Subscribers & churn
One subscriber definition across billing, CRM and provisioning, with suspends and downgrades ruled on once.
Industry · Telecom
Churn that matches the bill. Get your Telecom metrics right.
Telus trusts their semantic computations to AtScale. AtScale defines each telecom metric once, so AI assistants, agents and BI tools give marketing and finance the same number.
When someone asks about churn, Ava (Ask, Verify, Answer) checks which version they mean. ACE (the AI Computation Engine) then calculates it the same way in every tool. Below are six places telecom teams use AtScale, followed by one metric traced from the question to the agent that acts on it.
Use cases
Where billing, network and marketing count the same subscriber.
In each one, two teams can calculate the same metric in two reasonable ways and get two different answers. AtScale settles the definition once, and every person, tool and agent uses it.
Network performance
Counter data from 3G, 4G and 5G hardware, comparable across vendors.
Customer experience
Network events tied to the problems customers report, so support and engineering trace one root cause.
Self-service analytics
Departments reach the metrics directly instead of queuing at a central data team.
Cost optimization
Automated aggregates cut query cost on every platform the queries land on.
Governance & security
Role-based access at the metric layer, so regional and vendor systems stop carrying their own copy of the logic.
Churn, from question to agent.
A suspended line still gets a bill, but your churn number says it left.
Each team defines churn a little differently, and the subscriber base it’s measured against changes every month. Neither problem shows up until finance checks the numbers against billing.
Churn analysts don’t always prompt what they mean.
People type short questions. Before ACE computes anything, Ava (Ask, Verify, Answer) works out which metric and which rules each question means.
What a person types | What they meant |
|---|---|
What a person types “What was churn last quarter?” | What they meant Churn computed across the quarter against the average subscriber base, suspends left out. |
What a person types “Which plans are churning?” | What they meant Churn by plan and tenure cohort, each cohort against its own base. |
What a person types “What’s the margin on this plan?” | What they meant Plan margin with the handset subsidy loaded in. |
How AI agents act on telecom churn, and when a person signs off
When agent ask questions, AtScale checks metric definitions the business already confirmed and knows how to calculate. Once a metric is trusted, agents can fully automate decisions the agent takes on its own, or ensure the humans step in to sign off. Here are three examples in telecom:
agentic workflow | Atscale calculation | Who acts |
|---|---|---|
agentic workflow At quarter close, run an agent to “Report new customer churn to finance compared to last quarter.” | Atscale calculation Churn across the quarter against a specified subscriber base, suspended accounts left out. | Who acts Finance approves. The agent drafts a report. Finance validates, signs off, edits, and builds the narrative about the impact on the business. |
agentic workflow Every day, run an agent to “Send a retention offer to any plan cohort where churn is rising.” | Atscale calculation Churn by plan and tenure cohort, each cohort against its own baseline. | Who acts Fully automated. The agent sends offers on its own, capped per customer, according to specified thresholds, considering promotions. |
agentic workflow Every week, run an agent to “Flag any plan whose margin went negative.” | Atscale calculation Plan margin with handset subsidies loaded in. | Who acts Pricing approves. The AI agent flags negative margin plan. Analysts decide whether to change the rate card. |
Proof points from this sector
Case StudiesTelus
Case StudiesCloud-Based OLAP Transformation: How Vodafone Portugal, Celfocus, and AtScale Modernized Analytics
White Papers & Reports