Each Click is an intent – where customers show
how serious they are in purchasing a product.

A Global Automobile Giant increased online conversion by 6.7% through redesigned lower-funnel activities.

About the organization

Our client is one of the “Big Three” automobile brands from the USA. Established over 90 years ago, our client organisation has a global presence and recognition.

Business Problem

Our client’s website attracted over 700,000 visitors a month. The goal of the exercise was to identify online activities which foster conversion and purchase decisions. A conversion path, which is the process by which an anonymous website visitor becomes a known lead, was identified for different customer segments.

Our Approach

Our analysis included website activities such as the pages typically visited, the content browsed which highly correlated to a subsequent conversion – essentially a mapping of the digital path to conversion for website visitors. We also moderated for the device, time spent on each page, originating and destination websites and cookie data.


Dissect user behaviour to identify consistent patterns for conversion paths

Data was collected on website activities, click-through rate, page-wise time spent browsing, page visits to specific pages such as shopping tools and vehicle content sites, new inventory search page visits, tools that were in demand. All data analysed was derived at a weekly level. The level of granularity was each individual access to the website.

Key aspects of the exercise

Anomaly detection techniques that were used to identify visits that originated from bots.

A conversion propensity model across behavioural attributes was built to identify upper-funnel activities which drive traffic to the website. Conversion was measured for each user segment; deep learning models were built to identify attributes that were highly correlated to conversion.


Our analysis revealed that overall conversion rate was 0.12% of all visits on the site.

A bi-directional effect was noticed: higher conversion may be caused by higher time spent on the website, or a self-selection possible where higher customer interest drives both results.

High-value-activities that were instrumental in purchase decisions were identified. An optimal observation window was obtained (leading weeks to sales) during which such activities were monitored for sales impact.

Time series models were built, and this combined with Granger causality, prove a reliable way to capture impact of high-value activities on sales.

Action Items for Business

Based on the pages and website activities that led to higher conversion, we used a Bagging Model based on Classification Trees to create a tool that identifies users who are more likely to convert than others (“Hot Visitors”), so, that these qualified leads can be subsequently targeted through a personalized message.

We created a data transformation module that quickly classifies visitors using a “Hot Visitor” identification tool in real time. Our tool was able to identify 9.5% of total visits where there was a high likelihood of conversion, but the existing interface did not present the option to the visitors. This also allowed for actionable, specific points of engagement to be built on the web interface to engage with prospects in a more meaningful, impactful manner.

The TEG Impact

Using machine learning techniques, TEG Analytics was able to identify patterns of user behaviour and conversion drivers, as well as specific areas of improvement and easy to implement suggestions that could improve conversion. Our analysis helped the client build a more impactful website by targeting prospects better.

Insights@Speed of Business


Hide dock Show dock Back to top
Loading