Operators use machine learning so customers make confident choices
As techies argue over the merits of Samsung‘s Galaxy S8 or holding out for the Apple iPhone X announced yesterday, spare a thought for Jo or Joe Average Buyer. Many of them just want a device that meets their current needs –
without breaking new ground in emojis that emulate 15 points of your face. Well, Jeremy Cowan may have an answer for them.
What chance, you may ask, do they have of finding the phone that most closely matches their needs?
Answer: Better than you’d think when browsing mobile network operator (MNO) websites in Germany (Ziggo and Deutsche Telekom), the UK (for all MNOs except, so far, Vodafone), New Zealand (Spark), and maybe soon in the US, Australia and Asia.
Experience at one specialist machine learning company, 15Gifts, shows that if a product is “techie” then people delay making a buying decision. International business development director, Jon Morgan says, “working on all (Telefonica subsidiary) O2‘s websites we monitor customer behaviour online. If they are confident we leave them alone. If they are pausing we anticipate their questions and needs.”
Developed, as the company name suggests, for the business of gift selections, “the 15Gifts engine uses machine learning and UX (user experience) techniques to help customers buy with confidence. It anticipates their questions when selecting a phone, it estimates their data usage, it knows the most likely price for conversion to a sale.”
“Our engine monitors customer’s behaviour to ensure we begin the conversation at just the right moment, with a perfectly tailored message,” says 15Gifts. “No two customers are the same. That’s why our engine quickly understands who it’s speaking to, and what they’re looking for,” tailoring the conversation to match their requirements. “Our algorithm improves with every single interaction – learning from the choices of like-minded customers’ to only recommend those products most likely to convert.”
A slider bar enables users to show what they expect to spend per month. The engine picks out the Unique Selling Points (USPs) of the proposed phone, but it also offers a ‘Next Best Match’. If a customer is not ready to buy just yet, the profile the engine creates becomes crucial for improving their next interaction – whether that is on the phone, in-store or via live chat.
Collecting data, too
“Equally important is the data we collect. Partners can login to the Tableau dashboard and every project has its own data scientist.” Morgan adds, “This isn’t a SaaS service (software-as-a-service); data must be constantly updated.” The analytics and insight produced through the engine is as valuable to 15Gifts’ customers as the conversion uplift it drives. The dashboard ranges from visitor demographic insight to individual customer profiles.
The engine makes more than 10 million recommendations a month, engines drives millions of recommendations, and the company can reportedly demonstrate significant uplift in conversion and revenue. Across all its partners, the 15Gifts engine manages a completion rate of 92% of customers. Research shows that 63% of sales would have happened anyway – “Hero products will always sell,” says Jon Morgan – but the flipside of that coin is that 37% of sales are incremental, purely through using the engine. The recommendation engine’s average share of all online sales is said to be 12%.
Call centre benefits
Viewed across all call centre locations, with 50% of agents given access to the engine and 50% acting as a control group – a call centre trial has seen a 39.6% increase in the number of sales per agent per day due to call time reduction and conversion uplift. “In addition,” says Morgan, “we have seen a significant uplift in overall call quality.”
Asked by VanillaPlus which companies 15Gifts is competing with, Morgan mentions Honeybee from Dixons Carphone Warehouse, but adds that here there’s no machine learning element.
Finally, as well as recommending devices and alternatives that are most likely to be purchased (based on sales to customers with similar profiles) the engine automatically adapts recommendations to changing trends due to product launches, special offers and pricing updates. Which brings us back to where where we started.
The machine learning specialist sums it up rather neatly, “Limitless choice isn’t all it’s cracked up to be.”
Comment on this article below or via Twitter: @VanillaPlus OR @jcvplus