Depending on who you ask, machine learning is either a breakthrough on par with the harnessing of electricity itself, or an overhyped marketing gimmick whose inflated expectations are bound to create a sense of disillusion at some point for those who choose it as a dance partner.
Wherever your opinion falls, the debate ends with an empirical look at the ways machine learning can potentially impact decision-making, whether the decision is strategic, tactical, financial, or simply makes everyday life more productive. More than a tool for face recognition or piloting autonomous vehicles, machine learning is now showing itself in social engineering and medicine — helping to predict when participants might drop out of a drug trial, and improving hiring managers’ candidate selection process.
There is a reason Eric Schmidt believes machine learning will be at the heart of every successful big startup exit event in the next five years.
Advertisement
The fundamental questions these startups will most frequently seek to answer — indeed the first thoughts that enter the mind of anyone responsible for profit, loss or operations — is, “How can we unlock machine learning to provide a measurable business benefit? What can machine learning actually do for our business? Where is the material effect?”
In most companies, but in retail in particular, business value means revenue, pure and simple. Here are three ways machine learning is creating revenue for retailers, right now.
Turns Flagging Customers Into Buying Customers
One of the oldest axioms in customer acquisition is that the best source of new business is old business. Customers who’ve purchased before are the most likely to purchase again, and are far cheaper to acquire than net-new customers. Machine learning algorithms have improved to the point that they can spot customers who are most likely to churn, and empower the marketing team to interact with them ‘just in time,’ with an engagement that inspires them to buy.
Systems analyze factors such as the following to determine the likelihood of a customer going away:
- Date of first purchase;
- Date of last purchase;
- Days between first and last purchase;
- Total lifetime spend; and
- Total number of orders.
Machine learning algorithms can ingest this data, then recommend the next best step in terms of driving additional value from each individual customer. For example, the system can learn over time how many days between orders is indicative that a customer is close to falling out of the funnel. Then, when a customer is just about to reach that threshold, machine learning triggers the marketing team to prioritize that customer with a ‘win-back’ campaign.
Using predictive analytics like this to inform marketing activities both increases the likelihood of a conversion and reduces irrelevant offers or messages.
Makes The Perfect Recommendation To Increase Average Customer Spend
While Amazon popularized and perfected the recommendation engine with its “people also bought” feature, the value is tied to helping consumers understand products that their peers purchase together. Wider availability of machine learning is providing the opportunity for smaller, more specialized retailers to expand on this concept; now, retailers can quickly gain insight about purchase patterns among their own products. It is a subtle, but extremely important, difference.
Essentially, machine learning has the ability to predict a consumer’s next likely move by extrapolating a host of interesting patterns, such as:
- Products that are purchased together;
- Products that generally trigger a follow-up purchase in the near term; and
- “Look alike” customers
Because machine learning systems continuously analyze customer demographic data — gender, age, location, purchase history, etc. — they are able to quickly spot purchasing trends among a particular subset, then make an informed decision about the buying potential of similar customers.
This entails much more than merely tracking products that are similar to one another, or are often purchased together. Rather, it provides a means to actively get a retailer in front of entire groups of customers who may hold a latent propensity to buy, and prompting the marketer to take appropriate action.
On a more linear basis, there certainly may be times where a customer’s purchase of ‘X’ leads directly to purchase of ‘Y.’ For example, when someone buys brown shoes, the system may spot a discernable pattern: they almost always return one or two weeks later to buy a brown belt. Machine learning continually works to connect these dots every time a customer buys, and puts the retailer in position to anticipate needs and make shopping more efficient on the buyers’ behalf.
These types of intelligent recommendations not only drive additional revenue in the short term, but enhance the overall customer experience, which is perhaps the single biggest differentiator among brands in the age of on-demand retail.
Eliminates Wasteful Spend, And Adjusts ‘On A Dime’ To Preference Changes
One of the coolest aspects of machine learning is that the algorithms are in a state of constant adjustment. This not only makes the algorithm “smarter” over time, but it also identifies new trends in lockstep with shifts in market and consumer behavior.
For example, with regard to the above-noted recommendation engine, products that are frequently purchased together have a tendency to change throughout the year. In the spring, a suitcase might be frequently purchased along with a specific beach bag, but in the winter customers might pair that very same suitcase with a ski bag.
Machine learning raises these subtle changes directly to marketers’ attention, instead of relying on manual analysis to find and feed the information in a timely manner. In turn, there are fewer steps in the process; front line marketers are simply more ‘in the know’ about seasonal adjustments as they happen, which enables them to take advantage of purchase trends as soon as they begin to surface.
While everyone may not agree machine learning is a breakthrough on par with the creation of the Internet, it can help retailers keep their revenue potential consistently fresh, and allow them to edge their share-of-wallet ever higher.
Paul Mandeville serves as Chief Product Officer of QuickPivot, leading the firm’s efforts in product strategy, innovation, and design. Prior to joining QuickPivot, Mandeville served as Chief Operating Officer at Conversen, a cross-channel marketing technology startup located outside of Boston, Mass. He has more than 15 years of marketing application technology experience, holds several MarTech industry awards, and contributed to QuickPivot’s first U.S. Patent grant for the design of QuickPalette.