How new machine-learning-based marketing technology can help retailers

Retailers are in a bind, say the reports. Falling footfall on the High Street, falling sales (take your pick of reasons why), squeezed margins… As it applies to technology and marketing: what, I was recently asked by a retailer, should we focus on?

It’s flattering to be asked, of course, since QueryClick is no retailer. But I was asked because we do work with quite a few retailers, applying several unique-to-QC technologies to solve their challenges – so, actually, we’re in the position of enjoying a privileged view of the retailing world. And from our advantageous position, it seems clear to me that, in the current climate, top of the agenda in respect of technology/marketing for retailers needs to be top of the funnel.

Winning new customers comes at a cost: Cost Per Acquisition (CPA). Data and the application of intelligence to those data can identify both where your future customers are and provide the ability to market to them. The earlier in the customer’s journey – the closer to the top of the funnel – that this takes place, the greater the opportunity to minimise CPA. And the greater, too, the opportunity to massage them through the funnel to conclude their purchase with you.

But two things are needed to achieve this. First, investment in customer data management technology that supports the ‘single customer view’ required; and, second, investment in accurate attribution technology that integrates fully with customer data and all marketing channels.

Investment why? Because while both such technologies exist, and you may already have them, there’s wide variation in their efficacy.

Yes: legacy data management systems typically hold sufficient data for machine learning approaches to work. But, equally often, they also have challenges in terms of scale, security and legislative compliance (GDPR, PII access control, etc.). Moving to a technology stack that allows solutions to these challenges while simultaneously allowing immediate business impact decisions based on finding customers in multiple marketing data points will become the industry standard within a year or so, but making the right decision to get there today is often challenging. Ask us about ours.

Meanwhile, we’ve also developed and operate custom machine-learning models that enable multi-channel attribution for our clients, both online and offline, removing the traditional headache of trying to use tags, codes and cookies to cover all customer-matching.  Our machine-learning attribution solution – Corvidae – has identified some hefty missed attribution and some equally substantial mis-attribution, such as the £9.7m revenue per day that we identified for Tesco, the £4.5m of paid-spend cannibalisation that we detected and removed for Airbnb, and the 48% of total turnover (!) that was incorrectly attributed for a well-known sports equipment brand.

If your attribution isn’t cutting the mustard then, as a retailer, you’re almost certainly leaving revenue on the table and paying more than you need to earn the revenue you are getting. And no retailer can afford that right now.

But when you do combine great data management with great attribution – and with great subsequent acquisition strategies – huge benefits can accrue. In one study we ran, we saw a reduction in CPA of up to 78% over trying to buy that customer in Paid Search using siloed data. Seventy eight percent! In today’s retail environment, surely every retailer could use results like that.