A common problem is mixing two questions. The first is whether this company is worth pursuing. The second is whether this person is the right one to talk to, and whether they are paying attention. A single score that blends both hides which question failed.
This guide separates the two. It lists what to score at each level, gives a decision flow for when to move from one to the other, and shows how to keep a signal from being counted twice.
Who this is for
This is for B2B teams that sell to a buying committee: several people at one company, each with a different role in the decision. It is most useful if:
- You already have an ideal customer profile, even a rough one. If not, start with How to Define an Ideal Customer Profile.
- You score leads today, or have read about lead scoring, and find that one enthusiastic contact at a poor-fit company keeps reaching your reps.
- You run, or are considering, account-based marketing (ABM) and need criteria for deciding which accounts get attention.
If you sell to individuals with no company behind them, account-level scoring has little to add.
The core difference
An account is a company. A lead is a person at that company. The two levels differ in what they describe and how fast they change.
| Account level | Lead level | |
|---|---|---|
| Question | Should we pursue this company? | Should we talk to this person now? |
| Describes | Fit, size, stack, trigger events | Role, authority, engagement |
| Changes | Slowly, in steps | Often, with each reply or click |
| Owner | Account owner | Whoever works the lead |
| Fails when | Company is outside your ICP | Person has no say or no interest |
Account-level facts are shared by everyone at the company. If the company has the wrong size or stack, every contact there inherits the problem. Lead-level facts belong to one person. A strong lead at a weak account is still a weak opportunity, and a weak lead at a strong account is a reason to find someone else there.
What belongs at account level
Score the company on things that stay true no matter who you contact.
Fit. Firmographics such as industry, country, size and domain. A CSV import can map your own columns to domain, name, country, size and industry, along with deal fields such as stage, amount, outcome and loss reason [[1]](#sources). Your CRM holds the same kind of data. HubSpot gives company name, domain, owner and lifecycle stage, plus deal pipeline, stage, amount and closed-lost reason [[2]](#sources). Salesforce gives accounts, opportunities and their stages [[3]](#sources).
Stack. The tools the company already uses, where they bear on whether your product fits. This is a company fact, not a personal one.
Trigger events. Changes at the company that can open a buying window. Vendors in this space describe signals such as job changes, funding rounds, executive hires, tech stack shifts, website visitors and G2 buying signals [[4]](#sources). Another lists hiring, technology use, company growth, LinkedIn engagement and competitor following [[5]](#sources). These are vendor descriptions of what they track, not measurements of how well it works. Pick the triggers that connect to a real reason your customers bought.
Competitor activity can also sit here. Outlevo's competitor tracking includes competitor signals on accounts [[6]](#sources).
Company-level engagement. Some engagement belongs to the company, not a person. Website-visitor tracking is one case. Outlevo shows which companies visit, which pages they read and whether they fit your ICP, and it never identifies a person [[7]](#sources). That is an account signal by construction. It costs 1 credit per company identified per day, however many pages that company reads [[7]](#sources).
Data quality. Score how much you actually know. A company with a strong fit score built on thin evidence should not be treated like one built on solid evidence. Outlevo's Readiness Card carries three scores: fit, intent and data quality [[8]](#sources).
What belongs at lead level
Score the person on things that describe them and their behavior toward you.
Role. Job title and function, and where the person sits in the committee. In HubSpot, job title comes with the optional contact import [[2]](#sources).
Authority. Whether this person can decide, influence, block or only use the product. Score authority at the lead level, because two contacts at the same company can have the same account score and very different authority.
Engagement. What this person did: replied, clicked, accepted an invitation, booked a meeting. Replies and bookings are individual events. Outlevo reads LinkedIn acceptances and replies, so follow-ups wait for an acceptance and a reply stops the sequence [[9]](#sources). A booking through Calendly counts as a meeting booked [[10]](#sources).
Reachability. Whether you can and may contact this person: a verified email, no opt-out. HubSpot's email opt-out field comes with the optional contact import [[2]](#sources).
Where the line gets blurry
Some signals touch both levels.
- A job change concerns one person but can also point to a new decision-maker at the account. Record the person-level fact on the lead and the account-level implication separately.
- An executive hire is the same event seen from the company side. Score it once, at account level, as a trigger.
- Committee coverage is the number of distinct roles you have reached or engaged at one company. It is an account-level view built from lead-level events. Count distinct roles, not the sum of each person's score.
The rule: ask what the signal is about. If it would be true whichever employee you contacted, it is an account signal. If it changes when you swap one contact for another, it is a lead signal.
Steps
This is a decision flow. It does not give point weights or score cutoffs. Set those yourself from your own closed deals, and revisit them as you learn.
1. Start at the account. Check fit first. If the company is outside your ICP, stop. No lead-level signal should rescue it, and no contact data should be bought for it.
2. Check the evidence behind the fit. A fit score is only as good as what is under it. Outlevo's Readiness Card needs at least three observations, each with a source and a date [[8]](#sources). Facts read from a website are kept only when their quote is found on the page, and guesses are labeled as hypotheses, never findings [[8]](#sources). Apply the same standard to your own scoring: a "fit" based on a guess should be marked as a guess.
3. Look for a trigger or company-level engagement. Fit tells you who could buy. A trigger or a visit tells you whether now is a reasonable time. Keep this at account level.
4. Decide whether the account clears your bar. Use your own threshold. Accounts that fit but show no trigger can wait in a lower-attention group. They do not need to be dropped.
5. Only now find people. Move to lead level once the account passes. Spend on person-level work after the account decision, not before. Contact discovery costs 4 credits [[11]](#sources), an extra cost you avoid for accounts that fail fit.
