So, machine learning in AdWords is great and definitely the future. But Dynamic Search Ads? Not so much, in my opinion.
In this column, I would like to talk about Dynamic Search Ads (DSAs), which are generally for people new to paid search, and I’ll share a handful of caveats.
When to use DSAs
Dynamic Search Ads are generally for beginners and are used in the following situations:
- If you have a very high-volume account, and your website is basically perfect in terms of SEO. (However, even if perfect, most SEO sites are far from optimal.)
- To pull in some new keyword ideas to a managed account.
- As a catch-all, or to sweep up extra traffic.
- For an Inventory Management Feed (IMF), where you don’t have the time and money to do this manually or use bespoke technology.
- If you are new to paid search or have absolutely no time to manage campaigns.
Outside of those situations, I can find little reason for using DSAs. I am sure Google will crack it one day, but that day is definitely not here yet.
Why DSAs fall down on the job
Whenever I audit an account and see dynamic search ads, I know there is much opportunity to improve performance. A well-managed account will always outperform DSAs. Here’s why.
1. DSAs make bad headlines
Anyone who’s ever used DSAs understands what I’m talking about here. It’s actually quite impressive how far Google’s machine learning has come, in terms of generating a high proportion of relevant, legit-sounding headlines. Still, with any DSA campaign, there will always be mistakes.
Websites are full of opportunities for a nonsensical, off-key sounding piece of ad copy. Sometimes copy is taken out of context, where it might have made sense within the website; sometimes random words get thrown together; sometimes sub-domains somehow make it into the headline. Bad headlines get fewer clicks, and lower click-through rates (CTR) drive up costs per click (CPCs).
Not to mention the damage it can do from a branding perspective.
2. DSAs don’t find the best keywords
There’s no doubt that Google’s artificial intelligence (AI) is thorough; a dynamic search ad campaign will reliably include just about every relevant keyword for your website. However, it will not necessarily find the highest-performing variants.
A typical example comes from a client’s account. An important search term for their account was “memory foam mattress,” which DSAs did not pick up.
To a human, this sort of cultural intuition comes quite naturally, but Google’s machine learning is not advanced enough to share this intuition.
Dynamic search ads pull in a lot of long-tail queries, and they won’t all be relevant. You can add negative keywords and targeted exclusions, but if you haven’t got a good setup, it might not see good performance.
This also means it’s likely to be more expensive than manual campaigns since the ads will normally be less targeted than for keyword-targeted campaigns.
3. DSAs limit control
This is the most important point. Businesses need to be able to use paid search flexibly. If you’re in the retail sector, sometimes there are outstanding inventories you need to push, which needs to be reflected in the structure of your account.
Many companies update their websites on an hourly basis, which is problematic because dynamic search ads are based on website scrapes that occur roughly every two days.
Ad copy and bidding decisions sometimes depend on factors outside of what’s measurable within AdWords. The targeting options associated with a DSA account — cost per acquisition (CPA), return on advertising spending (ROAS), maximize conversions, enhanced CPC — may fail to take into account wider brand objectives.
The headline of an ad needs to be shaped by a brand’s objectives. Advertisers need greater control than DSAs permit.
4. Personalized automation is the way forward
Until Google’s machine learning improves significantly, I recommend you stop using DSAs.
In my opinion, it might be decades before the situation changes. And even if there are major breakthroughs in Google’s natural language processing and automated bidding functions, DSAs won’t truly be superior to a human-managed account until AI is at the same level as human intelligence.
Google may keep tweaking keywords like exact match and shopping, and over time, it might get harder for human-managed accounts to have a clear advantage. But from the past six years of running a PPC agency, I can tell you the average growth rate of our accounts has not been achieved by using DSAs.
Do you agree? If not, I challenge you to an A/B split test!
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.
About The Author
Daniel Gilbert is the CEO at Brainlabs, the best paid media agency in the world (self-declared). He has started and invested in a number of big data and technology startups since leaving Google in 2010.