ICPHow it works

Find the right people. Know what to say.

Describe your buyer in a sentence, or let Proven ICP identify your Ideal Customer Profile (ICP) from your website. Get real people ranked by fit, with cited reasons and context for a more relevant first conversation.

  1. 01Describe a buyer, or analyze your website
  2. 02Get ranked prospects with cited fit reasons
  3. 03Use their context for a better first conversation

The free preview is an illustrative sample. A free account includes 1 real search.

Revenue operations leaders at freight companies replacing their routing software

4 matches

Tomas Reyes

Director, Sales Ops · Corridor Logistics

Fit score89

Why they fit · public evidence

  1. [1]

    Posted an RFP summary for route optimization vendors.

    corridor.example/vendors

  2. [2]

    Team page lists a new revenue systems function formed this year.

    corridor.example/team

Suggested opener

Noticed Corridor opened a routing RFP. Happy to share how similar fleets evaluated vendors.

Illustrative example, not a live search. People and sources are fictional. Always confirm details before outreach.

The process

Three steps from brief to better conversations.

One sentence in, a ranked and cited shortlist out. Every step runs live against the public web.
  1. 01Define who you want

    Describe a buyer, or start from your website.

    Write who you need the way you would tell a colleague. Not sure yet? After sign up, paste your URL and Proven ICP drafts your ICP for you to review and edit. A follow up question or two sharpens it.

    SampleDescribe your customer, or analyze my website
  2. 02Get a ranked shortlist

    Real people, ranked by fit, with the reasons cited.

    Proven ICP reads the live public web, not a stale contact database. Every match shows why it fits and the public source behind each reason, so you can check it yourself.

    SampleAnatomy of one match, fictional sample
  3. 03Start the conversation

    Open with something that matters to them.

    Use what each person has written, said, or shared to make your first message relevant. Paid plans add available contact details and a drafted opener for every match.

    SampleSuggested opener, fictional sample
The full five step process, in detail
  1. 01

    Plain English description

    You describe who you are looking for in natural language. No filters, dropdowns, or field mapping. Examples: "Early-stage SaaS founders who write publicly about product-led growth and have raised a seed round in the last 18 months." Or: "Heads of RevOps at Series A B2B companies who post about their sales stack." The only input is a sentence or two about the person you need.

  2. 02

    ICP refinement through follow-up questions

    Proven ICP asks a small number of intelligent follow-up questions to sharpen your Ideal Customer Profile. These questions target the dimensions that matter most for your search: geography, company stage, specific signals of intent, and any constraints that would disqualify an otherwise strong match. Sharpening the ICP at this step meaningfully improves the precision of what comes back.

  3. 03

    Live public web search

    Proven ICP searches the public web in real time using neural search. It does not query a static contact database. Results reflect current public information: what people have written this year, where they have spoken recently, what companies they are at today. Three parallel search strategies run simultaneously to maximize coverage across different signal types.

  4. 04

    Multi-layer signal analysis

    For each candidate found, Proven ICP reads and analyzes the actual page text from their public presence. It extracts writing, talks, affiliations, credibility signals, and behavioral indicators. It then scores each person on how well those signals match the ICP, with a specific cited reason explaining the match. Candidates who do not meet a minimum quality threshold are filtered out before the results are returned.

  5. 05

    Ranked shortlist with cited reasons

    Proven ICP returns approximately 25 matches ranked by fit. Each match includes name, role, company, location, a numeric match score, and a cited reason that explains specifically why this person fits the description. The reason references actual public signal: what they wrote, where they spoke, what their company does. Paid plans also include full contact details and an AI-drafted outreach message personalized to each match.

Signals

What Proven ICP reads to judge fit.

Proven ICP analyzes 8 categories of public signal for each candidate. The categories analyzed depend on search mode: B2B searches focus on professional signals, B2C searches extend to personal profile and lifestyle signals.

