Conversational search has been around for a while now, and although that has not immediately changed the paid search landscape, I believe that in the near future it will, and I’m going to run through why I believe that to be the case. The questions I’m asking in this blog post are why does it need to change and what will that mean for paid search?
Our existing model is simple. We display relevant ads based on the search terms used and attempt to match user intent to our products and services. Short head-terms provide a high volume of searches but, with less information in these short queries, we can’t match the query to a specific product. Middle to long-tail keywords are more specific, so we can take them to highly relevant products or services, but volume is usually less.
Take the example of a search for ‘jewellery shop’. Where ‘conversational searches’ build off the original query, without re-stating the original keyword, this starts to cause problems for paid search as it stands, in that there are multiple signals that are not currently supported in the infrastructure, but more crucially there is a lack of a discernible keyword after the initial search. At the moment, within our normal paid search query behaviour, we expect a move from short ‘head term’ to a longer more refined, more intent laden search query that allows us at each stage to deliver a relevant advert, and ideally more tightly focused as the queries build up, with appropriate and cost-effective bidding. With our ‘jewellery shop’ example that might look something like;
Search 1 - Jewellery Shop
Search 2 - Jewellery Shop London
Search 3 - Ethical Jewellery Shop South London
The last query would allow us to deliver an ad that contained several elements to hook the searcher and ensure they are going to find a relevant and useful service for them.
For mobile the head terms have proven to be even more important as the search queries can often be shorter than on desktop, due to the restrictions of typing on smaller devices. As devices become more sophisticated the length of mobile searches may increase, but conversational searches will lead to longer, more fragmented, query strings rather than simple search terms. For our example, but this time using conversational search, it may look more like this, where the first query contains the keyword, the rest do not.
Search 1 - Jewellery Shop
Search 2 - near me
Search 3 - with an ethical approach
We could easily bid on the first query using any of the match types available, but exact, phrase or modified broad would not work on any of the subsequent qualifying queries.
There is a solution to the ‘conversational search’ problem that we already see with search terms on pure, non modified, broad match. Those keyword types can trigger impressions that are ‘session based’. In these cases your ads can appear within a user’s search session, matching to the broad keyword of another search within that session, even though the new query has nothing to do with the original search to which your broad match keyword was matched.
Most decent PPC managers will use exact phrase and modified broad match types, with pure broad match a lesser used tool. We want tight matching and more control over our keywords, so much so that even losing control of close variants, as happened last year, led to howls of derision.
The future of mobile search will have to be pure broad match, as the most valuable searches, further down the query path, will not contain the original keyword around which we currently build our account structure. This will be the single biggest shift of control from the advertiser to the paid search platforms we have seen, far exceeding the removal of our ability to create device specific campaigns.
The future of mobile search is poised to be pure broad search.
If you think that this sounds far fetched, bear in mind Sergey Brin’s original vision was this;
Quite obviously this is not compatible with the existing paid search structure and breaks the current model. Google will still want to serve you paid ads of course, but the keywords you bid on in this version of events would be broad or maybe even completely dynamic, pulling information from the website.
Why do I think this is likely to happen? Firstly because of the move towards mobile, which lends itself towards the type of voice and conversational search where this is most problematic. Secondly because when you start to factor in relevant signals such as motion, location within individual buildings, various context signals and the analysis of personalised trends and patterns in user behaviour, there simply isn’t going to be a way to sensibly take all of these into account. And thirdly, because it is in Google’s best interests. If Google can take away control from advertisers and dictate when ads will appear, using all their vast data, then they will be able to improve performance for the vast majority of advertisers and ultimately continue to grow profits.
Many people will say that Google won’t be able to manage that mess more effectively than they could, but that may not be the case if recent investments into AI are anything to go by. As money is being poured into this area of research by Google and Baidu most noticeably, with Google paying $400 million for DeepMind, and Baidu boosting R&D by 85 percent to $284 million, the power to run automated paid search campaigns with minimal human interaction will encroach on the existing model.
Baidu are embracing artificial intelligence.
AdWords Express has been around for a while now, as have Dynamic Search ads within AdWords. Both are able to work out your keyword targeting and pull ads based on the pages of your website, but are pretty crude tools at the moment compared to a well run and professionally managed account. However they don’t need to become as good as their human counterparts just yet to have a major impact.
If Google can provide a dynamic platform that works effectively, a turbo charged AdWords Express, then there are a lot of businesses out there that would be able to put more budget into AdWords. I’ve reviewed enough accounts over the years to know that basic mistakes are still made that make accounts inefficient. This is particularly true of accounts that are run by over-worked small business owners, who have had to learn about AdWords on the go, or employees who are given the responsibility to run the account with no previous experience and plenty of other responsibilities to distract them, but it is also sadly true of many agencies. If Google can create a system that allows businesses to move away from the high maintenance keyword based accounts that drain time and resources, but crucially are able to do it with equal or better results, then Google should be able to welcome many more advertisers to their auctions.
Who knows what 10 years may do to the industry? If we look at some of the experiments going on now, there are hints at a complex, signal rich system, with rapid testing, greater personalisation and, in the case of this test below, reactive technology. How would a paid search account look with some of these layers of complexity baked in?
Is artificial intelligence the future of advertising? (The Guardian).
However the longer term pans out, in the short to medium term the changes in technology, and the greater comfort of users interacting using voice based searches with smart assistants that provide context and complex signals to those searches, will require changes to how we manage our accounts.
Perhaps we will see a version of AdWords where users are given the Express option for free, but the larger advertisers will be able to afford a premium model, with more control.
Anyone who has worked with a Google manager will know that the advice/mantra is “go broader, spend more”. In a scenario where you have to go broad to be in the race for the best converting clicks, where it is harder to keep spend down due to the broad approach, it will be more essential than ever that the management and monitoring of accounts is in good hands. Tight structures, solid ad testing, use of all the available features and great attention to detail will be at a premium. Many advertisers hitting Google for the first time have been stung by the use of broad, and that is where agencies and PPC specialists have always been able to prove their worth, refining flabby search query reports into trim, focused accounts. In the medium term future, with conversational search giving more and more power to broad match, that value will only increase.
I started with the question why does paid search need to change, in light of the rise of conversational search, and hopefully I have made a compelling argument that we will reach a crunch point, where keywords are no longer fit for purpose. How the industry reacts and takes paid search forward from that point is clearly a more difficult one to answer. I believe that machine learning, backed by the vast stores of data that Google has at their disposal, will lead to a more dynamic ad system, where we will control our budgets, discounts and offers, and the pages we wish to promote, but Google will dictate when the ads appear to a far higher degree than at present.
Where I think this leads us is to a happy marriage of organic and paid interests, where landing page experience becomes the new Quality Score. Websites with rich content, fast loading times, easy navigation, with reviews, special offers and benefits all in an easy to use format, will be rewarded with better paid search performance, focusing our digital efforts into creating great websites. Ultimately wouldn’t that be an online world where everyone benefits?
If you have any thoughts on this topic, or on the broader question of the evolution of paid search then I’d love to hear from you in the comments.