How a Domain Lookup Site Became an Intelligence Company: DomainTools Has Sold DNS Data for Twenty Years

At two in the morning, a bank security analyst stares at an unfamiliar domain in a phishing email. He types the string into DomainTools' Iris, and seconds later a relationship graph spreads across the screen: the domain was registered just three weeks ago, seven other domains hang beneath the same registration email, and the server it resolves to has historically hosted a batch of sites long flagged in red, with a risk score of 99. The analyst didn't run the resolution himself, nor did he dig th

Chahu Team2026-11-075 min read

At two in the morning, a bank security analyst stares at an unfamiliar domain inside a phishing email. He types the string into DomainTools' Iris, and seconds later a relationship graph unfolds on the screen: the domain was registered just three weeks ago, seven other domains hang off the same registration email, and the server it resolves to has historically hosted a batch of long-flagged sites. Risk score: 99. The analyst didn't run the resolution himself or dig through historical archives. Those answers were bought.

The company selling those answers has spent more than two decades accumulating data around DNS and domains.

It started with a Whois lookup site in a garage

In 2002, Jay Westerdal wrote Name Intelligence in his parents' garage; it was later renamed DomainTools. Back then it did one simple thing: tell you who a domain is registered to now, and who it was registered to in the past. Its Whois history reaches all the way back to 1995, and that stockpile became its hardest asset.

In 2008, Westerdal sold the company to Thought Convergence for a reported price between $16 million and $18 million. It changed hands several times after that, secured backing from the investment firm Battery Ventures, set up headquarters in Seattle, and has long kept its team at around 140 people. A small domain-lookup site slowly grew into an intelligence supplier for banks, law enforcement, and security teams.

How it actually makes money from DNS

Today DomainTools slices domain and DNS data into a row of products and charges for each one separately. Analysts use Iris Investigate, an investigation console that drills down through registration details, resolution records, SSL certificates, and email addresses layer by layer. It's sold as a per-seat subscription, with public estimates putting the cost at $15,000 to $45,000 per analyst per year. To feed intelligence into a company's own SIEM and SOAR, customers switch to interfaces like Iris Enrich, billed by query volume; it can enrich 6,000 domains per minute in batches, and a mid-sized security team typically spends roughly $30,000 to $80,000 a year on the interface. Customers watching for brand impersonation buy Iris Detect, priced by brand portfolio at $25,000 to $60,000 a year. Put these together into a deployment for a mid-sized security operations center, and $80,000 to $200,000 a year is a common figure.

Beyond subscriptions and usage-based billing, it also lists its data on AWS Marketplace, Microsoft Sentinel, and the UK government's G-Cloud procurement framework, letting customers order directly through cloud bills they already have. There's another line of business too: packaging data as feeds—such as the daily-updated high-risk domain list Domain Hotlist and the newly observed domains feed—and embedding them directly into other vendors' security products for sale.

Where the core value lies

What it really sells is a capability: given an unfamiliar domain, answer two questions in seconds—who is behind this domain, and what else is it connected to. Anyone can do resolution itself; the hard part is tying twenty years of registration changes, global resolution observations, and who else has lived on the same infrastructure all to this one domain.

That drill-down is valuable in investigations. Following the registration email of a malicious domain can uncover a string of domains registered by the same person; following the IP it resolves to reveals what else that server has hosted historically; and layering on a risk score from 0 to 100 lets analysts put the most suspicious targets first without having to judge for themselves. Time depth locks in its advantage: an attacker can change Whois and swap resolutions today, but he can't change history—and history is exactly what DomainTools has been storing for over twenty years.

Where the data comes from

DomainTools' data comes from three distinct collection lines.

The first is Whois history. Since 1995 it has been crawling and archiving domain registration information—who registered it, when, how many times it changed hands—storing it record by record. The second is active resolution: it regularly queries authoritative servers for domains, snapping current records like A, MX, NS, and SSL into time series. The third is passive DNS, which it only completed by acquiring Farsight Security in November 2021. Farsight's flagship product, DNSDB, bills itself as the world's largest passive DNS database, and it's led by DNS veteran Paul Vixie.