  • [1]Public writing
  • [2]Talks and podcasts
  • [3]Open source work
  • [4]Affiliations and boards
  • [5]Media coverage
  • [6]Peer recognition
  • [7]Financial capacity
  • [8]Personal profile (B2C)
What each signal category covers
[1]

Public writing

Newsletters, blog posts, LinkedIn long-form articles, personal websites, Substack publications, and Medium essays. Writing is the highest-signal category because it reveals what someone thinks about, cares about, and knows, in their own words.

[2]

Conference talks and podcast appearances

Keynote sessions, panel discussions, podcast episodes, and recorded interviews. Speaking engagements signal credibility, domain authority, and the specific problems a person is publicly associated with.

[3]

Open source contributions and technical projects

GitHub repositories, open source project contributions, technical specifications, and public demos. For technical roles, this category often reveals more about depth and focus than any profile field.

[4]

Professional affiliations and board memberships

Named board roles, advisory positions, investor portfolios, professional association memberships, and working group participation. Affiliations reveal relationships, commitments, and organizational alignment.

[5]

Media coverage and third-party citations

Press mentions in trade and general publications, analyst reports that name an individual, and awards or rankings from credible third parties. Third-party coverage is a strong credibility signal because it represents external validation.

[6]

Peer recognition signals

Who cites, recommends, or refers to this person in their own public writing and talks. Co-authorship and co-speaking patterns. Peer signal is particularly valuable for identifying genuine authority versus self-reported expertise.

[7]

Financial capacity indicators

Publicly available indicators of budget authority, company funding stage, role seniority, and deal history for investors. Proven ICP derives a capacity tier from cited public sources. No paid data sources are used.

[8]

Personal profile signals

Education, homeownership, family, vehicle, pets, lifestyle, hobbies, collections, entertainment preferences, travel, and civic affiliations publicly shared across social profiles and personal sites. Used primarily in consumer-facing searches to build a richer picture of the individual.

Accuracy

Every reason points to a real page.

Every signal extracted by Proven ICP is grounded in a source URL from the actual page where it was found. The system uses three independent enforcement layers to prevent fabricated or inferred data from reaching a lead record.

Cited sourcesNothing guessedYou can verify
[1]

Temperature zero inference

All extraction calls run at temperature zero. This means the AI model reads and reports what it finds rather than completing or embellishing with plausible-sounding content.

[2]

Strict prompt rules

The extraction prompt requires the target person's name and the specific fact to appear on the same page, descriptions to be direct quotes or minimal paraphrases within 20 words, and omission of any fact where there is any name ambiguity.

[3]

Source URL allowlist validation

After the model responds, every signal's source URL is checked against the set of URLs actually fetched in that request. Any URL the model invented, shortened, or modified is dropped before it can reach a lead record. Dropped signals are logged as warnings.

Under the hood

Technical foundation

The short version: live public web, no stored contact list.

Proven ICP is built on large language models for ICP interpretation, signal extraction, match scoring, and reason generation. Web retrieval uses semantic search to find relevant public information.

The pipeline runs multiple parallel search strategies for each query to maximize signal coverage across different parts of the public web. For enrichment, additional parallel searches target specific signal groups (credentials, lifestyle, media presence, financial and community) independently, then merge and deduplicate results before they reach the output.

Proven ICP does not maintain a proprietary contact database. There is no static dataset to go stale. Every search is a live operation, which means the results reflect what is publicly available at the time of the search.

FAQ

Frequently asked questions

No. Proven ICP does not maintain a proprietary contact database. Every search queries the live public web, which means results reflect current public information rather than data that may be months or years out of date.

Try it

See it in action

Open the free preview with no account to see an illustrative ranked shortlist with the reason each person fits. Create a free account to run 1 real search on your own brief.

Try a free search

Find the right people, then go close more deals.

Describe your ideal customer and review each match with its cited public sources.

Free accounts include 1 real search. No credit card to try.