Remember when consumers were panic buying toilet paper, hand sanitizer, and disinfecting wipes? Remember the rise of the stay-at-home economy in 2020, when visits to Amazon surged?
Those days are long gone.
Consumer behavior has shifted dramatically again. People are returning to brick-and-mortar stores en masse. Amazon reported its first quarterly loss since 2015. Sales of hand sanitizers are down 50%, and Clorox expects its company revenues to decline in coming months because people are no longer panic buying home cleaning products such as sanitary wipes.
This shift has confounded many who thought that the pandemic would usher in a permanent demand for cleanliness-related products. And it won’t be the last time consumers shift courses. Retailers and consumer products firms in particular have an incredibly difficult challenge trying to anticipate the constantly changing nature of consumer behavior.
[Read more: A Future Redefined — 2022 Retail and Consumer Goods Analytics Study]
By combining first-party data with third-party signals of consumer intent, machine learning can predict what consumers will do next.
Retailers are sitting on a mountain of first-party data — especially chains such as Target and Walmart that can track every aspect of the shopping journey through native apps. They’re getting better at sharing that data with CPG firms to understand where consumer behavior is headed.
For example, Kroger and Walmart both use cloud-based platforms to share first-party data with CPG firms in order to collaborate more effectively. Kroger’s 84.51° Collaborative Cloud draws upon 2 billion annual customer transactions to provide CPG brands insights into demand patterns. Walmart’s Retail Link platform can help suppliers make better decisions about feature placement in the store but also aids in the sale of adjacent products.
A CPG firm might notice that consumers who buy potato chips also routinely buy a ranch dipping sauce on the same visit. The supplier may consider developing their own brand of ranch dip to capitalize on these extra sales. And those are just a few examples of how powerful first-party data can be.
Sharing first-party data is helpful, but not the complete answer. Consumers do not shop in a vacuum. Their behavior is affected by factors outside a retailer’s control, such as weather conditions. So, retailers need to tap into third-party data to read signals that could change purchasing behavior dramatically.
That data exists in real time, too. For example, both Amazon and Google Search are vast, all-knowing barometers of purchase intent. Google processes 8.5 billion searches a day. Those searches provide potential clues about what consumers are interested in buying.
According to Google, during the height of the pandemic, there was an uptick for searches for “candle making kits” (a 300% increase). Searches for “patio heaters” increased by 600% as people moved to eating outdoors. Right now, searches for “spring shoes” and “cheap gaming chairs” on Google are trending, among many other products.
Third-party signals also proliferate on social media. Retailers actively monitor TikTok to understand signs of purchase intent especially among the cool kids that TikTok caters to. Many people on TikTok love to showcase their latest fashions and personal purchases — so often that the hashtag #TikTokMadeMeBuyIt has accumulated 11.7 billion views as of this writing.
A casual scan of posts tagged #TikTokMadeMeBuyIt reveals countless purchases being discussed, ranging from levitating bulb lamps to waterproof shower phone holders. Now, consider how many people on TikTok have become one-person influencing machines by accumulating high follower counts, and you can see why TikTok alone is essential to retailers to understand trends and adapt their demand forecasting models accordingly.
The retailers that can incorporate third-party data with first-party customer data will have an edge as consumer demand shifts suddenly — which it will. But there’s a catch with third-party data, and first-party, for that matter: it’s impossible for a human being to track it, analyze it, and act on it. This kind of data explodes in real time by the second.
Not all consumer signals are useful — many could create a needless and very expensive distraction to a business that lacks the right tool to really dig into data at scale and then segment it at the store level.
This is where machine learning can help. Machine learning is a branch of artificial intelligence (AI) in which machines teach themselves with minimal programming from a human being. Many businesses, including retailers, already use machine learning in their marketing and promotions.
[Learn more about these technologies and the role of data at Analytics Unite 2022, being held in-person in Chicago this June.]
For instance, machine learning can be helpful in direct marketing and email by ingesting vast sets of consumer data and using it to determine things such as the best times to send emails. Machine learning also represents a largely untapped opportunity for demand forecasting.
Machine learning is especially adept at finding nonlinear connections that are crucial for demand forecasting — such as less obvious search behavior, where intent to purchase is not overt. Even an automated platform would have difficulty uncovering those nonlinear associations without machine learning.
Machine learning also makes it possible for a retailer to separate useful signals from irrelevant data. That’s because with initial programming, a computer using machine learning can literally get smarter at learning how to zero in on data that matters before analyzing it — for instance understanding how to detect a surge in demand for umbrellas in a particular city versus searches about the show The Umbrella Academy.
With machine learning, retailers and CPGs can take third-party data and shared first-party data to another level of performance. For instance, they’ll be able to do better scenario planning. CPGs and retailers can do “what-if” analysis with computer simulations. They can analyze the likely impact of running a promotion at a certain date based on the anticipated actions of a competitor. The potential scenarios are endless.
How, for example, might the rising cost of air travel affect a promotion for products that appeal to people planning shorter road trips in their cars during the summer?
For machine learning to succeed, retailers need to keep humans in the loop. People are needed to identify the data that matters. People are needed to train machine learning models with the variable data they’ll need to sift through unstructured, third-party data.
A good place to start is to identify a specific milestone coming up — say, back to school season later in 2022 — and collaborate with your technology teams to collaborate on a path forward with machine learning, first-party data, and unstructured third party data.
Successful retailers won’t wait. You should not, either.
—Vasudevan Sundarababu, Senior Vice President, Head of Digital Engineering, Pactera EDGE