Lead Scoring with IP Data: Turning Network Signals into Priority

Lead scoring with IP data works at account level: resolve the address, keep only business-network verdicts, then add points for company fit, repeat visits and high-intent pages. Never score an ISP, a carrier or a cloud host, and never treat the result as a named individual.

What IP data adds to a score

A scoring model usually mixes three families of signals.

Family	Examples	Where IP data helps
Fit	Industry, size, region, tech stack	The resolved company and its domain let you look these up
Behaviour	Pages viewed, repeat visits, time on site	IP groups sessions into an account when no one has identified themselves
Engagement	Form fills, replies, meetings	Email domains confirm the account; IP is secondary

IP data supplies the account key for anonymous activity. Without it, a prospect reading your pricing page three times from the office is invisible until they fill in a form, and BusinessMCP’s guide on visitor identification notes about 98 percent of visitors never do.

Gate first, then score

The most important rule is ordering.

1. Resolve the address.
2. Gate on verdict. Only company rows proceed. ISP, mobile, hosting, VPN and unidentified rows get no account score.
3. Enrich the domain with firmographics from a company data source. BusinessMCP’s company data API is one option for that step, and its IP to company API handles the first.
4. Score the account.

Key takeaway: a scoring model that skips the gate will rank internet providers at the top, because they generate more visits than any real customer.

A starter model

Treat these numbers as placeholders to be tuned against your own win data, not as benchmarks.

Signal	Points	Note
Company fits ICP (industry, size)	+20	From enrichment
Pricing or docs page viewed	+15	Evaluation behaviour
Comparison page viewed	+10	Shortlisting
Second visit on a different day	+10	Sustained interest
Three or more pages in a session	+5	Engagement
Already a customer	Route elsewhere	Do not alert as a lead
Competitor or agency domain	Exclude	Maintain a list
Blog-only single visit	+0	Not worth sales time

Set a threshold, for example alert when an account passes a total you choose, and review the alert volume weekly. If sales ignores alerts, the threshold is too low.

Time decay

Interest fades. A simple approach: halve the behaviour points every 14 days without new activity. This keeps stale accounts from outranking fresh ones. The right half-life depends on your sales cycle.

Calibrate against outcomes

After a month or two, compare:

• Accounts that scored high and became opportunities.
• Accounts that scored high and went nowhere.
• Opportunities that never scored, meaning they came from unresolvable traffic.

Adjust points to match. The last group is a reminder that IP resolution covers a minority of traffic, so scoring built on it will always miss accounts. See IP to company match rate for why.

Combining with deterministic signals

When a contact submits a form, the email domain confirms the account and can be matched to earlier anonymous sessions. Treat this as higher confidence than an IP match and let it override. Also reconcile conflicts: if the IP says Company A but the email domain says Company B, trust the email, and consider that the visitor may have been on a third party’s network.

Alerting without being creepy

How you act matters as much as what you score.

• Alert the account owner with the pages read and the date, not a stream of raw visits.
• Reach out to roles that make sense for the account, not to guessed individuals.
• Keep language neutral; never imply you watched a specific person.
• Respect opt-outs and your privacy policy, as covered in GDPR and IP lookup.

Data hygiene

• Re-resolve addresses every few days so ranges that changed owner do not produce stale accounts.
• Store the verdict and class with each score so the reasoning can be audited.
• Exclude your own and partners’ networks. See identifying companies visiting your website for the full setup.
• Run the audit from how to test an IP to company API each quarter on the scored accounts.

Cost sense

Scoring does not need every visit resolved. Resolve unique addresses once per day, skip filtered infrastructure and cache results. IP to company API pricing shows how to estimate the bill, and bulk lookups covers the batch route.

A model like this will not find every buyer. It will help a small team spend its first call of the week on the account that read the pricing page twice, and that is a realistic, valuable promise.

Worked example

Take two accounts seen in one week, both resolved to company verdicts.

Signal	Account A	Account B
ICP fit	Yes (+20)	Yes (+20)
Pricing page viewed	Twice (+15)	No
Comparison page viewed	Once (+10)	No
Visits on different days	Two (+10)	One
Pages in best session	Four (+5)	One
Total before decay	60	20

Account A goes to the account owner today with the pages read and dates. Account B stays on a watch list. If a contact from Account B later fills a form, the email domain confirms the account and its score rises with the new evidence. Notice what is missing from both rows: any attempt to say who the individual was. The system scores accounts, and people are identified only when they choose to identify themselves.

Common failure modes

• Scoring volume rather than intent. A noisy network with thousands of visits outranks a quiet prospect. Cap per-account points per day.
• Never removing customers. Existing customers browse docs constantly. Route them to account management instead.
• Letting competitors in. Competitor and agency networks look highly engaged. Maintain an exclusion list.
• Ignoring decay. Without decay, last quarter’s visitor still tops the list.
• Trusting one signal. Use at least two independent signals before alerting.