It had only been a couple of months after the project started. We were working with a world-class retail company with both in-store and online presences at a time when COVID-19 response measures triggered an economic downturn.
When I joined a significant call with our client-partner, it was clear that the expectation was to deliver a super early “end of project” – since many retailers were curtailing their operations and/or projects because of uncertainties due to the pandemic.
However, as we presented the details of our work, our partner was amazed. We posited that there was a significant opportunity to lower the cost of returns through predicting if customers will return an item before they purchase it.
The company offers a 30-day free return for most purchases, which comes with huge expense for the company (including shipping costs, damages, restocking costs, merchandise carrying cost, store associate time, non-congruent inventory and store transfer costs).
By predicting if a customer will return a product prior to purchase, the retailer would then be able to intervene and reduce the rate of return. At Theory+Practice we call this “intelligent intervention”.
The intelligent Interventions we proposed focused on three key questions:
- What is the probability of return for each item in a customer’s basket?
- What is the relative value of each potential intervention – given these predicted probabilities?
- What is the most appropriate intervention for each individual case – given a cocktail of inputs, including the probability of return, the relative valuation of feasible interventions and any number of optimising objective functions?
At the time of our presentation, we received the feedback:“We have never worked with anyone who thinks quite like you.”
With this enthusiastic response, it would have been reasonable for the team to take a moment to bask in the warm glow of praise, but – in the internal debrief call that followed immediately after the conversation with our client-partner – I was impressed to see that everyone on our team was ready to dive deeper right away.
The big question on everyone’s mind was: “What would an AI solution for limiting the possibility of negative returns look like?”
Since I had only recently joined Theory+Practice, this incident was welcome confirmation of my belief that there was an underlying culture that embraced questions, no matter how challenging the situation may seem. This is a culture that is passionate about pursuing inquiries that would lead to the creation of true value for our partners.
The team designed two AI models for predicting the probability of return for each item in the shopping basket – and while the models were developed for a particular retailer brand, we ensured that their architecture would enable our client to easily transfer them for use with all their other retail chains.
Given the sensitivity of this project, against the backdrop of the economic downturn due the COVID -19, it was particularly pertinent to be clear about the individual risks and rewards of each potential customer intervention. By triangulating a set of targeted intervention strategies (such as final discount offers linked to no return), a set of assumptions (such as a customer’s propensity to accept a final discount offer), and a range of AI model outcomes, our predictive ROI revealed a range of values.
On the top end, our work determined an opportunity of a staggering 4,400% return on investment.
At the same time, at the bottom end, we highlighted the significant risk of million-dollar negative returns. This proactive revelation – that an intervention lacking in individual context-sensitivity (such as the original 10% discount the retailer had preconceived) could actually lead to such damaging unintended consequences – not only averted potential disaster for our partner but opened up opportunities for a much more personalised and targeted approach.
Significantly, by linking a range of outcomes to an Intelligent Intervention model, our client-partner will now be able to implement a targeted value maximisation strategy – where the most appropriate interventions are applied to customer interactions in real-time.
I was particularly pleased to see that our partner truly valued our approach – and understood the importance of going ahead with the project even at such an uncertain time.
It is now clearer than ever that it is essential to weigh risks against any return on investment for AI projects and I believe Intelligent Interventions are needed, now more than ever, to help amplify such possibilities. The market is shifting and companies need to be prepared. They need to continue to deliver value – to customers as well as stakeholder – while remaining financially viable.
Implemented appropriately, intelligent interventions – underpinned by the pragmatic use of AI, technology and data – can not only unlock viable solutions to help businesses succeed but can also minimise the real risks of negative returns on investments.
By Edosa Odaro (LinkedIn Profile)
FT Top 100 Most Influential Leaders