Google Insight: How Is AI Impacting on Paid Search?
As a Google Premier Partner Agency we were recently invited to attend an event at Google Dublin, at which a range of speakers from Google and its external partners held forth on the subject of machine learning and its growing impact on the areas of analytics, insight and attribution. This field has long been vitally relevant to paid search and the wider online advertising industry, and it was fascinating to hear what Google had to say on the issue. Along with a precis of Google’s views from the day, here are some insights on how this new tech is already impacting on our paid search work.
Perhaps counterintuitively for a conference concerning machine learning, an early speech covered the growth in human IQ. Taken globally, the intellect of the average human increases measurably each year: people are getting smarter. This growth in human mental faculties is having a direct impact on our capacity to engineer and educate machines, and on their ability to learn from us: computers are getting smarter too.
Google use “deep learning” within their products to improve services for consumers and advertisers, and as a consequence the performance of this tech is also improving rapidly as the curve of artificial intelligence gets steadily steeper. For example, although unimaginable even a few years ago, it is already possible for a user to take a picture of text written in a foreign language and have Google recognise and translate it. Google can also translate from spoken text. The Babel fish can’t be far away now, surely? Don’t forget your towel…
Elsewhere, the emergence of an “internet of things” has ensured the proliferation of home devices that can turn the lights on, change the temperature and play music on demand. While they are being more widely used, these web-based systems are accruing data more quickly and so becoming smarter: the uptick in their use is turning them into a more successful product, as machine learning optimises their performance.
Over and above its value in the domestic space, machine learning is proving invaluable to industry and in particular to science and medicine. One example given during Google’s event was an algorithm written by a fifteen year old high school student, Abu Qader. This tool has been created to detect breast cancer cells within mammogram images. (https://www.youtube.com/watch?v=El4OCwR6qxo). It is estimated that medical error is the third most common cause of death in the USA, suggesting that the use of AI to narrow the space for human blunders could save a great many lives.
Within our own industry, Space and Time are already using computer code to fulfil some of the more repetitive elements of PPC account management. Provided it is kept under the watchful eye and guiding hand of a paid search expert, an automated script can take a lot of the strain when it comes to optimising and reporting on our campaigns. We also use code to respond quicker than a human could and to work round the clock- such as the use of an API feed to make our ad copy weather responsive; updating creative in real time as the weather changes in a given location. Crucially, all this labour-saving technology frees up valuable time for our search experts to do the tasks that only humans can do: analysing data and realising insight; looking into new strategies; researching the marketplace and talking to our clients.
One theme that emerged throughout the Google event concerned the dangers inherent in an over-reliance on assumptions. Hundreds, perhaps thousands of times each day we all instinctively make instant, snap judgements based on heuristics, ill-informed assumptions and minimal data. It’s an essential evolutionary neurological shortcut that prevents us from collapsing into inertia while what might or might not be a tiger takes her time over eating us. Within the professional sphere however, where robust datasets are readily available and tigers are rare, it is necessary that this rush to assume is resisted.
The examples Google cited were compelling. It might be reasonable to assume, we were told, that the majority of Google searches for “new cars” are made by men. Similarly, surely most of the people searching for recipes online are women? Wrong and wrong. Ill-founded assumptions of this sort can be as costly as they are outdated. Google has over a billion users of its various products and incorrectly ignoring half of these people would be a major error for any marketing team.
At Space and Time we find it is essential to base decisions and recommendations on the data we have collected, not on what we assume. Even when beginning a project with a certain set of assumptions is inescapable, we use the available data and a range of research tools and proprietary systems to challenge these assumptions as often and as early as possible. This means we can always support a decision with data and so will not become trapped by our assumptions.
Budgets and Accrediting Conversions
The use of fixed budgets is common practice within search advertising. It’s wholly reasonable that not every advertiser has sufficient budget to appear for every last relevant search query that is made. Perhaps predictably, Google would like to see a good deal less of this sort of thing and have a number of reasons why we should all be spending more money with them. In many ways their stance makes good sense: once a campaign is delivering leads or sales at a quality and a cost that is acceptable and is essentially self funding, which business would not want to increase the quantum indefinitely? Cash goes in, sales come out. All jolly good. Of course in the real world there might be any number of reasons to limit the amount spent on even the best-performing campaign, but understandably Google’s ultimate commercial imperative is to increase their own revenues.
In order to help us open our pockets more liberally, Google has introduced Smart Bidding, an AI tool which helps to optimise performance towards the delivery of low-cost goals. Possessed of sufficient data and daily spend, Smart Bidding is able to increase the volume of conversions from a given budget. We are already using it to good effect wherever it is suitable.
As consumers have become steadily more capable and comfortable browsing from their mobiles over the course of the last decade, the phenomenon of “cross-device” activity has become ubiquitous. We are all now likely to use more than one device to search for something and then switch to another before deciding to make a purchase or to part with our information. Previously AdWords would only accredit a conversion to the device that was used to generate it: all other devices in the process were ignored. Now that the technology is in place to track and understand cross-device activity properly, we are able to analyse the contribution made by various devices at each stage of the purchase funnel. Aggregated across a whole campaign, this attribution modelling can give vital insight on the true value of each medium and platform within the marketing mix. At Space and Time we have been using various tracking tools to model attribution across different touchpoints and devices for years. The facility to attribute across device within Google’s Adwords product was introduced at keyword level in late 2015 and its usability and accuracy has been improving ever since, thanks to the performance of the algorithms and the volumes of data involved.
Google Premier Partner Awards
However insightful and interesting a solid day’s conferencing can be, it’s always nice to round it off with a celebration. Accordingly, while in Dublin we joined up with 350 other influencer agencies to take part in the Premier Partner Awards. The invitation alone meant a great deal, since it placed Space and Time Media within the top 3% of agencies in 26 countries across Europe, the Middle East and Africa. In addition we had our efforts and expertise recognised when we were shortlisted for the ‘Search Innovation Award’ for the work we did for one of our clients: Quality Formations.
From the full day spent at Google we gained plenty of insight and there was lots for us to take away and reflect upon.
Senior PPC Specialist