Building an effective multi-model AI lead generation solution

The client

Leading Biotechnology Company

Our client is a market-leading biotechnology company headquartered in the UK that sells to more than 100 markets across the globe. They have a proven track record of delivering disruptive technologies to the markets that they serve and are focused on automation and increasing ease of use through the provision of more cost effective solutions and an increasing number of products that can be used without internet connectivity. Our clients’ technologies are used by research scientists worldwide as well as to support ‘real-life’ decision making in healthcare, industry and other applied settings.

ONT_anon case study_illustration_3_Client

Key Outcomes

We helped a leading biotechnology company to address their digital marketing challenges through the use of AI. Working together, we built a multi-model AI lead generation solution that is highly accurate and integrates effectively with their existing systems. This has resulted in increased conversion rates and a more effective sales process.

Conversion time 2 Weeks to accurately convert marketing qualified leads into sales qualified leads
Conversion rate 4x Increase in lead conversion rate
Accuracy 91% Predictive accuracy in the ‘likely to buy’ AI model
Accuracy 95% Predictive accuracy in the ‘customer fit’ AI model

The Challenge

A rigid, complex marketing product that was holding the client back

Our client had been using a software-as-a-service (SaaS) product to apply machine learning (ML) to multiple aspects of their digital marketing including: personalised website experiences, ad targeting, user segmentation, segment value prediction, and lead score generation.

Unfortunately the product did not meet the expectations of their marketing team due to: a lack of lead scoring explainability, and thus their ability to fine tune. High degrees of complexity and long lead times in making adjustments. A lack of additional benefits beyond the rules set in their customer relationship management (CRM) system, and very limited opportunities to adapt the core functionality of the product.

As a result they challenged us to produce a flexible AI tool that would allow for the discovery, categorisation, and prediction of leads with an emphasis on high-value leads. This tool needed to employ reusable APIs for the integration of lead scoring across different client systems including Salesforce, which was the sales and marketing teams primary tool, so as not to disrupt business operations. In addition, we were asked to include dashboard functionality so that key performance indicators (KPIs) could be visualised and performance tracked effectively over time.

ONT_anon case study_illustration_3_Challenge

THE SOLUTION

A multi-model AI lead generation solution

Working in sprints over a period of around 6 months we partnered with the client to build a multi-model AI lead generation solution that integrated with their existing systems in order to predict both customer fit and likelihood-to-buy, which when combined gave an accurate, relevant lead score.

The multi-model AI solution was built on AWS SageMaker to focus on performance and accuracy, whilst maximising compute efficiency. It ingests data from multiple sources including the clients’ website and CRM systems. Data is then processed and transformed using machine learning to generate accurate, reliable lead scores. These scores are then visualised in a dashboard for ease of understanding by the clients’ marketing and sales teams.

ONT_anon case study_illustration_3_Solution

The Results

Accurate marketing leads that convert quicker and at a higher rate

Our AI lead generation solution helped the clients’ marketing and sales teams to focus their efforts on more reliable sources of new business, improving efficiency and maximising their return on investment. Whilst our solution predicted 50% fewer monthly leads than their existing solution, the quality was significantly higher meaning that when contacted their propensity to buy was much greater.

Following extensive testing of the dimensions used to build the AI solution, we embarked upon a further phase of work lasting 1 month during which the machine learning models were retrained to include revenue as a feature, resulting in highly accurate predictive AI models. This resulted in a 91% predictive accuracy rate for the ‘likely to buy’ AI model which is used to generate a probability score of how likely a customer is to buy a product. As well as a 95% predictive accuracy rate for the ‘customer fit’ AI model which is used to predict whether a registered user is a good fit for a particular product.

This enabled the clients’ marketing team to be able to successfully and accurately know the grading of leads sent to sales. In turn giving them confidence that 4x more of the marketing qualified leads (MQLs) provided turned into sales qualified leads (SQLs) with an increased potential of going on to become paying customers.

For their counterparts in sales, significant efficiencies have been gained so that time isn’t wasted contacting low-quality leads. Allowing them to confidently focus their outreach efforts on leads with a good customer fit and high likelihood-to-buy, which currently convert to an active sales opportunity in an average time of just 2 weeks.