3 Ways Machine Learning Helps You Find Your Blindspots

Each one of us has one thing we know fairly well. A topic which we could be quizzed on and remain unfazed. Then there are topics we know exist, but don’t really know much about, for myself that’d be, say astrophysics or neoclassical art. But then there’s things that we can’t even fathom because sometimes we just don’t know, what we don’t know. Yes, even in our areas of expertise.

When this leads to things happening that you couldn’t see coming, it’s called a “Black Swan” event. So named after a poet in the second century that imagined a bird as rare as a black swan. At the time, these were thought to be as real as werewolves and unicorns, but of course, we now know that not to be true. Yet, before stumbling upon Australia, even the most esteemed ornithologist (bird expert) of the Western world would have argued against this fact.

What does this mean in the world of e-Commerce?

Well, it means that despite knowing your own business better than anyone else, you may still have blindspots. Actually, you might not even know that they’re there.

 

Have you ever seen a drop off of customers, despite knowing your target audience and the way they tick? Or, have you ever spent time matching products to form recommendations but still not seen the results you wanted?

 

What might seem logical or obvious to us, may be quite the opposite when actually put into practice.

 

Enter Machine Learning

Opposed to manual solutions that try to spot patterns and pre-empt your customers needs, machine learning (ML) is faster, more accurate and reaps bigger rewards. Plus, studies find that as ML creates outputs that are logical to the human thought process. So, it can work in harmony in with humans, not in place of us.

 

There are many ways we can use ML to help our own processes, here’s just three:

 

Did you mean…?

Because of modern search engines, we are used to seeing corrections for our typos. It’s no big leap to add this into your search function but the winning factor with machine learning, is that it is ALWAYS adjusting to customer queries. Where we may set up a search engine to recognise “read” as “red” as a common mistake, with ML if customers also start searching for “r e d”, it will be added to the library of terms for ‘red’.

Matching products

It’s not that uncommon to see webshop owners taking individual products and matching them to others. This can be a mammoth task as the catalogue grows, not to mention that it may miss out on some winning combinations.

When selling a bike, a human may find it logical to add a bell, basket and helmet as recommendations. That isn’t too far-fetched, but ML may find that many customers who buy bikes, also buy a book about Tour-de-france. Again, not so far-fetched but also, not a first thought.

Assume, Verify, Tune, Repeat

I’m talking about targeted email campaigns. These four steps are the tried and tested way campaigns have worked for time immemorial. Whilst campaigns like billboards, print copy or TV advertisements have limited ways to verify engagement, email open and click-through rates are more easily measured. This way it’s easier to see which customers are responding well to content, allowing you to tune both your campaign and audience, before repeating the process. With automated emails, there isn’t even a need for manual verification and tuning, it’s done for you, so you can sit back, relax and enjoy the conversions.

 

ML helps you to save time and notice things you didn’t even realise you were missing out on. Why then, would you want to use ML software that takes time to learn your webshop, its products and its customers. Not all products are built equal, good software works from day one, no matter the size of the data.

Sidenote: Yes, I am currently reading Nassim Nichholas Taleb’s book, ‘The Black Swan’. How did you guess?

Try Clerk.io, free for 7 days and see how we can help your webshop:

Free 7-day trial 

 

 Check out more content like this.