you asked for the mini thesis — here it is

A map into every account within your niche.

A niche is finite — which means it can be written down completely: every company judged with written reasoning, every buying committee derived from each company's actual state. This is one we built and delivered, anonymized.

Audit the method yourself:
↓  Download the pack
or$npx tam-map
1 — download and unzip it (or run the command)
2 — drop the files into ChatGPT, Claude, Cursor — anything that reads files
3 — paste EVALUATE.md — it audits the method, then retrofits it to your market
▶
Watch the build, narrated
how the map was made · how to use the pack — 4 min
the method — four connected steps
01 · SCAN
Extraction
Five source angles, merged into one universe.
02 · HARNESS
Context
Every page and profile, read in full.
03 · PROCESS
The filter
Judged against a written rulebook — pass and fail, both kept.
04 · INTERNAL MAP
Head of SOCchampion
CISOeconomic buyer
IR Managerend user
Sales Directorcut
CTOtechnical evaluator
CFOfinance gate
Threat Hunterend user
CEOcut
Security Engineerend user
Leiter IT-Sicherheitend user
Every title scored inside its own company's roster.
01SCAN

Every way in, taken.

A market's companies don't sit in one place, and no single source holds them all. We come at the universe from five directions and merge what comes back — coverage is a product of the angles, not the size of any one list.

Generic company sourcesLinkedIn · Prospeo · PDL — the commercial layer, cross-checked against each other
Signal-based sourcesCrunchbase · TheirStack · Bombora · BuiltWith — funded, hiring, tooling, in-market
Web keyword matchingwe index what a site actually says, then find every other site that says it — the lookalike-by-language angle
AI matchingan agent reads the shortlist the way a human would and judges fit before a single row is enriched
AGGREGATEmerge · dedupe
domain-resolve
Reverse job titlesfind people holding titles that only exist inside companies like your buyers’ — then walk back to the employer
Public directoriesevery association, marketplace and vertical index — sitemap-enumerated in full, never sampled
Groups & communitiesLinkedIn · Facebook · X · Slack — where the market gathers; membership is a membership list
Lookalike expansionruns lastthe judged set teaches the search what to look for, then we sweep again
→ everything lands in one table, keyed to the domain. the fifth angle runs after processing — which makes this a loop, not a line.
02HARNESS

Read everything the company
says about itself.

Before a single judgment, every account is read in full — not sampled, not summarized from a database field.

aurigadefense.io [ 200 OK ] [ SCRAPING ]
website · full sitemap
/homepage
/services/mdrservice page
/services/incident-responseservice page
/aboutorigin, team
/careers4 open roles
/certificationsiso27001 · crest
/locations3 offices
linkedin
company profiledescription
specialities11 tags
headcount74 · by function
personnel roster38 titles
[ .MD ][ .JSON ] harvesting…

          
→ 21 evidence columns per account, in the company's own words. This is what the rubric reads — nothing is judged from a database field.
03PROCESS

State → structure → inference.

This is, more or less, everything money can buy about a company and a person. The map isn't in the payload. It's between its fields.

layer 1 · static state — the payload, as sold
Company planeper account
identity + classification
nameprimary_domaincompany_typeyear_foundedindustrysub_industryspecialities
firmographics
size_bandhq_countryhq_locationrevenue_est_range
headcount + composition
headcount_totalheadcount_growthteam_composition function→countseniority_mix
technographic + web
techstackweb_keywordsweb_trafficfollowerscompetitor_domains
Person planeper contact
identity
full_nameheadlinesummarylanguages
role
current_titlenormalized_title dept·sub-deptseniorityfunctiontenureyears_experiencepast_employers
education + skills
schools·degreesskills
network
connectionsfollowers
not state: funding events · hiring activity · job changes · open-to flags — those are signals: motion, not structure. They stack on the map later, scanned by agents. →
layer 2 · structure — merge, then mint
One typed record per account
Providers disagree; shapes differ; fields arrive broken. Everything cross-validated and keyed to the domain.
domain-keyedconflict-resolvedprovenance kept
Fields nobody sells, minted
Computed between the payload's fields — or read off the company's own website. Filterable like any other column.
sales:tech_ratiotitle_presence / absenceage_vs_sizegovernance_postureservices_offeredcore_servicedelivery_modellocations_countgeographies_servedverticals_servedcustomer_size_servedcertifications_heldclient_count_claimclaimed_vs_delivered+ whatever your niche demands
fully custom per niche — defining this layer is exactly what our rubrics do.
layer 3 · inference — the point
Permuted, the fields read like a diagnosis of the company's actual state:
sales:tech 3:1×services: generic
Reseller, not an operator.
→ disqualified — reasoning written down
age 3y×60 fte×no CISO
Escape velocity — budget, no process.
→ champion sits in engineering
iso 27001 on site×no compliance titles
Compliance is bought, not built.
→ partner motion, not replacement
age 14y×12 fte×flat
Lifestyle business.
→ deprioritized, reason attached
own-SOC keywords×analyst-heavy roster
Delivery-led operator.
→ bullseye segment
enterprise clients claimed×no enterprise titles
Overclaiming — mid-market in reality.
→ message the reality, not the website

structure holds still. signals move. next: agents scanning the map, stacking signals live →

04INTERNAL MAP

Every company is
its own universe.

