IP to Company Match Rate: What Is Realistic and Why It Varies

A realistic IP to company match rate on B2B traffic is roughly 20 to 35 percent of visits, according to BusinessMCP’s published guidance, and lower for SMB or consumer-heavy audiences. A rate far above that usually means internet providers are being counted as companies. Precision on the names matters more than the percentage.

Match rate is not one number

People quote “match rate” as if it were a property of a vendor. It is really a property of a vendor and a traffic mix. Consider the same service applied to two audiences.

Audience	Typical network mix	Expected company-level resolution
Enterprise software buyers, office-based	More corporate egress ranges	Higher end of the range
Mid-market B2B	Mix of office and home lines	Middle of the range
SMB, freelancers, agencies	Mostly home and mobile	Lower
Consumer or content traffic	Almost all ISP and mobile	Very low

The 20 to 35 percent figure appears in BusinessMCP’s guide to how visitor identification works and its evaluation guide. Both are vendor-published. They are useful as a sanity check, not as a promise.

What drives the rate up or down

Remote and hybrid work. A person on home fibre shows the provider’s name. As more work moves home, fewer visits come from office ranges.

Secure-access services. Many large firms send staff through security gateways (SASE or secure web gateways). The visible address then belongs to the vendor. A careful resolver flags those ranges instead of attributing them. BusinessMCP’s product page says it does exactly this. We cover the topic in VPN and proxy identification.

Mobile traffic. Carriers use shared addresses at scale. A carrier name is not a prospect.

Cloud and bots. Automated traffic arrives from hosting ranges. It inflates visit counts but never resolves to a customer. See filtering bot and cloud traffic.

Company size and region. Larger organisations tend to hold their own address space. Small firms rely on shared providers, so they rarely resolve.

Data freshness. Address blocks are leased and sold. A graph rebuilt nightly can still be a day behind on a range that changed hands today, as BusinessMCP’s own page notes.

Key takeaway: the rate is a mirror of your audience. Improving the vendor changes the number less than changing who visits.

How vendors inflate the number

Be sceptical when you see:

• 70 percent or higher quoted without a traffic description.
• Counting ISP hits as matches. If broadband providers appear in the “identified companies” list, the number is padded.
• Matches per session versus per address. A busy network visiting 500 times counts as one company, not 500.
• No class or confidence fields, so you cannot see what was counted.
• Person-level claims. Names of individuals from an IP alone are not something a reputable company-level service provides, and the legal footing is weak in the EU. See the GDPR article.

Three metrics worth tracking

BusinessMCP’s evaluation guide suggests three measures. We agree with them and add one note on each.

Metric	Definition	Note
Precision on named results	Correct names divided by names returned	Hand-check about 50 per vendor
Carrier and vendor leakage	Share of “companies” that are really ISPs, carriers or gateways	Should be near zero
Coverage of resolvable visits	Named results divided by visits that could in principle be named	Hard to know; use as a relative measure between vendors

Raw fill rate is deliberately absent, because it rewards guessing.

Measuring your own rate

1. Pull one day of server logs or analytics exports. Remove bots you can already identify and your own office.
2. Deduplicate to unique client IPs. If behind a CDN, use the forwarded client address, and only for requests that verifiably passed through your proxy.
3. Resolve the set with one or more vendors, for example using bulk lookups.
4. Tally verdicts: company, non-business (ISP, mobile, hosting, VPN) and unidentified.
5. Sample about 50 named companies. Check each against a known customer list, a form fill from the same domain or a plain web check.
6. Report both the named share and the precision.

The step-by-step is in how to test an IP to company API.

Should the rate change your decision?

It should change the business case, not necessarily the decision. If 25 percent of your visits resolve and precision is high, you gain a modest set of accounts to prioritise, at a cost you can calculate from pricing. If 5 percent resolve, a free tool may be enough, or you should lean on form-based identification instead.

BusinessMCP notes that the value comes from acting on identified accounts rather than from the percentage. That is the right lens: one correct enterprise account noticed in time can matter more than a hundred noisy ones.

For a view on turning identified accounts into prioritised leads, read lead scoring with IP data, and to try a lookup yourself use BusinessMCP’s IP to company API with its trial credit.

What a healthy report looks like

A good internal summary after a one-week test includes the number of unique addresses, the verdict breakdown, the precision on a sampled set, the share of leakage, the cost per correct answer and a list of mistakes found. Keep it short, date it and rerun it after a quarter, because your traffic mix and the vendor’s data will both have drifted.

Setting a target before you buy

Teams often ask what rate they “should” hit. A better question is what outcome would make the spend worthwhile. Work backwards from a result.

Step	Example reasoning
Value of one good account alert	Estimate from your win rate and average deal size
Alerts needed per month to matter	Pick a figure sales will actually act on
Unique addresses per month	From your logs
Required resolve rate	Alerts needed ÷ unique addresses
Compare with the realistic range	If you need far above the published range, the plan is unrealistic

If the required rate sits inside the realistic band, proceed to a trial. If it sits far above it, either widen the funnel (more traffic, better targeting) or add deterministic sources such as form fills and tracked email links. Match rate is an input to the plan, not a goal in itself.

Segment before you average

A blended rate hides useful differences. Split your sample by country, by landing page and by traffic source, then compare. Paid social traffic often resolves worse than direct or referral visits from partner sites, and documentation traffic often resolves better than blog traffic because of who reads it. Segment-level rates tell you where identification is worth the effort and where it is not, and they make the case to stakeholders far more convincingly than one average.