By Angie Knibb, Head of Search
Hands up if you’ve heard the sentiment “machine learning will make search marketeers and agencies redundant.” I certainly have. I feel like for most of my career I’ve been hearing: “In 15 years machines will replace humans.” Well it’s more than 15 years later and I am yet to have a machine replace me, so I thought I was pretty safe.
However, the more I look into machine learning, the more I start to wonder if there is an element of truth in the sentiment. Once you really start to think about it, the concept starts to ring alarmingly true, alarmingly quickly. And that’s because it’s already such a big part of everyday life. Machine learning powers pretty much every dating app we use. It’s behind the recommendation engines on the likes of Amazon, Spotify and Netflix. And of course the digital duopoly (Google & Facebook) are using an abundance of machine learning to aid in advertising, customer service and more.
We’re enabling the machines to teach themselves to speak and interpret what we’re saying. Did you know that the biggest voice search at the start of 2018 was ‘I love you’? Even more scary was that one of the runners up was ‘Alexa, will you marry me’!
The biggest thing since the creation of ‘the internet’
But what is more tangible is that it’s now predicted that by 2020 50% of searches will be via voice. The driving force is that there are now 8 billion connected devices worldwide, which is more than there are people. Combine that with the fact that 10% of UK households have at least one smart speaker, and it is clear why we can already see the impact of this in our paid search accounts – by seeing longer, more conversational, and more informational queries coming through. As a result 24% of advertisers cite voice as a priority for this year, so if it’s not coming up in your discussions for next year, you may find yourself falling behind your competitors.
We’re also having machine learning teach itself how to see. With instant visual differentiation of two different, but similar looking, products on Google plus Pinterest’s visual lens queries growing 140% YoY, it’s hard to deny that this has changed the landscape, and will continue to do so.
For me, this is really the culmination of a clear societal change. The biggest previous technological advancement in my lifetime was the internet becoming a ‘thing’ – I remember my Dad sitting us down and showing us how to dial up with the dreadful dial tone, and I vividly remember us all using the phrase ‘I’m going online’. When was the last time you said that? Not for years I expect. And why is that? Because, to quote Google, ‘we don’t go online anymore, we live online’. We expect to always be connected to the internet, to take it with us wherever we go, and to always be able to find exactly what we need, when we need it.
Machine learning is not the enemy
The good news for us as marketeers, though, is that this creates huge amounts of data for us to analyse and utilise. These data points include useful things like user location, device, browser, search behaviour, purchase history, weather, and so on. To put a number on it, 2.5 exabytes of data are produced every day according to Google. I am reliably informed that this is the equivalent of 530 million songs. Now I for one couldn’t differentiate between that many songs, much less take each of them into account when optimising for an individual search, but a machine can. Which is exactly why machine learning has a really positive and exciting impact on paid search.
That impact comes in many forms, all the way from Quality Score, which was introduced in 2008, through to bid strategies optimising towards a set goal, or Dynamic Search Ads allowing us to appear for relevant searches we’re not bidding on. But the important thing to remember here is that we have really good models now, which go beyond the simple ‘if this, then that’ functionality to work towards a set goal, and we can reliably process vast amounts of data in real-time, allowing for the machine-learning models to consider billions of parameters.
The computation power and ability of the models will no doubt continue to evolve, bringing with them ever-improving analytical insights, greater and faster optimisations, and I am sure we will continue to see development and focus on bid and budget strategies. But what I see as our core pillars for innovation, both machine and human, are Audience, Integration and Attribution.
I believe we are moving into a much more audience-centric approach.
If you’re not there already then you certainly should be moving towards it. And that means thinking about why customers are coming to you, and what they need from you. The market leaders are using Personalisation across their channels and websites to really deliver on this, and that’s something the industry seems to be moving more towards, as we accept that we’re in a world of audience first, keyword or channel second.
Which leads me nicely to my second pillar: integration.
We need to stop thinking about what we want our channels and marketing budgets to do, or what the tech allows, and think about what our customers want. That means understanding how all channels support them on their journey, and then understanding the impact each channel is having overall. It means aligning messaging, budgeting and goals – which isn’t the same as saying they should all be the same, but it is saying they should be working together, not against each other, to focus on how we as a business meet the user needs.
And the final pillar is attribution.
Being able to track and understand all touch points is key, and that attribution needs to be in real-time, cross-channel, cross-device, full-funnel, data-driven, and above all it needs to be actionable – there’s little point in having an all-singing all-dancing attribution model if you can’t do anything with it.
Humans still have a crucial role to play.
Machine learning is best used to cut, collate, and react to the data in the blink of an eye, something we as humans can’t possibly do, and it’s invaluable when it comes to finding potential tactics, and the manual implementation and optimisation side of our roles. You do, however, need humans to look at how to best utilise those tactics to further develop or support your strategy, to focus on how we grow, and to understand what it means for your business.
That’s why I believe machine learning changes our focus, but not our value. Machine learning essentially provides the tools for greater audience targeting and attribution. We need people to identify how best to apply those tools to continue to grow and develop our strategy, and how to better integrate our channels to aid the consumer needs.
If you’re wondering how to apply this to your business, I’d recommend starting by asking yourself the following questions:
Are you keeping up with what your customers want?
Are you responding to their needs in the way in which they want you to?
How can you capitalise on machine learning for this across your channels?
Where do you invest the saved time?
What skills and people will you need in the next 2, 5 and 10 years to deliver this?