A title means nothing on its own. It means something relative to the other titles standing next to it. So we never score a person as a row — we score them inside their company's own array.

1
Pull the whole rosterEvery senior person at the company — not a sampled three, not the first CISO we find.
2
Collapse it into one title arrayDeduplicated, normalized by function across languages — RSSI and Leiter IT-Sicherheit resolve to what they actually do.
3
Score each row against its own arrayChampion · economic buyer · technical evaluator · end user · finance gate · cut. The verdict is relative to who else exists.
4
One judgment call per companyThe whole array goes in together — because you cannot judge a row without seeing the room it stands in.
~600 peoplea security org exists
CISOeconomic buyer
Head of SOCchampion
IR Managerend user
Head of ITcut
CEOcut
the same title,
opposite verdicts
15 peopleno security org
Foundereconomic buyer
Head of ITchampion
Security Engineerend user
Sales Directorcut
the map, live

The database. Not a screenshot.

Every qualified account, read live from the delivery spreadsheet. Search it, cut it by tier, open any company — its committee is inside, each keep with its reasoning.

CompaniesPeople
⌕  Search company, domain, or name…
1,544 companies · live
Company
Ctry
Employees
Core service
People
Tier
Auriga Defenseaurigadefense.io
DE
74
managed detection response
4
bullseye
Nordwall Securitynordwall.se
SE
31
socaas
3
bullseye
Helios IT Grouphelios-it.fr
FR
210
socaas
5
good
Cleric Cyberclericcyber.co.uk
UK
18
incident response
2
bullseye
Vantano Systemsvantano.it
IT
96
managed detection response
4
good
Bastion Noordbastionnoord.nl
NL
44
siem management
3
bullseye

switch to People and cut by committee role — every row carries the rule that placed it. Click any company → its buying committee slides in, ranked by economic power, reasoning attached. (names above are illustrative — the live page reads the real 1,544)

signals

Custom agents, scanning for motion.addon

Structure holds still. Signals move — orthogonal to the map, landing on a company or a person. Agents defined for your niche sweep on schedule; every change flagged, reasoned, routed.

the agents — custom per niche
leadership-watch
new security leadership, exits, promotions
hiring-watch
role openings by family and velocity
funding-watch
rounds, investors, acquisition events
stack-watch
tooling adopted or dropped
web-watch
new services, pricing pages, positioning shifts
+ your niche's signals
defined custom, per niche — like the rubrics
the map, being swept
Auriga Defensebullseye · mdr74 fte⚡ new Head of SOC
Nordwall Securitybullseye · socaas31 fte⚡ 3 analyst openings
Helios IT Groupgood · socaas210 fte⚡ adopted CrowdStrike
Cleric Cyberbullseye · ir18 fte⚡ Series A · €8M
Vantano Systemsgood · mdr96 fte⚡ launched MDR page
Bastion Noordbullseye · siem44 fte⚡ CISO changed jobs
Corvus Managed
good · edr130 fte⚡ hiring SOC lead
Sable Operationsbullseye · mdr52 fte⚡ new UK office
the output — signal → play
the map ships as the pack + your database as csv. with agents: a live warehouse — signals typed back into each account, queryable, always current.

every flag lands with the account's full structure underneath it — that's why the play writes itself.

how we work

Three weeks. Two hours of yours.

The build is ours. The judgment calls stay yours — placed exactly where they matter.

week 1week 2week 3
us
rubric definition
companies built + judged
committees built → delivery
you
strategy call60 min
rubric sign-off30 min
sample approval15 min · async
→ you pay here
after you've approved the companies. we build on trust — by this point, most of the heavy lifting is already done.
your total contribution: ≈ 2 hours. everything else is the machine and the rubrics.
01 · YOUR ICP THESIS
Who's in, who's out — as you see it
One call. Your qualification language becomes the rubric's doctrine.
02 · THE CAP
People per company
Your cost lever. Contacts beyond the cap are free overdelivery.
03 · A SAMPLE VERDICT
15 minutes of judgment
You review a random sample before the full run. The gate only opens on your yes.
pricing

One price for the build.
Then only what you add.

