outricher/Visa ICP playbook/Method
The machinery, for the reader who wants it.
Which extraordinary-ability criteria career data can actually see, what one delivered record contains, how fast the signals drift, every defect we found by reading rows rather than counting them — and the lines we hold on your behalf.
Why an ICP beats a list
Three laws
Law 01
The arc, not the address
The signal is never “lives in country X”. It’s a degree earned outside the US next to a job held inside it — the pair that means “here on someone else’s visa”. That pair held across every segment the practice re-ordered; the address changed constantly.
Law 02
Urgency is a date
Graduation month, the date a founder title switched, the number of years someone has actually been in the country — all of these are fields with dates in them. That is what lets a letter land in the month a person is already looking for counsel, instead of a year early.
Law 03
Nobody advertises it
In one qualified pool of 13,926 founders, exactly nine mentioned visa intent anywhere public. Keyword search on this population measured ~92% false positives. You cannot find these people by searching for the problem; you find them by the shape of their career. The full count is on the ICP 03 page.
Which criteria the data can see
All ten criteria, graded honestly
Numbered as in the EB-1A regulation, 8 CFR 204.5(h)(3)(i)–(x). We would rather tell you what we cannot see than have you find out on a delivered file.
| # | Criterion | Read from | Coverage | How we use it |
|---|---|---|---|---|
| i | Nationally or internationally recognized prizes or awards | awards records | Partial | 22.1% raw in one qualified pool — but must be classified by issuer before it counts (see findings below). |
| ii | Membership in associations that require outstanding achievement | weak proxies only | Weak | No clean field; professional-group membership measured zero non-empty across an entire pool. |
| iii | Published material about the person in professional or major media | — not in career data | Not in data | External media search only; we flag, we don’t claim. |
| iv | Judging the work of others | self-description / maintainer status | Weak | Inference from self-declaration only — flagged, never asserted. |
| v | Original contributions of major significance | patents records + open-source impact | Partial | Present on 8.5% of one qualified pool — an evidence booster, never a filter. |
| vi | Authorship of scholarly articles | publications records | Partial | 13.2% of the same pool; strongest on PhD-anchored leads. |
| vii | Display of work at artistic exhibitions or showcases | — | Not relevant here | Not relevant for the technical and business cohorts in this playbook. |
| viii | Leading or critical role for distinguished organizations | career-position graph + employer profile | Strong | Best-detected criterion — every delivered row carries a dated title and employer; title compliance measured 100% on shipped files. |
| ix | High salary or remuneration | inferred salary band | Weak | A screening proxy at ~9% density, not evidence. |
| x | Commercial success in the performing arts | — | Not relevant here | Not relevant for the technical and business cohorts in this playbook. |
The honest funnelCriterion viii and advanced degrees are the dense signals — reliably detectable at scale, which is why every profile is built on them. Awards, patents and publications are sparse boosters: we attach them as evidence when they are there, and never filter on them, because filtering on a signal present in one profile in twelve manufactures a tiny list that looks impressive and converts badly.
What one delivered row contains
Not a name and a title — a record an attorney can qualify from
Candidate A•K•-2026
Foreign-degree engineer in the US — O-1A / EB-1A shape
Senior machine-learning engineer · computer vision · 11 yrs
US metro (region-level here, exact on delivery) · institute-of-technology undergraduate, outside the US · ~900 connections · 14 positions on file
Identity & contact
full name · profile URL · personal email · secondary emails · phone where held · city / region / country · connections · followers · last observed activity
Current position
title · normalized role · seniority band · start date · employer · employer size · industry · employer country · domain · description
Career history
every prior role with start and end dates · employers and their size bands · computed total experience · computed years inside the US · promotion sequence
Education
institution · degree · field · years · foreign-vs-US verdict with the reason it was reached · bachelor country and region · gap between first and graduate degree
Petition evidence
certifications · languages · awards · publications · patents · projects · volunteering — each counted, with the underlying text where present
Screening & ranking
lead score with its components shown · title band · contactability · activity recency · inferred salary band · the criteria this person visibly hits
Delivered as XLSX, CSV, JSON and JSONL in the same package, plus a one-row-per-person summary sheet for the people on your team who will never open a JSON file. Measured on the engineer cohort: 129–145 fields per delivered record. Other cohorts measure differently — we quote the depth of the file you are actually buying. Values shown are a representative composite, not a real individual; the field structure is real.
Freshness
A list is a photograph. This is a feed.
