The Challenge
A real-estate platform came to us with a familiar problem for a fast-growing business: the data needed to run marketing and sales well existed, but it was scattered. Marketing, web, and sales data each lived in their own systems, in their own shapes, with no reliable way to join them.
That fragmentation had a direct commercial cost. The telesales team was working leads without a clear signal of which ones were worth the call. Effort was spread evenly across prospects that were nothing alike, and the result showed in the numbers: a telesales fill rate sitting at 7 percent. Most of the calling day was spent on contacts who were never likely to convert.
The platform did not need more leads. It needed a way to see the leads it already had clearly, and to point its people at the ones most likely to buy.
Our Approach
We treated this as a data problem first and an automation problem second. There is little value in scoring a lead on top of numbers you cannot trust, so the sequence mattered: get the foundation right, then build intelligence on top of it.
Crawl phase. The first task was to bring the disparate sources into one place and make them agree with each other. We consolidated the platform’s marketing, web, and telesales data into a central store, resolving the inconsistencies in definitions and formats that had made cross-source analysis unreliable. The goal of this phase was modest and important: a single, trustworthy view of a lead and its history that the rest of the work could stand on.
Walk phase. With a dependable foundation in place, we built the layers that turned data into decisions:
- Modular BigQuery data warehouse. We designed the warehouse in composable parts rather than as a single rigid model, so new sources and metrics could be added without reworking what already existed. This gave the platform one governed place for marketing and sales data to live and be queried.
- Predictive lead scoring. On top of the warehouse, we built a model that scored inbound leads by their likelihood to convert, drawing on the signals the consolidation had finally brought together.
- AI-driven marketing and qualification automation. We wired the scores into the day-to-day workflow, so qualification stopped being a manual guess and became a repeatable, data-informed step ahead of the call.
- Sales prioritisation. Finally, we ordered the telesales queue by score, putting the highest-probability prospects in front of the team first and giving them a clear reason to spend their time where it counted.
The Outcomes
The combined effect was a sales operation working from evidence rather than instinct. The team stopped treating every lead the same and started spending its hours where conversion was most likely.
- Telesales fill rate rose from 7 percent to 44 percent, a direct result of putting the highest-probability prospects in front of the team.
- Marketing and sales data moved from scattered, inconsistent sources into a single governed warehouse the platform can query and trust.
- Lead qualification shifted from a manual, subjective step to a repeatable, score-driven one.
- The modular design left the platform with a foundation it can extend, rather than a fixed system it will outgrow.