Quick answer
Account-level analytics, also called group analytics, measures behavior by the organization a person belongs to (a company, team, workspace or store) instead of by the individual. In a B2B product the customer is the account, so twenty people each using a little differs from one person using a lot.
Formula: Health (0–100) = 40 × share of members active + 30 × share of days active + recency (30 if seen within 7 days, 15 within 30 days)
What is account-level analytics?
Account-level analytics, also called group analytics, measures behavior by the organization a person belongs to (a company, a team, a workspace, a store) instead of by the individual. Every event is tied to an account as well as to a user, so you can ask how many accounts are active, which accounts are growing, and which depend on a single person.
The reason it exists is that in a B2B product the customer is the account, not the person. A company that pays for the product has one renewal decision, made by an organization. Person-level numbers cannot tell you whether that organization is still using the product: twenty people each using it a little and one person using it a lot look alike in a total count of events, and they are very different situations.
Product-analytics vendors describe this the same way. Mixpanel lets you count by a group key such as a company ID instead of by user. Segment's group call associates a user with an account-level entity such as a company, organization or team, using a group ID, optional traits, and the user's own ID. A user can belong to several groups.
Person, account and device: which unit to count
| Unit | What one "row" is | Good for | Weak for |
|---|---|---|---|
| Device or visitor | A browser or phone | Traffic, anonymous behavior | Anything about customers |
| Person | A signed-in user | Onboarding, feature use, retention of individuals | How a customer organization is doing |
| Account | A company, team or store | Adoption, health, expansion, churn risk of customers | Behavior of individuals inside it |
You do not replace one unit with another. You pick the unit that matches the decision. "Did this user finish setup?" is a person question. "Is this customer still using us?" is an account question.
When to count accounts: a decision rule
Use three questions before you set anything up.
- Does a paying customer contain several people? If most customers are one person, count people. If a typical customer has five or more users, count accounts as well.
- Is the decision made by the organization? Renewals, seat expansion and churn are account decisions. Onboarding steps and feature discovery are mostly person decisions. Measure each at its own unit.
- Can you identify the account at sign-in? If your product knows which company a user belongs to when they log in, you can tag events with it. If it does not, fix that first: nothing downstream works without an account ID.
Account metrics: what to measure
- Active accounts: accounts with at least one event in a period, such as 30 days.
- Active members per account: distinct signed-in people active in the period.
- Breadth: the share of an account's members who are active. This is the number that separates wide from shallow use.
- Depth: active days or events per active member.
- Recency: how recently anyone in the account was seen.
- Health score: a single 0–100 number combining the above. It is a usage summary, not a prediction; see customer health score.
A worked example: same activity, different accounts
Sample data: over 30 days, two accounts produce a similar amount of activity.
| Account A (Northwind) | Account B (Contoso) | |
|---|---|---|
| Members | 24 | 12 |
| Members active in 30 days | 20 | 1 |
| Share of members active | 20 ÷ 24 = 83.3% | 1 ÷ 12 = 8.3% |
| Events in 30 days | 300 (about 15 per active member) | 320 (all from one person) |
| Active days of 30 | 18 | 27 |
| Last seen | yesterday | today |
Compare event counts alone and B looks as busy as A. Now score each on a 0–100 health scale with this weighting: 40 points for the share of members active, 30 points for the share of days active, and 30 points for recency (30 if seen within 7 days).
- Account A: 83.3% × 40 = 33.3, plus 18 ÷ 30 × 30 = 18, plus 30 = about 81.
- Account B: 8.3% × 40 = 3.3, plus 27 ÷ 30 × 30 = 27, plus 30 = about 60.
B scores lower because its health rests on one person. If that person leaves, goes on leave or changes role, the account goes dark. A looks quieter day by day, but twenty people depend on it.
Choosing the unit also changes funnel and conversion answers. Sample data: 40 accounts with 400 people. 120 people used a feature, so the person-level rate is 120 ÷ 400 = 30%. In 18 of the 40 accounts at least one person used it, an account-level rate of 18 ÷ 40 = 45%. In 9 of the 40 at least three people used it, which is 22.5%. All three figures describe the same behavior; they answer different questions.
Common mistakes
- Using accounts for a consumer product. If there is no organization above the person, group analytics adds work and no information.
- Expecting anonymous visitors in accounts. An account needs identified members. A visitor who never signed in has no account.
- Calling the group before login. The account call has to follow the sign-in, or there is no member to attach.
- Reusing or changing account IDs. One company with two IDs becomes two customers in every report.
- One-person accounts everywhere. If each account has one member, the account view equals the person view. Check typical account size first.
- Reading health as a forecast. A usage score tells you who is active now. It does not know a contract date.
- Mixing accounts with lead handoff. Whether an account is ready for a sales conversation is a different question (see product-qualified lead); the data here is an input, not a verdict.
For how a customer organization moves from first touch to purchase, see the B2B customer journey. For grouping people who share a behavior, the idea is cohort.
Account analytics in TapCub
Take the two accounts above. This is how to get both views.
- Tag each account after login. Once a user signs in, call
tapcub.group('company', 'acme', { name: 'Acme' }). Later events in that session stay on the account. A person can belong to several accounts, and a call before login does not create a member. Servers that know the organization can send the same account field on a server event, and web, app and mini program share an account when they send the same type and ID. - Read the account list. Open All-platform analytics › Users › Accounts. The top cards show account count, accounts active in 30 days, members and account types. Sort by Last active, Members or Health. Open a row to see 30-day activity, top events, hierarchy, members and traits.
- Read health the way TapCub scores it. It is 0–100 and appears only when the account has a last-seen time. The share of members active in 30 days is worth 40 points, the share of active days in those 30 is worth 30, and recency adds 30 if the account was seen within 7 days, 15 within 30 days and nothing after that. That is the formula used in the example.
- Set the hierarchy. In the account drawer you can set a parent account, for a store under a region or a team under a company.
- Switch the unit in analysis. In All-platform analytics › Behavior › Behavior analysis, set the identity to Account for a funnel, retention or event analysis. Rows without an account key are left out of the count. With account identity you cannot open the individual people or add segments to that analysis.
TapCub does not do CRM or lead scoring. It shows account behavior and a usage-based health number; deciding what to do with a customer is outside it.
Sources
- Mixpanel, Group Analytics, counting by a group key such as company ID, checked 2026-11-17.
- Twilio Segment, Spec: Group, the group call, group ID, traits and membership, checked 2026-11-17.
- TapCub documentation, Groups & accounts; Measurement × Behavior; Counting modes, checked 2026-11-17.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Product Analyst
Focuses on event analysis, conversion funnels, and retention, using clear metric definitions and structured analysis to explain product performance.