Every signal these profiles rest on is a date that keeps moving. We measured how fast that happens, because “fresh data” is a claim anyone can make.
drift in 15 days
Of 500 profiles vetted against a spec and re-tested 15 days later, 473 still qualified. The rest had moved — new role, new tenure, changed situation. That is the rate at which a static list stops being true.
held on re-enrichment
A 1,991-profile re-enrichment and re-filter held at 98.2% compliance — so the drift is genuine movement in people’s lives, not noise in the pipeline.
the window refills
Every month a new cohort crosses into your criteria: a class graduates, a founder title switches, a tenure counter reaches your threshold.
repeat rows
Every delivery after the first is checked against what we have already sent you. Send your CRM export and it joins the ledger.
Why we read the rows before we quote a number
Four findings from reading records, not counting them
Doctorates — clean
All 22 sampled doctorate records were real and the fields mapped straight onto national-interest and extraordinary-ability arguments. No false positives. This signal can be trusted as-is.
Awards — depends entirely on the issuer
A global sample returned genuine distinction — world-body fellowships, major business awards. A same-size regional sample returned roughly 15 of 22 long-service and corporate commendations that are not petition-grade. The raw award count is worthless until it is classified by who issued it.
Secondary schools posing as degrees
An audit found 61 profiles whose only “foreign education” was a school-leaving board — 23 of them an international baccalaureate — plus seven where a bootcamp had been read as a university. All removed before shipping.
Keyword search, measured
Naive substring matching on petition types returned 9,219 hits; word-boundary matching returned 298. “TN” matches inside partner and attorney; “perm” inside permanent; “visa” inside advisor. A ~97% false-positive rate, invisible unless you read the rows.
This is the unglamorous part, and it is the part you are paying for.
Anyone can run a query. The difference between a list your team works all the way through and one that quietly dies in a spreadsheet is whether somebody read a few hundred rows first and asked what each signal actually means. Every defect described here was found by reading — sometimes by us, sometimes by the practice — and each one is now a test that runs before anything ships.
Guardrails
Lines we hold, on your behalf as much as ours
Final wording for your jurisdiction is your counsel’s call, not ours.
Flat fee or fixed scope — never a share of the engagement.
Fee-splitting with a non-lawyer is a conduct problem for you; we price so it cannot arise.
Research, not solicitation.
We deliver research; the firm decides who is contacted, how and when. [TO BE WRITTEN WITH COUNSEL — how ABA Model Rule 7.3 and your state’s version on solicitation and attorney advertising apply to how the rows are used]
We describe evidence, never eligibility.
Scores and tiers are evidence-strength for an attorney to qualify. No dataset can make a sustained-acclaim determination, and we never imply one.
We do not claim to know anyone’s status.
Nationality, visa type, filing history and lottery outcomes are not in career data. Everything here is a structural proxy, and we say so on every delivery.
Every batch is audited before it ships.
Geography, title, blacklist-leakage, duplicate and corrupt-record gates, with the audit result attached to the delivery. No fabricated contacts: a missing email is delivered empty, never guessed.
Signals we cannot build, we decline to build.
Real-time activity, group membership, citation counts and granted-patent status have all been asked for and reported back as unavailable rather than approximated.
Data sources and legal basis
Where the data comes from, and on what basis
What the career graph is built from
[SOURCES — to be written with counsel: what the graph is built from and on what legal basis]
California (CCPA / CPRA)
[CCPA — how Californians’ data is handled, opt-out and deletion requests]
EU and UK contacts (GDPR)
[GDPR — legal basis for EU/UK contacts, or confirmation they are excluded]
Removal requests
[HOW A PERSON ASKS TO BE REMOVED — address or form]
Method
Every figure is either a measurement from a delivered file (row counts, coverage rates, audit results, defect counts) or a live read-only query against the Outricher career graph — dated career positions, education records, awards, patents, publications, certifications, languages, and a contact layer. Volumes describe work delivered to one visa practice between March and September 2026 and are not a statement about market size. Pool figures are counts of people meeting a stated set of conditions at the time of the query.
Integrity
No client, employer or individual is named anywhere; every example row is anonymized to role, employer size band, region and signal. We publish no campaign metrics — open, click and reply rates were never measured by us and are not claimed here; the evidence we do have is that segments were re-ordered. Award records must be classified by issuer before being described as petition-grade. The dossier is a representative composite; its field structure is real.
Test the method on your own ICP.
Send one paragraph and get a free sample back — plus one fully assembled record like the one above. Judge us on rows, not on a deck.