There's a pattern I watch out for across the B2B accounts I worked with at Mixpanel: measuring the right things for the wrong entity. The most common problem here is that teams tend to want to dive deep into user insights when many of the answers to their product or customer health questions are best found with group or account analytics.
Take user retention. A Mixpanel account with 25 individual users where more than half have gone dark looks like high churn—until you realize their analytics project is still as active as ever, with tons of new reports and queries coming in. That gap is exactly the problem. The metric looked alarming for the wrong reason.
Tracking individual behavior (activation, retention, funnel drop-off, and feature engagement) is foundational. But in B2B, user behavior and account health are different things. They don't always move together. Sometimes they move in opposite directions. A team looking only at users won't see it coming.
Here are five places where B2B teams limit their insights by not keeping user-level and account-level analytics equally top of mind.
User activity and account activity tell different stories and can point in opposite directions
Monthly active users is a hard metric to interpret in isolation for a B2B business. What matters the most is whether accounts are healthy according to the dimensions that drive revenue: ARR band, contract tier, industry, and company size. User insights play a part in this, but instead of over-indexing on questions like, "What percentage of users came back?" you need to also ask, "Which of our customer businesses are retaining?" Those answers are often very different.
Back to the example teased at the top, an account that started with a set number of active users (or seats) can show a drop in that number and still maintain a healthy account status. You just have to keep an eye on the aggregate activity in the group of users that makes up the account. Maybe some restructuring at the account consolidated their activity to fewer seats, and as long as the account is still active enough to be getting value from your product, that’s the top-line takeaway.
Power users make at-risk accounts look healthy
Of course, the inverse can be true, too. An account can look active, with engagement metrics that look fine. Then you look at the distribution and it's two people doing everything.
The CEO and founder are running a million queries and reports. That account looks healthy until it isn't. One job change, one reorg, and there's nothing holding it together. It's what I'd call brittle.
Account analytics doesn't replace user analytics; it gives you the context to interpret it correctly. You need both: account-level totals and user-level distribution. One tells you whether the number looks right; the other tells you whether it's sustainable.
The goal isn't to be suspicious of power users. Every product has a frequency distribution, and some users will always engage more than others. What matters is what a healthy distribution looks like for your specific product. For some accounts I work with, a healthy account could mean 25 value actions per week, with individual users averaging around five. If an account hits that total with two people doing all of it, that's the signal. Volume alone doesn't surface it.
The right unit of analysis depends on what you're asking
In some B2B scenarios, especially in B2B2C products, "track users" can also become a limited approach because, for many important questions, the relevant entity isn't a person.
A food delivery platform is a useful example. Merchants have different workers on every shift. What matters isn't which individual worker took an order. It's that the store is updating its catalog, responding to orders, completing the key actions that make the platform work. Some platforms set the store ID as the user ID. Individual user identity is almost irrelevant for the core product question.
Level 1
Individual user
The person taking actions
Question it answers
What did specific users do? Where did they drop off? Who activated?
Examples
Activation funnels Feature engagement User retention curves
Watch out for
User churn ≠ ARR churn. Power users can make fragile accounts look healthy.
Level 2
Sub-account entity
Workspace, project, team, API key
Question it answers
Is this workspace active? Which projects are healthy? Is this API key doing the right things?
Examples
Project activity API key usage Store-level behavior
Watch out for
An account can look stable overall while 4 of 5 workspaces have gone dark.
Level 3
Company account
The customer driving revenue
Question it answers
Is this account healthy? Which segments are at risk? Where are expansion signals accumulating?
Examples
Account retention Churn signals Expansion readiness
Watch out for
This is the level that drives ARR. Missing it means missing the signals that matter most to the business.
Developer-facing products work the same way. A company with an AI product tracks API key activity (actions per key) because that's the entity that reflects actual usage. An AI lab we work with uses exactly this approach. They care that the key is active and the right sequences of actions are being taken, not which individual engineer triggered any particular call.
We use this internally at Mixpanel. Since plenty of our B2B accounts have separate Mixpanel instances (projects) for their different-sized and different-shaped apps or product efforts, we often rely on Group Analytics to track project ID activity (instead of user- or even account-level) to learn where our own infrastructure performance can be improved.
Learn more by segmenting account- and group-level behavior by properties. Is churn concentrated in your SMB segment? Mid-market? Which specific accounts are showing early behavioral signals right now? This turns a reactive problem into a proactive one—from "why did this account churn?" to "which types of accounts are heading in that direction?"
Your product roadmap can end up optimized for active users instead of valuable accounts
When product teams prioritize by feature engagement, they're measuring usage volume: which users touch this feature, and how often. That's a useful input. But in B2B, it can systematically mislead you about what's holding your most valuable customer relationships together.
The features driving the highest individual usage aren't necessarily the ones your enterprise accounts depend on. A product team can ship an entire quarter's worth of work that moves engagement metrics without touching a single enterprise renewal conversation.
There's also a subtler version: building for users who are already active while neglecting features that help more users reach that level of engagement. Onboarding, team setup, and the moments that make the product feel essential across an org. Both sides matter.
Again, account-level or group-level property data surfaces these patterns. The Mixpanel example from the previous section is instructive: When we looked at which project types were suffering the worst query performance, we found consistent patterns in group-level properties attached to account projects—specific configurations struggling in ways that weren't visible from user behavior alone. That drove targeted product investment, measured at the account level.
The practical test: "Did the features we shipped for fintech customers actually land with fintech customers?" You can only answer that if you can filter feature adoption by industry, ARR band, or contract tier—account-level properties, not user-level ones.
Expansion signals accumulate at the account level, not the user level
The signals that precede expansion (more users, a higher tier, a new module) are account-level patterns. They don't surface cleanly from individual user data.
Hitting usage limits is the clearest signal. If an account uses all 20,000 of their monthly session replay credits, that's visible at the account level. They've maxed out and almost certainly need more. At the user level, you'd see several people viewing a lot of replays, but not the account-wide picture that tells you they're at capacity.
Subtler signals accumulate over time, too. Sometimes, accounts expand from one use case to several or they activate features at a rate that outpaces their user count. Accounts can even change their entire behavioral profile to resemble other accounts right before they expand.
Over time, you can refine the playbook. Compare current accounts to historical expansion patterns, identify the leading indicators that most reliably preceded expansion, and move earlier. Account-level groupings give you a sharper signal than any fixed rule.
Don't limit your insights to part of the picture
User analytics is foundational. Every pattern I've described here builds on it. The shift is toward also measuring accounts, workspaces, and the other sub-entities that reflect how customers actually use your product as a business, alongside the users you already track.
What the teams I work with who do this well have in common: They've made it structural. Account analytics is baked into how they prioritize roadmap work, how they measure feature launches, how they identify expansion opportunities before they become obvious—not a one-off analysis they run before a renewal or when something goes wrong.
If you want to see what this looks like in practice, you can learn more about Mixpanel's Account Analytics and Group Analytics here.
Analytics for everyone.
Peter is a customer value architect at Mixpanel. He works with different teams and projects for large enterprise accounts on feature implementation and enabling use cases.












