Large Automotive Organisation: explainable lead scoring for offline conversion
A large automotive organisation had no unified view of the digital customer journey, and its sales teams chose leads on intuition. Unolabs consolidated CRM, marketing and behavioural signals, then built explainable account scoring, attribution modelling and continuous feedback from sales outcomes. Revenue teams now see why an account ranks where it does.
A concise view of impact and engineering focus.
Improved digital-to-offline conversion
Explainable account scoring
Continuous model feedback loop
Why the sales team ignored the scores it already had
A large automotive organisation lacked a unified view of the digital customer journey, and its sales teams relied on intuition rather than data when deciding which leads to pursue. Intuition is not always wrong, but it is not reviewable and it does not transfer between people.
The causes were structural. Customer signals were fragmented across CRM, marketing and behavioural data sources, and the existing scoring process was a black box that revenue teams neither understood nor trusted. Two separate problems presenting as one symptom, and fixing either alone would have left the other in place.
The practical result was leakage at the join between channels. Online leads were generated at scale but were not consistently converted into offline sales, and nobody could say with confidence which leads deserved attention first, or why. A rank order without a reason does not survive contact with a salesperson who disagrees with it, and the salesperson usually has the last word because they own the call.
A scoring model nobody trusts is functionally the same as no scoring model — the failure here was as much about explainability and adoption as about data fragmentation.
Four layers, built so the scores could be argued with
We built a transparent account prioritisation framework in four connected layers. Enhanced user experience research came first, identifying where the digital journey actually created friction and grounding the modelling in observed behaviour rather than in assumption.
Second, we consolidated CRM, marketing and behavioural data into a unified signal base and applied explainable ML scoring using activity, sentiment and propensity signals, so every account score arrived with the reasons behind it. Third, attribution modelling allocated conversion credit across channels and campaigns, giving marketing an honest view of what was driving qualified demand.
Fourth, and most important for longevity, continuous feedback loops fed sales outcomes back into the model, so the scoring improved with use instead of drifting away from reality. The framework was designed to align digital channels with sales actions, not to replace sales judgement with a number — the distinction that decided whether the sales team used it or routed around it.
- Enhanced user experience research to identify digital friction points
- Digital lead scoring using activity, sentiment and propensity signals
- Attribution modelling to allocate credit across channels and campaigns
- Continuous feedback loops to refine the scoring model over time
Explainable scoring plus feedback loops is the difference between a model that gets adopted and one that gets overridden — revenue teams could see why an account ranked where it did, and their outcomes retrained the model.
What changed once the scores carried their reasons
The framework improved digital lead quality, increased conversion confidence and delivered clearer guidance for marketing spend. The adoption came first; the numbers followed it.
Sales and marketing teams could see why accounts were prioritised, concentrate effort on the highest-opportunity leads, and trace the connection between digital investment and offline results. Because the feedback loop runs continuously, the model stays aligned with current buyer behaviour rather than freezing at its launch-day assumptions. A scoring model is a perishable asset, and most of them are commissioned as though they are not.
That is what makes digital-to-offline conversion a measurable, improvable process rather than a one-off modelling exercise. It also creates a dependency worth naming: the loop closes only if sales outcomes are recorded, so the weakest link in the whole framework is a CRM field that somebody has to fill in after a conversation has ended. No amount of modelling sophistication upstream compensates for that field being left blank.
What we'd flag: digital-to-offline attribution is only as good as offline sales data capture — the reported conversion improvements are directional, and they depend on sales teams consistently recording outcomes to close the feedback loop.
What to carry into the next sprint
Takeaway
Ship the reason alongside the score, or expect the score to be overridden.
Takeaway
Design the outcome feedback path before the model, not after go-live.
Takeaway
Treat offline outcome capture as infrastructure — attribution rests entirely on it.
Frequently asked questions
- Why do sales teams ignore the lead scores they are given?
- Usually because the score arrives without its reasons. A model nobody can interrogate is functionally the same as no model: a salesperson who disagrees with a rank order and cannot see why it exists will simply work around it. Explainable scoring built on activity, sentiment and propensity signals is what made this framework adoptable rather than merely accurate.
- What makes a lead-scoring model degrade after launch?
- Buyer behaviour moves and the model does not. Without a feedback path, scoring freezes at its launch-day assumptions while the journey it describes keeps changing underneath it. This framework fed sales outcomes back into the model continuously, so it improved with use. The design decision that matters is building that loop before go-live rather than retrofitting it afterwards.
- How reliable are digital-to-offline attribution numbers?
- They are directional. Attribution across channels and campaigns is only as good as the offline sales data captured against it, so the reported conversion improvements here should be read as indicative rather than as audited figures. Where sales teams record outcomes inconsistently, the attribution and the retraining loop degrade together, and neither failure announces itself.
- Does explainable scoring replace sales judgement?
- No, and it was deliberately not designed to. The framework aligns digital channels with sales actions by showing which accounts look most promising and why, leaving the call with the person making it. Scoring that presents itself as a replacement for judgement tends to get overridden; scoring that supplies evidence for judgement tends to get used.
- What has to exist before lead scoring is worth building?
- A unified signal base. CRM, marketing and behavioural data sitting in separate systems cannot produce a coherent account score, because each source describes a different fragment of the same buyer. Consolidating those signals came before any modelling in this engagement, alongside user experience research that identified where the digital journey was actually creating friction.
Find out why your sales team overrides the lead score
We will look at your CRM, marketing and behavioural signals together, then show you where the journey leaks, which scores arrive without reasons, and whether your feedback loop really closes.
Book a Lead Scoring Review