Free · both sides Free for buyers · Free for sellers · Funded by neither · Launching Q3 2026
Methodology v3.1 · Charter ratified · Inquiries open
Free for buyers · Free for sellers · Funded by neither

Grade every audience before you buy it. A through F. Free for both sides.

Until now, every audience verifier got paid by the audience side. VRA grades any audience — A · B · C · D · F — against a public benchmark, with zero revenue from the suppliers we rate. Score it before you commit a dollar. Get rated without writing a check. Independence in the funding model, not the marketing.

$0
If you're aBuyer or agency
$0
If you're anAudience seller
★   Verified Reachable Audience   ★
A+
VRA Grade Certified
Specimen · Methodology v3.1
Free to earn  ·  Free to display
$0
Always
Both sides
What's free

Buyers don't pay. Sellers don't pay. The two sides on the trade pay nothing.

There's only one way to be a credible third-party rater: don't take money from the people you rate, and don't make the buyers who depend on the rating pay either. We do neither. The registry, the API, the audit client, and the badge are free for both sides of the buy — written into the charter, not just the marketing.

— For buyers & agencies $0

Score the audiences you're about to buy.

Look up any rated seller or segment. Score your own audience without sharing IPs. Wire the grade into your DSP or RFP automation. We don't charge per query, per seat, or per workspace.

  • Public registry — every assessed seller and segment, current grade, 90-day trend
  • Scoring API — JSON in, JSON out, full v3.1 payload, no field gating
  • Private audit client — grade your own audience without sending us the IPs (Apache 2.0)
  • Downgrade webhooks — signed POST within 30 sec of any grade change
  • Methodology & replication kit — recompute any grade yourself
Get buyer access
— For audience sellers $0

Get rated. Display the mark. Pay nothing.

Submit a sample, we score it under v3.1, and the grade is yours — A through F. No payment, no premium grade, no expedited tier. The mark is licensed at zero cost while the rating holds. Sample is destroyed after scoring.

  • Free assessment — match-file in, grade out, no integration cost
  • Free badge license — use the VRA Certified mark in decks, RFPs, and marketing
  • Free re-rating — quarterly reissues + on-demand re-review after fixes
  • Sample destruction guaranteed — written into the seller agreement
  • Right to decline — non-cooperating vendors are still listed for transparency
Get rated

So who pays for it?

DSPs, ad-tech infrastructure, and platforms that license the grade for bid-time integrations or seller qualification — paying on behalf of the buy-side, not the seller.

Industry research firms, hedge-fund equity research, and regulators who consume the data feed for diligence and oversight. None of them are rated by us. None of them get a thumb on the scale.

The Grade

Residential. Reachable. Right. The three tests a real audience passes.

Before any letter is computed, every audience under v3.1 has to clear three plain-English bars. The grade just measures how cleanly it does. If the answer to all three is yes, you're in A/B territory. The further any one drifts, the further the letter falls.

R.

Residential.

The audience lives on real consumer connections — home ISPs, household Wi-Fi, fixed broadband. Not datacenter-routed, not hosting-routed, not a server farm wearing residential clothes.

Fails when More than a thread of traffic resolves to commercial IP space, hosting, or proxy infrastructure.
R.

Reachable.

Each profile resolves to one addressable identity — a single household, device, or person you can actually serve. Not relayed, not anonymized, and not a thousand mobile users sharing one carrier-grade NAT egress IP.

Fails when Traffic comes through CGNAT (most mobile carriers), iCloud Private Relay, VPNs, anonymizing proxies, or recycled identifiers — anything that collapses many users into one apparent address.
R.

Right.

The audience actually matches what's claimed. If the segment says "auto intenders 25–54," it has to look like auto intenders 25–54 against the benchmark — not just any household with a pulse.

Fails when Composition diverges materially from the segment's own description, measured in σ from a paid-tier ground-truth dataset.

One letter. One central signal. Four supporting checks.

Behavior — how many distinct identities use an IP in a day — is the one thing a residential proxy can't fake by buying real ISP credentials. IP fanout sits at the center of every grade. Four supporting signals reinforce or contextualize what fanout already saw. All five are published with weights and reproducible from raw data.

A
B
C
D
F

A and B are certified. Audiences earning A or B carry the VRA Certified mark and pass independent benchmark validation.

C is acceptable. Real audience, measurable signal, but elevated noise on at least one input.

D and F fail validation. Composition diverges materially from what's claimed. Listed publicly so they're visible, not hidden.

i.

