It’s getting harder and harder to stand out from the pack. Most supermarket chains carry the same range of products, margins are tight, consumers have more and more online and offline options…so what’s a retailer to do?
A recent industry report predicts that global spending on AI by retailers will skyrocket from $3.6Bn in 2019, to $12Bn in 2023. Investments are being made in a range of departments, including personalized marketing, shelf-replenishment, smart checkout and more, all designed to help improve sales and margins.
Some of the technological kinks and logistical implications related to computer vision for self-checkout and shelf-replenishment initiatives are still being worked out. But the field of personalized marketing is already mature and improving every day. Hard-working machine-learning algorithms analyze shopping behavior in a never-ending loop, always on the hunt to identify potential marketing content that will truly influence purchasing decisions. And it works. Retailers who have implemented advanced AI marketing solutions enjoy higher sales, improved margins and more loyal customers when compared to their peers who have not yet gone down that path.
So how does it work? Without requiring a PhD in data science, let’s take a layman’s look at how leading retailers deploy algorithms within their personalization platforms to drive specific marketing objectives.
To begin with, sophisticated models take a high-level look at shopper behavior in general, analyzing their historical activity (up to 24 months), scoring each shopper and household on loyalty to the chain, price sensitivity etc. Shoppers that spend a higher share of their wallets at the chain will be treated differently than those who are less loyal, receiving marketing offers that are designed to maintain loyalty and grow their baskets. Those who are less loyal will be tempted by offers to visit more often and venture into more departments within the store.
Another statistical model will constantly analyze how shoppers are consuming and reacting to market content, specifically via which communication medium. The model learns from breadcrumbs left by shoppers in every interaction, for example, did they open an email with coupons, did they mouseover any of the overs, did they click on / load any offers to their membership cards, did they visit the store following the email, did they redeem any of the offers viewed or loaded etc. Each and every data point is a clue leading to how and what to communicate to shoppers next time around.
The above algorithmic models are run in a never-ending cycle on the data, along with dozens of others related to offer relevancy, shopping locations, brand loyalty and many more.
But however smart the machines are, they need to operate within real-word commercial constraints such as making sure budgets are adhered to and are allocated to have maximum impact, ensuring that offers are only distributed to shoppers in geographies where the product is in inventory, taking care to give shoppers a mix of offers spread over the entire store, not including competing offers or brands in a particular piece of communication etc.
And all of this needs to be optimized to ensure that both the retailer and the suppliers that fund many of the offers are jointly meeting their marketing objectives.
More importantly, the ultimate measure of success needs to be focused on how shoppers react to these activities. Shoppers who believe that effective, personalized marketing benefits them by saving time and money vote with their feet. And that’s highly measurable and proven by a number of statistical metrics.
Running an automated program based on the principles outlined above, for tens of millions of shoppers, day in and day out is simply mind boggling. It requires a special blend of capable technology and deep domain expertise. The retailers that manage to find that blend will continue to outperform those who don’t, period.
Happy trading
Chen Katz
