The Future Of Ads In AdWords

Ok, I don’t really know for sure what the actual future of ads will be in AdWords, but I’m reading some things in the tea leaves lately that have gotten me thinking about it. So, here is my bold prediction – ads will soon be dynamic or responsive only. I know, that makes me hyperventilate a bit just thinking about it too, but my spidey sense is starting to tingle about more and more forced automation coming down the pike.

Let’s examine the evidence, shall we?

Exhibit A

I’ve been thinking this for a little while now, ever since the Google Display Network did away with text ads and is now pushing advertisers HARD to just create responsive ad units. From the AdWords Answer page on GDN responsive ads:

Responsive ads automatically adjust their size, appearance, and format to fit just about any available ad space. For example, your responsive ad might show as a native banner ad on one site and a dynamic text ad on another, as it automatically transforms itself to fit precisely where you need it to go to meet your advertising goals. As such, responsive ads can increase your reach and impact while also saving you time. (emphasis mine)

That line I bolded “increase your reach and impact while also saving you time” – get ready to hear that a lot in the coming months. This is going to be the narrative that is pushed right along with all of the “innovations” and “improvements” that I expect we are going to see along these lines. It is clear to me that AdWords wants to push this narrative because it pretty much tries to quash the three most common objections to using AdWords in one tidy clause.

Exhibit B

This week’s post by Matt Lawson of Google entitled “Three Foolproof Steps To Excellent AdWords Ads” really got me thinking about this topic, as I think its message is quite clear.

From the post:

Subtle nuances and customizations aren’t limited to emails or text messages, though; they can also be applied to your AdWords ads. Think about how many impressions your ads serve each day. Now think about how much you could improve your performance if you set up your campaigns so the AdWords system customizes your entire ad unit the same way you do your emails or texts.

You want your readers to be receptive to your message, and, with a bit of work in AdWords, you can utilize the system to tailor each and every ad to a user and their auction-time context. (emphasis mine)

This is not referring specifically to GDN ads, this is about all ads in AdWords.

Further from the post:

Provide a bunch of great components and set them free

A great ad unit is a combination of your ad and your ad extensions. But what extensions to show when? The best way to deliver great auction-time ads is to create a whole bunch of great components, then let the system pick between them using all the data accessible to it. (emphasis mine)

Are you starting to see where this is going yet?

First foolproof method espoused? Optimize your ad rotation for clicks and conversions. This part is complicated. But one thing I think is pretty clear from this post (from an official Googler) is that Google really, really wants you to let them handle decisions about which ads are the better performers. And to a certain extent, it does have access more natively to information that is harder to compile and analyze manually:

An optimized ad rotation is a competitive advantage for you. It allows the system to choose the ideal ad for each context. Maybe one ad does better on mobile; maybe another does better in a certain geography; maybe another does better at a certain time of day. Maybe all of those performance trends are different when those settings are combined. A great mobile ad in the morning might be different from a great mobile ad during lunchtime.

Ok, I can see some of where this concept is coming from. Being able to take advantage of machine learning to apply these types of things to your advertising is/would be quite cool. Here is where the wheels start coming off for me though:

I know you’re all capable of figuring out these trends for yourself. But I also know that it’s impossible to actually implement all of those insights at the time of each auction. I totally get (and even applaud) people’s deep appreciation for split-testing ads, but I think the A/B approach to message testing is becoming outdated. If A works better here and B works better there, let the AdWords system serve them where they’re already known to perform better.

Yes, but I think quite a few PPC professionals can talk at length about differentials between what AdWords would deem as a “winning” ad versus what a client might, based on AdWords data + their data.

It goes on further:

Once you’ve started optimizing your ad rotation, you should factor in % served to help decide the outcomes of ad tests. If you already have five ads in an ad group, and something is barely serving at all, it’s an indication that you can probably remove that messaging and cycle in something new.

When managing your account, you should use a combination of metrics to decide which ads to leave running. CTR, % served, total impressions served, CPC and average position can all be helpful in determining which ads could be replaced. (emphasis mine)

I get that the idea here is that the system is using all of its magical knowledge to always serve the best ad in every situation, which in and of itself is not a bad thing, but using these metrics to make ad decisions, makes me nervous. No mention here of conversions as a factor for determining ad serving either?