IP fanout (md5)

Distinct identity fingerprints observed per IP per day. A real household: 1–3. A polluted IP: 20+. Behavioral — can't be defeated by buying real attributes.

40%
★ Primary
ii.

Non-residential share

Fraction sourced from datacenter, hosting, or commercial IP space.

20%
iii.

Privacy-relay share

Fraction routed through iCloud Private Relay, VPNs, or anonymizing proxies.

15%
iv.

Freshness

Days since the underlying signal was observed. Stale audiences degrade silently.

15%
v.

Geographic precision

Pin-radius accuracy of the IP geolocation, in miles. Catches CGNAT and aggregation-point residue.

10%

Why fanout sits at the center of every grade.

Static attributes of an IP — its ISP, its geolocation, its connection type — can all be bought. Residential-proxy services rent real subscriber connections from real ISPs in real homes, and every attribute filter passes them with high marks. Behavior is the one thing they can't buy. That's the job fanout does.

i.

Static attributes can be bought.

A residential proxy running on a Spectrum subscriber's cable line in Toledo has a residential ISP block, accurate geolocation, and a real consumer subscriber pattern. Every attribute filter passes. The connection IS legitimate — it's just being rented out.

ii.

Behavior can't be bought.

A normal household runs 1–3 distinct identity fingerprints per day. A proxy IP routing scraping sessions, AI crawlers, ad-verification jobs, and tunneled streaming users runs 20, 50, sometimes hundreds. The attributes lie. The usage pattern can't.

iii.

Different question. Different answer.

Other verifiers ask "what is this IP?" and accept the honest answer. Fanout asks "how is this IP being used today?" — and that question doesn't have a way to lie. One can be defeated. The other can't.

★ Why it's the headline

Most audience verifiers grade what an IP is. Residential-proxy services figured that out years ago and bought their way through every attribute check. Fanout grades what an IP does — the one signal that's invisible until you actually run the math. It's why a clean Toledo cable line can score A and a polluted one can score F, even when they look identical on paper.

Why do you discount mobile traffic? Three reasons that stack.

Because mobile fails all three R's at the same time — and it fails them by how the network is built, not because of anything we filter out. Each reason below is enough on its own to disqualify mobile as a household anchor. Together they make it structurally impossible.

i.

One IP, thousands of people.

Verizon, AT&T, T-Mobile, and every other major carrier route thousands of phones through a small pool of shared public addresses ("CGNAT"). The IP we see in a mobile bid request isn't your phone — it's a carrier gateway shared with hundreds-to-thousands of other subscribers in the same region. Whose phone is it? You can't tell.

ii.

Mobile IPs don't say where someone is.

A residential IP pins to a specific cable line at a specific address. A mobile IP pins to a cell tower with a 1–15 mile radius — or worse, a regional aggregation point 20+ miles away. Useful for metro-level. Useless for households.

iii.

Mobile IPs don't last.

A home IP usually stays with a household for weeks or months. A mobile IP can cycle through a dozen addresses in a single afternoon — every tower handoff, every reconnect, every load-balance. The IP that's "you" at 2pm belongs to someone else by 4pm.

★ The honest part

We don't throw away mobile users. We throw away mobile as an identifier. The same person, when they get home and join their Wi-Fi, becomes a residential observation in the cohort. The mobile-to-Wi-Fi handoff is actually the most valuable moment in the building — that's where the cross-device graph gets its strongest evidence. We're patient about waiting for the user to come home before we anchor them.

Methodology v3.1 — published, versioned, free to replicate.

64 pagesCC BY-SA 4.099.6% precision · 97.4% recallThe full pipeline, signal definitions, weights, benchmark protocol, and dispute process. Open for replication and challenge.

Request the working draft
Inquire · always free

Tell us which side you're on.

Whether you're scoring audiences or being scored, the answer is the same: it costs nothing. Send the form and we'll route it to the right person — reply within one business day.

Buyers & agencies

API access, methodology working draft, and a charter API key during launch beta — all free.

Audience sellers

Free assessment, free badge license, free re-rating. On the day-one registry if submitted in the cohort window.

DSPs, press, researchers

Volume terms, embargoed briefings, and methodology access on request. Tell us what you need.

What gets you the grade.

Reply < 1 biz day
We reply within one business day, from a real human.
Thank you. Your inquiry is in. We'll reply within one business day from a real human, with the launch brief and any documents matching what you selected.