As industries go, the world of eCommerce is still a newborn baby. From its origins in static ads and interstitials to the birth of increasingly intelligent algorithms, the technology being used is evolving quickly and without respite.
To reflect this in our latest series of blogs we’ve been speaking to the best and brightest in eCommerce and retail to find out what the future will hold.
This week we’re speaking to Chris Lake, co-founder of Empirical Proof and former Director and founding member of Econsultancy.
So what stage are we at with machine learning in marketing? Is it a reality yet or is it just talk?
Firstly, I think it’s obvious machine learning has been needed in the industry for a while. But as humans it’s difficult to prioritise what’s important to optimise in the first place, and it’s equally difficult to find the human resource to deploy tests.
The problem is real. A consultant friend of mine recently got a group of stakeholders together and asked them to make an optimisation wishlist. They ended up with 700 ideas! They were forced to cut that down to just 50, in the space of a day. Cue lots of opinions and political manoeuvring.
It would be better if you could run those 700 ideas through a machine learning engine that scored and ranked them in priority for you, without resorting to arguments, posturing and best guess scenarios. Better still if that happened in real-time, on an ongoing basis. That would be very valuable.
So what examples are there already out there in the wild?
For the last few years, big data has been a buzzword and the idea of pulling it all together has been a holy grail. At this stage human input is still very much needed. We need to build smart neural networks so these machines can learn and optimise effectively.
I suspect the technology isn’t mature enough right now, even for the top ecommerce firms. Take Amazon’s recommendation engine as an example. This is very basic machine learning. Sure, it shows related products and makes contextual suggestions, but it’s still a little dumb in that it doesn’t necessarily know that certain products you only buy once. Also, I’m not convinced it has analysed the data deeply enough to know what kind of buyer you are. And that’s Amazon, the biggest of the bunch.
I think machine learning is very much a work of progress, among ecommerce firms and the tech vendors that support them.
Because of this do you think it’s harder to learn? Who’s leading the charge?
There are retailers doing a brilliant job with conversion rate optimisation and UX, but machine learning? I’m not so sure. You’ve had retailers like Zappos who have been renowned for customer service for years, but I don’t know that there is a poster boy for machine learning in the same way just yet. In terms of competitive advantage, you’ve got to think it will happen sooner or later.
According to Forrester, for every $100 spent driving traffic to an ecommerce site, only $1 is spent helping it convert. Thinking about this, what do you see as the relationship between acquisition and optimisation?
There are three key areas we look at in the customer journey: acquisition, conversion and retention. Optimisation is a layer that applies to all three.
I don’t think optimisation as a discipline has ever been in a better place but people are still so acquisition-focused that it’s crazy. So much money is being left on the table, and for me, the priorities are all wrong.
For new ecommerce companies, the bar is pretty high. Newcomers need to smash it out of the park with great mobile and customer experience. These things are a hygiene factor now, and a base expectation among consumers, though not all established companies do a great job in these areas.
Incremental – and continuous - optimisation is key to success. It can become a competitive advantage and point of differentiation.
What about the risk of marketing technologies damaging the customer experience?
This can be a problem, but it doesn’t have to be. At the heart of it is a lack of smart measurement, the lack of attribution.
Many people still have a very Web 1.0 way of looking at it. I don’t think it’s smart for marketing people to base a lot of decisions on the last click - there’s so much more to be gleaned from the customer journey that enables us to drive conversions without intruding on or irritating customers. Marketers don’t have to resort to the dark arts to win.
Is there anything else you’d like to add?
I think there’s a clear path toward success. If you wanted a competitive advantage a few years ago, you’d optimise for mobile visitors. 2011 really was the year of mobile, and anyone who didn’t jump on mobile would have been seen as a bit of a laggard.
There’s an opportunity in the same way now to get on board with machine learning and conversion rate optimisation. Getting proprietary, algorithmic ideas down and doing smart things behind the scenes is one avenue. The other is seeking out those vendors that will enable it.
One of the areas I’d expect to progress quickly is site search. Research by eConsultancy found customers who use site search are likely to spend more and are more likely to buy in that visit when they find a product in search. That’s a pretty obvious area to do some smarter stuff, with machine learning playing an active role in improving the UX and business goals.