How it works

How the Scout confidence score works

Last checked 5 min read

Scout gives every identified company a confidence score from 1 to 99, a band (high, medium or low) and a one-sentence reason. The score tells you how sure we are that the visit really came from that company, so you can decide who to contact first and which matches to treat with caution.

Why match accuracy matters

Identifying a company from a website visit is an inference, not a fact, so a tool that shows every match with the same certainty will mislead you some of the time. A score lets you act quickly on strong matches and check weaker ones before you spend time on them.

Visitor identification works by looking up the IP address of a visit and matching it to an organisation. That works well on a company's own network, and less well from home, a phone, a VPN or a shared office, where the IP address belongs to a provider rather than the visitor's employer. If a tool quietly counts those visits as companies, your list fills with names that never visited.

We would rather tell you when we are unsure. That is the reason for the score.

What goes into the score

The score starts from the strength of the underlying IP-to-company match, then goes up or down depending on the kind of network the visit came from and what the visitor did on your site.

Scout looks at these signals for each match:

Signal What it tells us Effect on the score
IP-to-company match strength How strongly the IP address matches a known company The starting point. A modest baseline is used if no strength is available
Network type Whether the IP belongs to a corporate network, or to an ISP, home broadband, mobile network, VPN or cloud host A corporate network adds to the score. Other network types reduce it sharply
Pages viewed How many pages the visitor looked at in the session Each page after the first adds a little, up to a limit
Google Analytics 4 session How engaged the session was, if you have connected GA4 Adds up to a set amount for engaged sessions
Google Search Console Whether a search query can be associated with the visit, if you have connected Search Console Adds to the score when present

You don't need GA4 or Search Console for Scout to work. Connecting them gives the score more evidence to use.

The three bands

Scores fall into three bands, so you can sort and filter without reading every number.

  • High (70 to 99). A strong match. The visit is very likely to be from this company, usually because it came from a corporate network and the visitor looked at several pages, or search and analytics data back it up.
  • Medium (40 to 69). A likely match with some weaker signals. For example, a single-page visit, or a network we can't classify with certainty.
  • Low (1 to 39). There is some basis for the match, but the signals are weak. Treat it as a lead worth checking, not a confirmed visitor.

The honesty rules

Some kinds of traffic are capped in the low band whatever else is true, because they are the most common cause of false matches in visitor identification.

  • ISP, home broadband, mobile, VPN and data-centre traffic is flagged. A match from one of these networks can never score above the low band. You will see the flag and the reason, for example "the visit came from a consumer ISP, not a company network".
  • Network-name guesses are capped too. When the only evidence is the name of the organisation that owns the network, rather than a direct match, the score stays in the low band.
  • Bots are excluded. Automated traffic is scored at zero and not counted as a company.
  • Nothing scores 100. The maximum is 99, because an IP match is never certain.

"Why this match?"

Every match comes with a short explanation written from the signals, so you can see the reasoning without opening a spreadsheet. A strong match might read "High confidence: matched this company on a corporate network and 6 pages viewed." A flagged one might read "Low confidence: the visit came from a mobile carrier, not a company network, so the match is flagged as a likely false positive."

Click any company in Scout to see the full breakdown: which signals contributed, by how much, and which version of the scoring model produced the score.

A versioned, rules-based model

The score comes from a fixed set of rules, not a black box, and each score records which version of the rules produced it. When we improve the rules, past scores are not rewritten. That means the numbers you acted on last month still mean what they meant then.

Using the score

Start with high-confidence companies that fit your ideal customer profile, check the pages viewed and the reason before contacting medium or low matches, and push only the matches you trust to HubSpot.

To learn how IP-to-company identification works in general, and where its limits are, read our guide to website visitor identification. If you are comparing tools, see our Lead Forensics alternatives page, or start a 14-day free trial and see the scores on your own traffic. Plans and credits are on the pricing page.

Frequently asked questions

What does the Scout confidence score mean? It is a number from 1 to 99 showing how sure Scout is that a visit really came from the company it has matched. 70 and above is high, 40 to 69 is medium and below 40 is low. Every score comes with a one-line reason.

Why is the score never 100? Because identifying a company from an IP address is always an inference. Scout caps the score at 99 so no match is presented as certain.

Why are some matches marked low confidence? Usually because the visit came from a network that does not reliably belong to one company, such as home broadband, a mobile network, a VPN or a cloud provider. Scout keeps these matches visible but holds them in the low band.

Can a confidence score change? The score is calculated when a visit is matched and is stamped with the version of the scoring model used. If we improve the model, older scores keep their original version so you can see which rules produced them.

Sources

Information checked 25 September 2026. This page describes Scout's scoring model as of that date (model version rules-v3).