Your numbers shouldrun on business rules,not on Excel

Semantic layers sit on top ofthe logicSemantic source is the logic

Book a demo

Client size

Avery Morgan

Retention

Dana Reyes

CRM.CLIENTS

BILLING.PAYMENTS

churn_by_size

Churn rate by client size

Micro clients churn the mostSmall clients churn the mostMedium clients churn the mostLarge clients churn the most

0.0%

Micro

5.8%

Small

2.1%

Medium

0.9%

Large

Avery MorganSales

No change goes unexplained.

You can use our agent,
your Claude or ChatGPT.

CreditUnit economics

Token cost per active user

$9.78−2%vs Sep

Exclude internal accounts from UsersApplied

Authors

Priya ShahReviewers

Mateo Silva1

Nadia Brooks1

Account metricsChanged by

Priya ShahonOct 6

Users

Humans who logged in this month, counted once per email, everyone included except employees and test accounts.

Untitled scenarioDraftPreview

Account metricsEdited

UsersHumans who logged in this month, counted once per email, except employees and test accounts.

CurrentScenario

Token cost per active user

$9.78

Token cost per active user

$9.78$10.32$10.90+6%+11%

Token cost

$8,850

Token cost

$8,850$9,700$10,950+10%+24%

Active users

905

Active users

9059401,005+4%+11%

+5 other affected results

Preview the effect on results

Why did Token cost per active user drop?

PolaxingThinkingHerding definitions

It fell 2% due to changes in definition, not data.

Metric = \frac{Token cost}{Active users}

Users definition change in Account metrics
−25% token cost, −14% active users.

Active users definition change in Engagement metrics
+5% active users.

Token cost: +18% in sessions.

With the same definitions, it rose 22%.

Would you like to simulate different scenarios?

Yes, please figure out the isolated effects of counting employees and test accounts separately.

PolaxingOn it. I'll set up a Scenario.

Create scenario

Add context

Edit definition

Preview results

4 tools with Polaxys

Counting test accounts again: $10.32 per active user, up 6%.

Counting employees again: $10.90 per active user, up 11%. Employees move it most.

Unit cost drop

Why did Token cost per active user drop?

Polaxing

It fell 2% due to changes in definition, not data.

Metric = \frac{Token cost}{Active users}

Users definition change in Account metrics
−25% token cost, −14% active users.

Active users definition change in Engagement metrics
+5% active users.

Token cost: +18% in sessions.

With the same definitions, it rose 22%.

Would you like to simulate different scenarios?

Yes, please figure out the isolated effects of counting employees and test accounts separately.

PolaxingOn it. I'll set up a Scenario.

Create scenario

Add context

Edit definition

Preview results

4 tools with Polaxys

Counting test accounts again: $10.32 per active user, up 6%.

Counting employees again: $10.90 per active user, up 11%. Employees move it most.

Ask about your data...

Claude

Why did Token cost per active user drop?

Clauding

It fell 2% due to changes in definition, not data.

Metric = \frac{Token cost}{Active users}

Users definition change in Account metrics
−25% token cost, −14% active users.

Active users definition change in Engagement metrics
+5% active users.

Token cost: +18% in sessions.

With the same definitions, it rose 22%.

Would you like to simulate different scenarios?

Yes, please figure out the isolated effects of counting employees and test accounts separately.

PolaxingOn it. I'll set up a Scenario.

Create scenario

Add context

Edit definition

Preview results

4 tools with Polaxys

Counting test accounts again: $10.32 per active user, up 6%.

Counting employees again: $10.90 per active user, up 11%. Employees move it most.

Reply to Claude…

ChatGPT

Why did Token cost per active user drop?

Codexing

It fell 2% due to changes in definition, not data.

Metric = \frac{Token cost}{Active users}

Users definition change in Account metrics
−25% token cost, −14% active users.

Active users definition change in Engagement metrics
+5% active users.

Token cost: +18% in sessions.

With the same definitions, it rose 22%.

Would you like to simulate different scenarios?

Yes, please figure out the isolated effects of counting employees and test accounts separately.

PolaxingOn it. I'll set up a Scenario.

Create scenario

Add context

Edit definition

Preview results

4 tools with Polaxys

Counting test accounts again: $10.32 per active user, up 6%.

Counting employees again: $10.90 per active user, up 11%. Employees move it most.

Ask anything

Unit cost drop

Why did Token cost per active user drop?

Polaxing

It fell 2% due to changes in definition, not data.

Metric = \frac{Token cost}{Active users}

Users definition change in Account metrics
−25% token cost, −14% active users.

Active users definition change in Engagement metrics
+5% active users.

Token cost: +18% in sessions.

With the same definitions, it rose 22%.

Would you like to simulate different scenarios?

Yes, please figure out the isolated effects of counting employees and test accounts separately.

PolaxingOn it. I'll set up a Scenario.

Create scenario

Add context

Edit definition

Preview results

4 tools with Polaxys

Counting test accounts again: $10.32 per active user, up 6%.

Counting employees again: $10.90 per active user, up 11%. Employees move it most.

Ask about your data...

Jordan LeeCOO

Enterprise-ready

We operate in highly and . Built with security and compliance from the start.

Native to your warehouseYour data, roles, policies, etc.
- Snowflake
- Databricks
- BigQuery
- and more
SSO readyOkta, Microsoft Entra ID or Google Workspace.
- Okta
- Microsoft
- Google Workspace
- and more

Proud to be a part of

for StartupsGoogle for Startups
for StartupsOpenAI for Startups
StartupsClaude Startups

Resident

EU data residencyStored and processed in the EU.
US data residencyStored and processed in the US.

Audited

SOC 2 Type IIndependently audited.
SOC 2 Type II · In progressAudit under way.In progress
GDPR-readyBuilt to fit your GDPR setup.

Private

Zero data retention (ZDR)AI providers keep nothing.
No AI trainingYour data is not used for training.
EncryptionAt rest and in transit.

Letter from the founders

The numbers don’t agree, and each team defines the business a little differently. Almost every company has seen a version of this.

This problem is hard to diagnose, partly because it doesn’t look like a problem. It looks like work.

Sales changed how it defines an active user, for a good reason. The growth team means something else by “big client”, also for a good reason. Meanwhile, the data team’s pipelines still run on last month’s rules, and engineers are catching up. All of it feels like part of the job description.

But it doesn’t have to be this way.

Engineers solved this problem for code some time ago. Their tools make sure that one person’s change doesn’t break another person’s work. Team boundaries can be expressed in code. The right people review each change before it ships. Changes are tested together and rolled back if something goes wrong. The whole system moves forward in lockstep.

The rest of the company never had these tools. How could it? Computers couldn’t read, and speaking computer language was a career of its own. Neither is true anymore.

Companies should spend less time getting everyone on the same page and more time discussing the business and making better decisions.

We’re building this future.

PS: Replace “person” with “agent” and it becomes clear why coding moved first and fast. The foundation for intelligence was already there.

About us

Rafael Martins worked at Goldman Sachs, and was VP of Finance and Investor Relations at StoneCo, where he led its NASDAQ IPO.

Bernardo Rufino spent a career building software at Microsoft, Meta and Google, where he worked on Gemini and AI.

Rafael MartinsCo-founder & CEO
Bernardo RufinoCo-founder & CTO

Your numbers shouldrun on business rules,not on ExcelExcelSQLPython

No change goes unexplained.

Enterprise-ready

Proud to be a part of

Resident

Audited

Private

Letter from the founders

About us

Your numbers shouldrun on business rules,not on Excel