They were selling to the wrong people
Yettel's prepaid base is large. Their goal was straightforward: get more of those customers onto postpaid plans.
Traditional segmentation couldn't identify which prepaid customers were genuinely close to switching. So the team called broadly. Most conversations went nowhere. The cost of every unproductive call added up.
"Which customers are ready to move, right now?"
Four phases. One permanent change
Find the signal in the data
We started with Yettel's customer data. The goal was to find the patterns that predicted upgrade intent before a customer even knew they had it.
Build and backtest the model
We developed and validated a predictive propensity model using historical campaign data. The model enabled more effective customer prioritization and delivered a measurable uplift in campaign performance.
Run the pilot, then iterate
The model went live in a pilot campaign. High-probability customers went to the tele-sales team first. The results justified scaling.
Runs itself now
We integrated the entire model into Bloomreach. No manual re-runs. No monthly rebuilds. The model now calculates, scores and triggers every month as part of Yettel's standard operations.
Why it works for everyone
Call only the people likely to say yes, and everyone comes out ahead. The tele-sales team stops burning time on dead-end calls and works stronger leads. Yettel earns more revenue for less spend. And customers only hear from Yettel when there's genuinely something in it for them.
Want to see what a model like this could do for your customer base?
Tell us what's not converting.