Build vs. Buy: What an In-House Company-to-Domain Resolver Really Costs

Build vs. Buy: What an In-House Company-to-Domain Resolver Really Costs

An 80-person SaaS team spent 14 months on a home-built resolver. Here are the real numbers, and what they learned.

Disclosure: FindCompanyDomain sells a commercial domain resolver. The scenario below is a composite of build-vs-buy patterns, not one named client. All figures, including the accuracy scores, are illustrative.

The risk nobody tracks: confidently wrong answers

The worst thing about a home-built resolver is not that it fails. It is that it keeps returning an answer for every record, so nobody notices when the answers go stale.

Owen, head of data engineering at an 80-person B2B SaaS company, found this out the slow way. His resolver scored 80% accuracy at launch. Fourteen months later the same 400-record test set showed 70%. That is 320 correct records falling to 280. No one on the SDR team flagged a pattern, because individual wrong domains look like normal noise. Only a fixed ground-truth set showed the drift.

Accuracy on the same 400-record test set

In-house, launch
80%
In-house, month 14
70%
Commercial tool
89%

Illustrative figures from the composite scenario.

The wrong answers also came with high confidence. On a first manual check of 50 records, 42 were right and 8 were wrong. Seven of the 8 wrong ones were non-US companies, and they carried an average confidence score of 0.84.

What the build took

Owen and his senior engineer, Marcus, built it over 11 weeks. The first plan had 4 components. The final system had 8:

  • Entity resolution
  • Data source connectors
  • Disambiguation engine
  • Confidence scoring
  • Domain verification
  • Data refresh pipeline
  • Monitoring and alerting
  • Fallback handler for low-confidence records

Entity resolution was the surprise. “Apex,” “Apex Solutions Inc.” and “apex-solutions” can be one company in a prospect list, and deciding that takes real matching logic, not text cleaning.

The numbers

Fully loaded engineering cost was $72 per hour.

Cost itemAmount
Build880 hours × $72$63,360
Data licences$19,200 a year, about $22,400 over 14 months (domain registration data $4,800, DNS and MX data $6,000, a second business registry for Europe and APAC $8,400)$22,400
Maintenance18 hours a month × 14 months = 252 hours × $72$18,144
Total over 14 months$103,904

A commercial tool at about $400 a month would have cost $5,600 over the same 14 months.

Licence prices vary a lot by provider, so plug in your own quotes. Not counted on either side: hosting and monitoring tools for the in-house build, and the 6-week migration to the commercial tool. Treat the total as a floor for the in-house cost.

The number that matters more: cost going forward

The $103,904 is already spent. The better question at month 14 is what each option costs from here.

In-house, per month

~$2,900

Maintenance about $1,296 plus licences $1,600

Commercial, per month

~$400

Flat monthly price

That is roughly $2,500 a month, or $30,000 a year.

In this scenario, a commercial tool scored 89% on the same 400-record test set, against 70% for the in-house build.

Why maintenance never ends

A company rebrands and moves domain. The registry updates 3 months later. The resolver keeps returning the old domain.

A data provider changes its response format without notice. The connector breaks silently for 4 days before monitoring catches it.

A batch of early-stage companies uses newer domain endings, and the matching logic treats the ending as part of the company name.

Each fix was small. Together they cost 18 hours every month.

Accuracy also fell because 2 of the 3 data subscriptions refreshed only quarterly. That was a design and budget choice, not a rule. A monthly refresh would have slowed the drift at a higher licence cost. Either way, the cost is real.

When building does make sense

Building is not always wrong. It can be right when:

  • Domain resolution is part of your core product, not a support tool.
  • Your volume is high enough that vendor pricing goes past the roughly $2,900 a month it costs to keep your own system alive.
  • Your records sit in a narrow region or industry where 1 or 2 data sources are enough and rarely change.
  • You have compliance or data-control rules that rule out sending records to a vendor.

If none of these apply, the build is likely to cost more than the tool it replaces.

What Owen got out of it

When Claire, his VP of Sales, asked what the company got for the money, Owen listed 3 things:

A clear spec of what a resolver must do
A sense of the edge cases
A way to judge vendors, because he had built a weaker version himself

It was useful, but not worth the extra spend. She approved the migration.

Before you decide, 4 checks

  • Build a ground-truth set of at least 300 to 400 records where you already know the correct domain.
  • Test any resolver, yours or a vendor’s, on that same set. Repeat it every quarter.
  • Count the forward monthly cost of your build, not only the money already spent.
  • Ask what happens to a record the system is unsure about. A system that returns a guess for everything is hiding its errors.

Frequently asked questions

How much does it cost to build an in-house company-to-domain resolver?

In this illustrative scenario, it cost $103,904 over 14 months. That was $63,360 for the build (880 hours at $72), about $22,400 for data licences, and $18,144 for maintenance (252 hours at $72). A commercial tool at about $400 a month would have cost $5,600 over the same period. Hosting, monitoring tools and migration are not counted, so treat the total as a floor.

Why does a home-built resolver lose accuracy over time?

Data keeps changing. Companies rebrand and move domains, but registries update months later. Data providers change response formats without notice. New domain endings confuse matching logic. In Owen’s case, 2 of the 3 data subscriptions also refreshed only quarterly. Accuracy fell from 80% to 70% in 14 months.

What is the real monthly cost of keeping an in-house resolver running?

About $2,900 a month: roughly $1,296 for maintenance (18 hours at $72) plus $1,600 for licences. A commercial tool is about $400 a month, so the gap is roughly $2,500 a month or $30,000 a year. The money already spent is gone, so this forward cost is the number to compare.

Why are confidently wrong answers a bigger risk than failures?

A failure is visible. A confident wrong answer looks like a normal result, so nobody flags it. In the scenario, 8 of 50 records checked were wrong, and 7 of those 8 were non-US companies with an average confidence score of 0.84. Only a fixed ground-truth set showed the drift.

How big should my ground-truth test set be?

At least 300 to 400 records where you already know the correct domain. Test any resolver, yours or a vendor’s, on the same set, and repeat it every quarter. This is the only reliable way to spot drift.

When does building a domain resolver make sense?

When domain resolution is part of your core product, when your volume makes vendor pricing go past the roughly $2,900 a month it costs to run your own system, when your records sit in a narrow region or industry where 1 or 2 data sources are enough, or when compliance rules stop you sending records to a vendor.

What should a resolver do with a record it is unsure about?

It should send the record to a fallback instead of guessing. A system that returns an answer for every record is hiding its errors. Always ask what happens to low-confidence records before you trust any resolver.

FindCompanyDomain refreshes its data continuously, so the domain returned reflects the company’s current setup. 500 free credits, no credit card, API from day one.

Get 500 free credits