App growthCase studiesServicesMeasurementBlogAboutBook a call
← All case studies
Atom Mobility · B2B SaaS, mobility tech

How we built an AI lead prediction model and cut Atom Mobility's CPA by 56%

Atom Mobility's Google Ads brought in plenty of demo requests, but only about one in ten was sales-ready, and the long B2B sales cycle kept Google Ads from learning which ones. Our account managers and data scientist built a model that scores every lead within 24 hours.

  • −56%
    Cost per acquisition
  • +54%
    Lead quality and conversion signals
  • 24h
    Lead scoring, down from weeks
Atom Mobility platform
THE PROBLEM

Plenty of leads, almost none of them real

Atom Mobility makes white-label software for vehicle sharing, rental and taxi businesses, so companies around the world can run mobility services under their own brand. It sells two products, Vehicle Sharing and Ride Hailing, each with its own audience, sales cycle and targets.

Their Google Ads were bringing in a healthy number of demo requests, but only about 10% were sales-ready. The other 90% were low quality or spam.

Every request went to a sales rep for review, and only a lead that checked out was sent back to Google Ads as an offline conversion. That took weeks, sometimes months. By then the signal came too late, and too rarely, for Google Ads to learn which keywords and audiences brought real customers.

THE GOAL

Two products, two targets

  • Vehicle Sharing: cut offline CPA by 55%
  • Ride Hailing: cut offline CPA by 22%
THE METHOD

An AI model that scores leads in 24 hours

Our account managers and data scientist built an AI lead prediction model to score incoming leads, and rebuilt the customer journey around it:

  • Each lead is judged on the model's prediction and on feedback from the sales team
  • The model updates its scores every 24 hours
  • Google Ads optimizes for AI-predicted conversions and sales-qualified conversions, instead of every form fill

Campaigns now got a quality signal within a day instead of weeks, so Google Ads could learn which keywords and audiences brought leads worth talking to.

THE ACCOUNT

What we changed in Google Ads

  • Keywords: bids moved toward the keywords that brought higher-quality leads, going by the AI scores and CRM data
  • Landing pages: copy, design and UX reworked using Microsoft Clarity heatmaps, so fewer people dropped off before requesting a demo
  • Budgets: money moved away from markets that produced clicks but not contracts
THE RESULT

CPA down 56%, lead quality up 54%

With the model and the account changes in place, CPA fell 56% on key cost metrics, and lead quality and conversion signals rose 54%. The work was a Search Engine Land Awards 2025 finalist for B2B search marketing, won Best Lead Gen Campaign at the Golden Parrot B2B awards 2025, and won the Data-Driven, AI and Innovation category at Digitegu 2024.

What we took from it: for Google Ads, a good-enough prediction in 24 hours is worth more than a perfect human assessment in six weeks. The sales team still qualifies every lead. The model tells them where to start.

Cost per acquisition: before vs. after AI lead scoring
Before AI
Base
After AI
−56%

Read the full write-up →

Search Engine Land Finalist 2025Golden Parrot B2B Winner 2025Google Premier Partner 2026

Is your algorithm optimizing for the right people?

Book a free strategy call →