened in the last 18 months — we price that in.

Evaluation Committee — how a grade actually happens

No single-reviewer grades. Every published AI Asset Score + Memo has been through at least two sets of eyes over 1–4 weeks.

  1. Intake (day 0). Seller submits via the submission form. An automated AI Asset Score runs; obvious no-go flags trigger a same-day rejection with reasons.
  2. Evidence request (day 1–3). If the asset clears intake, we ask for the concrete proof set — Stripe / Gumroad screenshots, analytics access, API-billing statements, transfer readiness notes.
  3. Draft memo (day 3–10). Lead reviewer writes the 10-section memo end-to-end. Numbers get double-entered from primary sources, not seller assertions.
  4. IC review (day 10–20). Second reviewer challenges the draft on Risks, Model Dependency, and Deal Terms. Disagreements are logged in the memo, not smoothed over.
  5. Publish or reject (day 14–28). Approved memos go into the Weekly Drop rotation. Rejected memos get a private write-up back to the seller with the exact gap to close.

Most submissions do not reach the Deal Memo stage. The bar is intentional — it keeps the signal clean for buyers.

Why our grade is honest

Six structural guarantees. Not marketing — actual mechanisms.

🚫

We take zero commission on sales

Traditional marketplaces earn 5–15% of the transaction. That means high scores = more revenue for them, which distorts the grade. We don't touch the sale — we charge for packaging services separately.

📊

Same rubric every time — no seller overrides

The 5 dimensions are fixed. Weights are fixed. Even if the seller pays for the Packaging Sprint, the final score is still calculated from the same rubric — the Sprint just helps them earn more points legitimately.

🔍

Every claim must be verifiable

Revenue = Stripe screenshot. Users = analytics dashboard. Testimonials = named contact. If we can't verify it, it doesn't count toward the score.

📉

Grades expire in 90 days

An AI asset scored 82 in January can drop to 65 by April if the market moved or usage fell. Grades are timestamped and re-scored on request. Buyers see the freshness.

📢

Failing grades are published (with consent)

We keep a public log of scored projects that failed to sell — and why. Sellers can opt out, but most agree because it's how the market learns.

🔁

Buyer feedback loops into the rubric

Buyers report post-purchase: was the transfer smooth? Did revenue hold up? These signals adjust the weights on Transferability and Proof over time.

Compared to other marketplaces

Not a smear — just where we differ structurally.

AIAsset.MarketFlippaMicroAcquire
Takes commission on saleNo10%Optional 4%
Public rubric with weightsYesNoNo
Score expires / requires re-audit90 daysNoNo
Verifies revenue claimsYes (required)PartialSelf-report
AI-native asset focusOnly AIEverythingMostly SaaS
Free grade availableYesNoNo
High-bar curationYesNoNo
Full VC-style memo per listingYes — 10 sectionsListing form onlyListing form only
Evaluation Committee reviewYesSingle seller-facing agentAutomated + optional broker

Why 12–24 month payback (not 3–5 years like Flippa)

Every valuation multiple on this site assumes a 1–2× annual profit payback. Here's the reasoning.

Traditional business marketplaces (Flippa, MicroAcquire, Empire Flippers) use 3–5× annual profit as the default. That's fine for a boring SaaS or an Amazon FBA — the market moves slowly, moats hold, and a buyer can plan 4–5 years ahead. For an AI asset, that assumption breaks in ways that are already visible today:

Our multiplier by asset type:

Asset typePaybackMultiple (annual profit)
AI wrapper (GPT/Claude API)12–18 months1.0 – 1.5×
AI SaaS with proprietary data or brand18–24 months1.5 – 2×
AI with distribution + low platform riskup to 30 monthsup to 2.5×
Classic SaaS with corporate contracts3–5 years3–5× (not AI, comparison only)

If a seller quotes a Flippa-style 4–5× multiple on an AI asset, we mark it as overpriced and recommend Pass — or renegotiate to 1.5–2× before proceeding.

Calibration case #1 — Flippa listing #12103259

Real Flippa listing tested through our AI Asset Score — August 2026.

The listing:

  • SaaS · Automotive · Netherlands · 4 years old
  • Monthly profit: €1,103 · Annual revenue ≈ €666k
  • 4 active subscribers · 0% churn
  • 1 major client contract: €2M over 3 years (€1.33M backlog remaining)
  • IP licenses: €5M in gaming / merchandising
  • Inventory included in sale: $948k
  • Asking price: $3,195,155

Our AI Asset Score output:

  • Score: 74 / 100 · Asset Candidate
  • Operator value (buy-and-run, 1–2× annual profit including revenue diversification): $144k – $288k
  • Strategic value (with contract backlog at 60–80% + IP at 20–40%): $1.33M – $2.0M
  • One-liner: "Strong contract & IP, but low user base"

Our verdict:

Asking $3.2M is 60–140% above our strategic-value ceiling and 11–22× above our operator-value ceiling. Buy only if seller drops to $2M for a strategic play, or $288k for operator run. Otherwise Pass.

This is exactly the honesty gap traditional marketplaces avoid — because they earn commission on the sale.

We'll publish a running Calibration Report as more listings are scored. If you have a specific listing you'd like tested — try the AI Asset Score widget.

Have an asset? Or looking for one?

Sellers — try the instant AI Asset Score widget or submit for a full Deal Memo evaluation. Buyers — tell us what you're looking for and we'll match you with relevant assets.

Grade my asset → I'm looking to buy →