No per-seat, no platform fee, no lock-in — you own everything that ships. Most markets fit inside the build; you pay by the row only if yours runs past it.

one price, the whole build
$5,500
up to 30,000 companies mapped  ·  up to 50,000 contacts found
beyond that, you pay by the row
+$0.03per companyabove 30,000
+$0.05per contact foundabove 50,000
+$0.03personal emailoptional · ~60% found
+$0.10phone numberoptional · ~60% found
Nothing is charged for a miss. Above 100,000 companies or 200,000 contacts, the build is scoped custom. Custom signal agents are quoted separately.
terms: you pay after the companies are approved · contacts on final delivery
build your quote
Companies in your market30,000
in the $5,500
+$0.03 each
100k+ custom
Contacts to find50,000
in the $5,500
+$0.05 each
200k+ custom
optional, per row found
60% is our measured average — change it to your own assumption and the quote follows.
your estimate
The build30,000 companies + 50,000 contacts$5,500
Companies beyondinside the build$0
Contacts beyondinside the build$0
+ Personal emails—$0
+ Phone numbers—$0
Total, one time$5,500
≈ $0.11 per contact, all-in · ≈ $0.18 per account mapped
Nothing is charged for a miss.
a recent work of ours

Audit the method yourself.

The European MSSP map — the one running through this page — packaged whole: both rubric stacks, the pipeline doc, the build report, and all three exports. Five minutes in your terminal and your own AI tells you whether the method holds.

1,544companies mapped
7,187rejected, with reasons
2,422committee contacts
$5,500the build covered it whole
1
Get the pack
Either way you get the same folder — tam-map-mssp/.
terminal
$npx tam-map
pulls and unpacks it into the folder you're in
no terminal
↓ download tam-map-mssp.zip
unzip it anywhere — same contents, ~6 MB
2
Give the files to an AI — any AI
In the browser (no setup): in ChatGPT or Claude, attach RUBRIC_GUIDE.md, HOW-IT-WAS-BUILT.md, the two rubric folders, and the CSVs from data/.
In an editor: open the folder in Cursor, VS Code, or Claude Code and let the assistant read it in place — better for the big CSVs. Either way, nothing to install and no key to paste.
3
Paste this into the assistant
It's also saved in the pack root as EVALUATE.md — open it and copy from there if you prefer. It runs four passes: spot-check the verdicts against the rulebook, stress-test the three claims with real row counts, list the weaknesses plainly — then ask what your market is and sketch this exact system rebuilt for it.
EVALUATE.md
You are auditing a market-mapping system I was sent. This folder (tam-map-mssp/)
contains a complete TAM build for a market that is NOT mine — European managed
security. I am evaluating the METHOD, not shopping the list. Be critical and
unbiased: I want an honest read, not a promotional one.

Read first, in this order:
1. README.md and HOW-IT-WAS-BUILT.md — what was built and how
2. pipeline/PIPELINE.md — the data machine (sources, Clay, context, waterfall)
3. company-rubric/00_precedence_canon.md + 03_qualification_doctrine.md — the rules
4. persona-rubric/00_persona_doctrine.md — the buying-committee engine
5. The exports in data/ (column spec: RUBRIC_GUIDE.md Part 2 — treat as ground truth)

Then work through four passes:

PASS 1 — VERDICT QUALITY
Spot-check 10 random disqualified companies and 10 random qualified ones against
the doctrine, using their own context columns as evidence. Do the verdicts hold?
Report any that look wrong, with the rule they appear to violate.

PASS 2 — THE THREE CLAIMS
a) ALL-INCLUSIVE: is the coverage genuinely exhaustive or quietly sampled? Check
   the funnel math, the per-directory yields, and whether rejections carry reasons.
b) MICRO-SEGMENTATION: could a campaign actually be cut into tight micro-segments
   from these fields? Build 3 examples and report their real row counts.
c) COMMITTEE FROM ACTUAL STATE: pick 5 qualified companies of very different
   sizes; check whether each committee reflects who actually exists on that
   roster rather than a fixed persona template.

PASS 3 — WEAKNESSES
What would you criticize? Where could this break, drift, or mislead?
Homepage-only evidence, language handling, model dependence. Say it plainly.

PASS 4 — THE RETROFIT
Ask me what my market is. Once I answer, sketch what this exact system would
look like rebuilt for it: the sources you would scrape and why, the single gate
question, the likely disqualifiers, the tier logic, the committee ladder for my
buyer, and what my funnel's shape would plausibly be. Flag which parts transfer
as-is and which need re-derivation from my ICP.
what you actually get out of this

Understand your market —
better than anyone else competing for it.

Not a list. A written-down market: every account judged, every rejection reasoned, every buying committee derived from who actually works there.

Let's map your TAM → 30 minutes. We look at your market live and tell you honestly whether it's mappable.
how it starts
  1. 1
    Book the callWe look at your market together — no deck, no discovery theatre.
  2. 2
    Rubric definitionWe draft your qualification doctrine from that call. You sign it off.
  3. 3
    The map, in three weeksPack, database, committees. You pay after the companies are approved.
Relevance TAM relevanceai.io we write markets down for a living