Passive DNS collects data in a completely different way from active resolution. It never initiates queries itself; instead it places a row of sensors above recursive resolvers and records the responses those resolvers get back from authoritative servers along the way. The sensors see only the resolution results themselves—query name, record type, answer, first and last seen times, and how many times it appeared—not who or which machine initiated the query. This kind of collection neither alerts the target being looked up nor fails to anonymize by design.

Scale shows the weight of this network. DNSDB holds more than 100 billion resolution records, takes in two terabytes of new data daily, observes 200,000 to 300,000 resolutions per second globally, and handles real-time throughput of one gigabit per second. The concept of passive DNS itself was proposed by Florian Weimer in 2004 to combat malware; Farsight turned it into a business, and DomainTools folded that business into its own investigation console.

How it preserves the value of its datasets

The first reason this data is hard to replace is time. Whois history reaches back to 1995, and passive DNS has accumulated continuously since 2010. That length of time can't be backfilled; anyone wanting to build an identical copy from scratch would need another twenty years.

The second reason is coverage. How much passive DNS can see depends on which resolvers the sensors sit above, and no sensor network can see the whole world's resolutions—so whoever has collection points spread wider and closer to the backbone has more complete data. Farsight also built a data exchange platform called SIE that pools observations from more participants, further expanding coverage.

Processing adds another layer of value. Raw resolution and registration records are linked into a relationship network you can jump across, then fed into data science models that compute predictive risk scores for domains, hosts, IPs, and name servers. What customers pay for is this processed judgment; the raw streams underneath are just semi-finished goods.

The real test comes from upstream changes. On January 28, 2025, ICANN officially retired traditional Whois for gTLDs in favor of structured RDAP, and DomainTools itself has posted notices that its Whois service may be discontinued. The way registration information is obtained is changing, but the historical stockpile and passive observation network it has built over twenty years are precisely what new protocols can't replace.

How it monetizes

Turning this into money follows a path of layer-by-layer markup. At the bottom is raw observation—a single resolution record isn't worth much on its own. Linked upward into a relationship network, where one clue can jump to a whole swath of infrastructure, the value rises. Layer on a risk score and hand the judgment directly to the customer, and you can charge another tier. At the top is embedding into customer workflows, so every SIEM alert automatically calls a DomainTools interface—the deeper the usage and the more frequent the calls, the higher the bill.

Pricing forks along the same path. The investigation console is sold per seat to analysts doing manual triage; interfaces are sold by query volume to security teams that need automation; data feeds are packaged for OEM embedding by other security vendors. The same data is sliced into different granularities and sold to different people.

A few ideas for commercializing DNS

Breaking down the DomainTools path yields a few reusable practices for commercializing DNS data.

First, accumulate one-off resolutions into an asset with a time dimension. A single resolution is worthless once used, but store it continuously for ten years and it becomes a stockpile no one can catch up to. Anyone in the DNS business should start storing history on day one.

Second, use passive collection to gain coverage while avoiding privacy landmines. Sitting above recursive resolvers and copying only responses, never touching user identities, accumulates massive observations without crossing the red line on personal data—this step determines whether the data can scale compliantly.

Third, focus on processed judgment. From records to relationship networks to risk scores, each layer of processing lets you charge another tier; what customers want to save is their own triage time.

Fourth, embed data into workflows customers already use—SIEM, SOAR, and MSSP platforms, for example—rather than expecting customers to spin up a separate system for your data. The deeper the embedding, the higher the switching cost and the steadier the renewals.

Fifth, think ahead about upstream changes. Shifts like Whois moving to RDAP can pull out from under you a source you depend on at any time; what truly withstands change is the historical stockpile you've built and the observation network you've laid down—things that don't disappear when upstream protocols change. DomainTools wasn't panicking when it posted notices that Whois might be discontinued, because it stopped relying on any single interface long ago.

References