6. Score each person on role, authority and engagement. Aim to cover more than one role at the company. A single contact is a single point of failure in a committee sale.
7. Switch back to the account when the lead stalls. If your one contact goes quiet or leaves, the account is still qualified. Find another person instead of rescoring the company.
8. Route by ownership. Decide who acts on a hot lead. Outlevo routes hot leads to the company owner in HubSpot and to the account owner in Salesforce [[2]](#sources) [[3]](#sources). Account ownership decides who acts on a person-level signal, so keep ownership clean before you turn on routing.
A short version of the switch: stay at account level until the company clears your bar, move to lead level to choose people and timing, and return to account level when the people run out.
How to avoid double-counting
Double-counting is a common way for a two-level model to go wrong. It happens when one real-world fact is added to several scores.
Give each signal one home. Write a short table of every signal you use and mark it "account" or "lead." A signal that appears in both columns is a bug.
Let leads inherit, not re-earn. A contact at a strong account can display the account's status as a read-only label. Do not add the account's fit points to each lead's score. If five contacts each receive the company's fit points, a company with a large committee looks five times better than one with a single known contact.
Don't add a company visit to each person. A company-level website visit is one event. If you copy it onto every known contact at that company, it becomes many events.
Don't turn lead activity into account intent without a rule. Three replies from three people at one company is real committee interest. Three clicks from one person's inbox is not three signals. Roll up by distinct people or roles, and say so in your rules.
Keep audiences at their own level. Ad audiences show the split clearly. LinkedIn Matched Audiences are built from company names and websites, not from people [[12]](#sources). Google Ads Customer Match uses SHA-256 hashes of verified email addresses [[13]](#sources), and only contacts with a verified address that pass the ads eligibility check are uploaded [[13]](#sources). The first is an account-level audience and the second a person-level one. Don't treat membership in both as two separate pieces of evidence about fit.
Count companies when you tune. Outlevo's win/loss analysis counts companies, not deals, so one account with several deals counts once [[14]](#sources). Do the same when you check whether a fit criterion predicts wins. A company that signed three contracts is one company.
Respect small samples. Under 10 won companies is labeled very low confidence in Outlevo's analysis [[14]](#sources). If you have only a few wins, treat any weighting you derive as a hypothesis. Outlevo labels AI-proposed ICP segments as experimental until a person approves them [[14]](#sources). A good habit for any scoring model is to label what has not been confirmed.
How this looks in Outlevo
Outlevo qualifies at account level. Each qualified account gets a Readiness Card with fit, intent and data quality scores [[8]](#sources). With HubSpot connected, Outlevo writes status, fit and intent scores, ICP version, reason, next action and due date to the company record, in fields it owns [[2]](#sources). Your standard HubSpot fields are not edited, and nothing is written until an owner or admin approves the field mapping [[2]](#sources).
Person-level data is a separate step. HubSpot contact import is optional and happens only after an admin turns it on with a stated purpose [[2]](#sources). That fits the two-level model: the company decision comes first, and the decision to bring in personal data is deliberate.
In Salesforce, Outlevo writes Outlevo fields on the Account, a Readiness Card record and a Task for the next step [[3]](#sources). A rep's Accept or Reject on the Account is read by the 15-minute sync, because Salesforce sends no webhooks to Outlevo [[3]](#sources). The Salesforce connection is a pilot, and the package is installed with you [[3]](#sources).
For the full account-level process, see How to Qualify B2B Accounts and the account qualification page.
What to check
Before you trust a two-level model, run through these questions.
- Does every signal have one home? Check your table for any signal in both columns.
- Can a lead score rise because of the account? If so, you are re-adding account points to people.
- Does one visit become many events? Look at how company-level engagement is attached to contacts.
- Is the fit based on evidence? Each account-level fact should have a source and a date. Guesses should be marked as guesses.
- Are you tuning on companies? Count companies, not deals or contacts, when you test a criterion.
- Is the sample large enough? With few won companies, label your weights as provisional.
- Do you cover the committee? An account with one engaged contact and no other roles is thin, whatever the lead score says.
- Who owns the account? Routing a hot lead only works if the owner field is right.
- Is contacting the person permitted? Company-level targeting does not remove other duties. The UK ICO says: "You can email or text any corporate body (a company, Scottish partnership, limited liability partnership or government body)." [[15]](#sources) The FTC says: "The law makes no exception for business-to-business email." [[16]](#sources) Check the rules for the places where your recipients are, and ask your counsel if unsure.
Next step
Take your current scoring rules and sort each one into "account" or "lead." Remove any rule that sits in both. Then write one sentence for each account you want to pursue: the fit evidence, the trigger, and the roles you have yet to reach.
If you want a starting point for the account side, the free ICP assessment reads up to 5 public pages of a website and returns facts labeled as verified observations or AI inferences, plus up to 5 experimental segment hypotheses [[17]](#sources). It asks which countries you sell into [[18]](#sources). If you have past deals, you can import them from CSV or your CRM and get a win/loss analysis that proposes ICP hypotheses [[19]](#sources).
If you are choosing data tools to feed account-level signals, see data and enrichment tools: pricing compared. To add competitor signals to your account criteria, see competitor tracking software for B2B.
Sources
- CSV and Excel integration
- HubSpot integration
- Salesforce integration
- 11x signals and triggers
- Reply.io Jason AI
- Release 2026-10-19
- Website visitors integration
- Readiness Cards
- LinkedIn integration
- Calendly integration
- Outlevo pricing
- LinkedIn Ads integration
- Google Ads integration
- ICP from evidence
- ICO: Electronic mail marketing (PECR)
- FTC: CAN-SPAM Act compliance guide for business
- Free ICP assessment
- Release 2026-10-19
- Release 2026-10-19