Predictive modeling and unlocking consumer journeys: MRS 2021 highlights

In the market research world, it's essential to keep exploring new tools, techniques, and trends to stay relevant and deliver the best possible results. That means that while we work on projects, we must often dedicate time to reading articles and attending workshops. One of the annual conferences that give an in-depth overview of current trends is organised by MRS (Market Research Society).  

A few weeks ago, I participated in the 2021 edition, which was full of exciting demos, case studies, and practical exercises. In this article, I will summarise sessions and share the key takeaways.  

During the morning, wrapped up in six varied sessions, the speakers focused on answering four different questions: 

1. How can we optimise predictive modelling? 
2. How can we understand the path to purchase? 
3. How can we get more nuanced consumer feedback through chatbots or text mining? 
4. How can we measure media effectiveness? 

Optimising predictive modelling 

Russell Bradshaw, a Senior Statistician at Money and Pensions Service, led the first session. He focused on targeting meaningful interventions and advice for a person in financial distress at the right point. MaPS use advanced analytics to help identify where financial distress is most likely to arise. Firstly, they use an iterative segmentation approach to understand and classify cohorts of people into appropriate levels and complexities of financial distress—secondly, they create a predictive micro-model of the UK adult working-age population.  

Using a combination of minimum spanning tree approach to help visualise the connections and traditional cluster analysis within two sets of the inputs drawn out by the visualisation, they were able to identify predictable and actionable groups to which interventions could be targeted.  

They then used secondary data to feed in their models to predict the level of financial distress in the future.  

Key takeaway: When it comes to predictive modelling, the benefits of transparency are crucial, as well as re-evaluating the model and learning from forecast errors. The more open you are about your model and its assumptions, the more you can be receptive to different people's inputs, as well as recognise any limitations.  

Understanding the path to purchase  

The next session was particularly interesting from a personal perspective, mainly because it touches on an essential point for our work here at boobook - how to unlock the consumer journeys. Nicolas Camargo, Analytics Consultant, SKIM and Michael Hetherington, Director – APAC, SKIM, discussed how consumer journeys used to be a linear set of predictable events. Still, today this is no longer the case. There are many more loops for consumers as they go back and forth across many touchpoints throughout their journey. Using a mixture of survey and passive data, as appropriate, they apply a pattern analysis technique to pull out relevant sections from customers' journeys. These can be used to reinforce provisions when they have a high chance of leading to a purchase or suggest further improvements when purchase attrition is high.  

Key takeaway: A simple and intuitive set of survey questions, combined with sophisticated pattern recognition modelling, can unlock critical journey insights. By uncovering naturally occurring customer journey segments, we can pinpoint which offer the most economical value and are most capturable by a brand based on the infinite number of potential sequences.

Gathering more nuanced consumer feedback through chatbots and text mining 

Ansie Lombaard, Director of Offer Innovation, Insights Division, at Kantar explained why brand purpose matters and how we can leverage advances in text analytics, using conversational data collected through personal, engaging bot interactions, to explore different ways in which a brand can manifest its purpose.  

Andy Crouch, Business Development Director, Pansensic, introduced an empathetic chatbot that empowers its questioning path/algorithms with AI-enabled real-time emotional, psychometric, and psycholinguistic understanding of the interviewee and their responses enabling a more engaging experience.  

Crouch took the example of their current projects focusing on mental health within the armed forces. Using many kinds of different text to feed into developing this chatbot, they also use real people to help train the data by categorising content manually. While looking at sentiment and emotions, we are reminded that people act on emotion, not opinion. If you can get to people's feelings, that's the key to driving behaviour. They've discovered through their research that, even though people know they're talking to a chatbot, people can be more open with a chatbot versus a trained counsellor partly because it's perceived to be non-judgmental.  

Of course, technology has a critical role when it comes to implementing chatbots and text mining. Maurice Gonzenbach, Co-Founder & Machine Learning Engineer at Caplena, showcased their advanced text analytics tool that we also tried at boobook. Caplena is easy to use, intuitive, and flexible. It lets you tap into all kinds of different data sources. You can either access directly online to different web services using API's, or you can use your collected survey data as your open-ended property. The software enables a subset of the data to be manually trained if desired to give more finely tuned results to your topic of interest or relevant needs. With data visualisation, continuously enhancing features, and flexibility with survey data, it feels like one of the more applicable text analytics tools tailored to the Market Research Industry.

Key takeaway: One way to get to help build a strong brand is to use AI. People open a lot more prominently by applying conversational AI chatbots instead of speaking to a natural person. Careful planning meant they could get the chatbot to focus a conversation on drawing out the fundamental nuances of the brand. Additionally, people can be more open with a chatbot because it's perceived to be non-judgmental.  

Measuring and analysing the impact of media channels 

Bethan Palmer, Client Consultant at DVJ Insights introduced a new RPS (Reality Performance Score) KPI, which she described as having more predictive power than traditional media expenditure. RPS considers the strength of each medium, the decay effect, and the strength of the creative execution, as well as accommodating the actual seeing of an advert, not just being exposed to it. A holistic approach covering different media channels, RPS has been presented as a good tool in summarising media effectiveness.    

Key takeaway: Reality Performance Score is the first real KPI representing the total sum of all media. RPS helps us understand the relationship and the strength of the relationship, with all relevant marketing KPI's, such as sales, revenues, website visits, Facebook visits, awareness attitude, consideration, preference, etc. It can help define the most effective advertiser in each category, the most effective campaign or medium. 

Helpful tools to use for your next research project 

During the afternoon session, we participated in a practical, hands-on activity in accessing, manipulating, and visualising data. Having watched him speak at other conferences, Preriit Souda, Director at Data Science & Strategic Insights, at PSA Consultants, did not disappoint, as he gave a practical and useful demo of Power BI and KNIME.  

He also drew attention to many helpful free to access resources. Touching on the pros and cons of these resources, such as data.gov.uk, or Kaggle, you can access a wide range of data that is great for enhancing and triangulating with existing research or simply testing out different statistical techniques. 

In conclusion

Overall, all the speakers were very interesting and gave great examples that embodied the flexible approach and combination of advanced tools powered by human creativity.  

Each time someone from the boobook team goes to a conference or a webinar, we always share new insights internally and externally, as it's crucial to keep learning and expand our knowledge.  

These sessions also gave another view on some topics we work on daily, like customer journeys. Others, such as brand purpose and media effectiveness, provide important information we can integrate into our projects and share with our clients as an overall marketing strategy guidance.  

Never miss a story. Subscribe to our newsletter!