Next section – Provide 3 to 5 Ads Per Ad Group

This most certainly bucks against the model of A/B testing. The logic for doing so is as follows:

It might seem like overkill, but as long as you’re optimizing your rotation, there’s almost no downside to enabling multiple ads per ad group. The system is going to choose what’s known to perform best for the context, and different ads appeal to different users at different times. The more ads you provide, the more options you’ll have to show the ideal message for any context.


Our internal numbers at Google bear this out, too. When compared to ad groups with one or two ads, ad groups with three or more ads can receive up to 15 percent more clicks/conversions, as long as you’re optimizing your rotation. Those extra options for messaging are valuable. (emphasis mine)

The “our internal numbers bear this out” without any more detail drives me insane. Mostly because it feeds in to the idea that all AdWords advertisers are homogeneous, that there is no difference between industries, business size, locations, etc. In my opinion, this does a disservice to all advertisers by implying that they are all the same and that their markets within AdWords are all exactly the same – we all know that they are not.

And, then offering up 15% more clicks/conversions. I am troubled by the use of clicks/conversions – as if they are interchangeable measuring units. Conflating clicks with conversions is also a disservice to advertisers. I suppose there could be a scenario where a click would be considered a conversion (I mean I can’t think of one, but I’m allowing for the possibility) but generally, a conversion is something that happens AFTER THE CLICK.

Here is a tiny tidbit that I think is among the most interesting in the piece, relative to my premise:

And one quick tip: use IF functions to make each of those ads even more dynamic for devices or audiences.

Mark this down – the IF function is going to become a big deal as the push toward responsive only ads happens. It is going to be the place where you can still have things behave or display differently under a limited set of circumstances.

Last of the 3 – Implement Every Ad Extension That Makes Sense

Extensions are a part of Ad Rank. They’re also great for performance, as ads with them are typically clicked a bunch more. And they’re great for users, as they’re helpful and add more information to an ad unit. And they’re already optimized to drive the best performance for you. (emphasis mine)

So, using extensions or not does impact your Ad Rank or as it could be called the “dark quality score” as it is what is used to determine your eligibility for each auction, how much you will pay for a click, ad positioning, etc. Interestingly, according to AdWords Answers on Ad Rank, it also factors in determining your eligibility for extensions…

Ads with extensions are clicked on more. Ok, are they quality clicks that lead to conversions or are they just more clicks? To me, at least, that really matters.

Extensions are already optimized to drive the best performance for me? What defines that “best performance”? Is it clicks or something more meaningful?

And concluding the piece:

And the process of serving a customized message in each of those impressions is very straightforward, so you can easily start delivering auction-time ads. Think about the beauty of communication in the digital age and its endless nuance, then bring that same level of nuance and customization to your creative messaging.

I think the message is quite clear here. I fully expect that the AdWords system is going to be morphing into one where we enter basic information, perhaps use a few IF statements to drive to specific landing pages or to choose what is seen on a specific device (maybe) and fill in all of the extensions and then the system will take it from there. It might start to have beginning “what are you trying to do” choices like GDN does now:

GDN Objectives

These are the options you see when creating a new campaign in GDN right now


I really don’t want this to be where things are going. I like being able to control more things (as many of us PPC pros do). But, it just is seeming more and more to me that this train has certainly left the station and is not only not going to slow down, I expect it to pick up speed over the next 12 to 18 months.

Where I think this could be potentially amazing would be if you could provide feedback into the system that would integrate actual client data (like what ads convert for them the best based on their internal numbers) and have the whole machine learning infrastructure use that data, along with the data described here to make dynamic or responsive or whatever they will be called ads. Part of me will always bristle at losing direct control over things, but if advertisers could provide more data into the system to influence the system’s choices for their particular account, that could have some exciting possibilities.

What about you? What do you think? As always, sound off in the comments or hit me up on Twitter (@NeptuneMoon).


  1. Having been an advertiser with Google since the inception of Adwords I can say with confidence that there has been an inverse relationship between anything Google automated and our overall ROI.

    Is the algo better than our personal experience with the brand we are advertising and the deep insights we have into the target audience’s behaviors? Does our ad campaign have enough traffic to make the algo even competitive?

    • Neptune Moon says:

      Thanks for sharing your thoughts!

      I agree that for professionals who manage PPC and want to be able to make their own decisions, based on more than just data from within AdWords, this is not a move in a positive direction. For advertisers that want more of a “set it and forget it” type of thing, this will appeal to a lot of them.

      And, as we see with “close variants” and even in choosing a “winning” ad, the machines are not always right.

Speak Your Mind