We tried to improve e-commerce business metrics using AI

Here’s what happened..

Customer retention with incentives

Most marketers frown on using discounts to retain customers. Discounts are however, undeniably ubiquitous. From small businesses to e-commerce giants — handing out incentives to drive sales seems to be the in thing. And it’s somewhat obvious that discount campaigns help with fast acquisition. But what happens post acquisition? Are discounts really capable of achieving long term customer retention? And if so, what’s the catch? Will it lead to customer addiction as we’ve seen in the recent past with deep discounts? Or can it revolutionize e-commerce marketing as we know it and become a sustainable way to achieve business growth?

We’ve already covered this in good detail here. So here’s a concise version: Customers today have a lot of options; too many for one’s liking. So why would they choose one over another? It could be a host of reasons, but one of the important ones is cost. By offering a discount, you’re giving customers an incentive to remain your customers. And if the FOMO on an offer isn’t enough to convert customers, we now have custom offers for each of your customers? (Check it out here). With customized offers the burn in your pocket won’t be as deep and you’ll get better conversions than you would by running a blanket offer for your entire customer base.

Having planned and executed a fairly large number of discount campaigns for e-commerce/mobile-commerce players of various sizes, we’ve got some data driven observations to drawing insight into these campaigns. We might have dealt with a small sample size of businesses, restricted to a few domains, however, the customer behavior and response learning when it comes to different domains could definitely be extrapolated.

How AI fit in

Once all the human made decisions were made with regards to budget, d uration of campaign and audience to be targeted, these inputs were fed into MARS, an AI powered tool that uses deep reinforcement learning to learn and understand customer behavior and send out an incentive worth just the right amount to lead them to an immediate transaction as well as nudge them onto the journey of long term retention.

When our customer businesses first started using MARS to run marketing campaigns, it operated in a slightly different way than it does today. Based on a customer’s present state and considering past behavior, MARS would pick from a pool of selected actions and send out the most appropriate one. Still intelligent, however there was a bit of an issue. If you gave MARS a pool of discounts ranging from low value to high value, and you ran a campaign on an audience which mostly consisted of customers who’d been inactive for a long time, it’s likely that MARS would send the high value discounts to most of them. This would lead to overspending on customers who might return to the platform only because of that one-time high discount and never transact again unless a high discount was offered. We came to realize this strategy was neither wise nor sustainable. So we added a budget allocation feature to MARS and added more options of discount strategies for MARS to pick from. This offered a larger pool of actions MARS could pick from and instead of just varying the amount of discount for every customer based on their behaviour it could also pick the right way of presenting that offer.

What happened once we clicked launch?

We’ll directly get down to the results. After running optimized discount campaigns, up to 10% higher retention was observed among customers. How does this happen? Generally businesses offer large discounts to first time customers to get them to try out the business in the hope that the customer will experience value and continue transacting with the business for the long term. However a price reduction during the very first time a customer is purchasing a product might also lead to wrong expectations about price which might be unmet later on. Thus it isn’t always the best decision to offer all first time customers a hefty discount. MARS picks the most suitable discount amount which will lead a customer to complete a transaction coupled with the best strategy to prevent customer addiction and ensure a repeat in transactions over the long term.

MARS discount campaigns also led to an increase of up to 65% in purchases with discounts. Since users expectations were being met with the incentives offered to them, purchases, and thus revenue increased significantly.

What’s next?

MARS is a deep reinforcement learning system. It learns from every discount it sends out and improves with it. Therefore as a campaign proceeds , it gets better and better at sending out the right discounts to maximize conversions and gets more and more efficient at spending the allocated budget. Moreover, MARS’ learning doesn’t end with one campaign, but is carried over to every subsequent campaign, so every campaign shows an improvement compared to the previous in terms of metrics improved as well as efficiency in spending the allocated budget

Wondering how you can use MARS to improve your business metrics? Try it out for free here.

Originally published at https://blog.marax.ai on October 15, 2019.