An Expert’s Thoughts On How To Scale ‘Single Keyword Ad Groups‘ (SKAGs)

Our Growth Analysts at Ladder noted some issues with ‘keyword stuffing’ in a few Google Ads campaigns recently. Initially, several keywords that performed well previously were added to these campaigns in order to increase the volume – because we were finding it difficult to spend with the old campaigns.

The Problem With Keyword Stuffing

Keyword Stuffing is defined as putting too many dissimilar keywords in the same ad group. This is an issue because matching the keyword the user searches to the ad they see is important for performance – if multiple keywords are in the same ad group they all get served the same ad, which might not be relevant for what the user searched. Additionally, you’re sending users with very different intents to the same landing page, which might not be the best fit for all purposes.

Let’s imagine an ad group in a campaign targeting Los Angeles with 5 keywords with the following monthly volume of searches (impressions), clicks, conversions:

The problem with this ad group is that all of these keywords have different levels of search intent and different performance levels, so ‘stuffing’ them all in one ad group hurts performance and makes it harder to optimize effectively.

“LA hotels” drove 8 conversions (vs 10 for “hotels”) despite 1/5th of the search volume: this is an important keyword that deserves its own ad group. By splitting it out we can show an ad that has “LA hotels” in the title to improve clickthrough rate, as well as landing them on a page with just LA hotels to improve conversion rate.

Now, let’s look at the opposite extreme, which is also problematic…

The Opposite Problem: Excessive Granularity

For most accounts, the vast majority of keywords will not have very much traffic and therefore should be grouped together in order to have enough data to make optimization decisions. If a keyword only gets 200 searches a month, for example, you might only have 5-6 clicks per week (and 0 or 1 conversions) with which to guess the correct CPC for that keyword. It also takes time and adds complexity to build and maintain a larger account with more ad groups, so we should always be biased towards simpler structures.

*Bonus Resource: Cascading Significance: How to Make Data-Driven Gut Decisions

Let’s look at an example where we split every one of the keywords out into its own ad group (called ‘SKAG’s or Single Keyword Ad Groups). This approach allows us to have the benefit of giving a unique ad to every keyword, with the keyword itself in the ad copy, which increases CTR. It also allows us to design specific landing pages for important keywords, like for example “LA hotels”.

Here’s a look at some SKAGs:

Not only would this take a lot of time, but we lose some ability to maintain and optimize this account too. We have an ad group called “ventura beach hotels” that gets only 50 searches a month, 5 clicks and 0 conversions. How do we optimize this keyword on a daily, weekly or even monthly basis?

This is a simple example, but typically 80% of keywords will only drive 20% of all volume (and vice versa) so the majority of keywords in our account will be like “ventura beach hotels”. If you scale this from 5 keywords to 50,000 it quickly becomes unmanageable to split out too granularly.

How would we split these keywords out effectively then? Where is the dividing line?

Consider this:

There are two keywords relating to “Ventura”, and though someone searching for a beach hotel might be different to just searching in Ventura, the 5 clicks a month makes it not worth splitting out with separate ad copy or a landing page.

You can now bid this ad group based on a 13% conversion rate (2 + 0 / 10 + 5), allowing you to calculate an effective bid that is ‘good enough’ until you get more data. “Hotels” is a broad generic term with low intent and high volume, so we should definitely keep this separate. “LA hotels” is important enough for its own ad group, and potentially is a good candidate for a separate landing page based on volume.

“Venues near me” is marginal: it hasn’t driven a conversion yet, but it has had plenty of clicks and is different enough for potentially splitting out. In this case, I would try running it at a lower bid and seeing if I do get a conversion, and if not, fold it back into the “ventura” ad group.

PRO TIP: Why did I call it “venues near me” rather than “cheap venues”, which might be a different ad group to add it to? Well, you have to think about what you want in the ad, and if we don’t want to include the word “cheap”, it’s not a relevant word to split out. Whereas “near me” might be a good converter if we just add location extensions or a map as a landing page test. It all depends on your strategy; if you find people who search “cheap” never book at your price range, you should maybe split them out and bid them down, or consider excluding the keyword altogether.

Our Solution For Scaling SKAGs Efficiently

My preferred method for dealing with ‘keyword stuffing’ vs ‘excessive granularity’ is to use actual historical account data to make the decision. You set rules or heuristics for what is an acceptable level of traffic and conversions in order to make bidding decisions, then you only split out based on those rules, and combine if the keyword can’t justify separation.

Obviously, you can use estimates from the keyword tool to figure out a rough separation of head and long-tail keywords, but what really matters is how the keyword performs for your business.

Here are some reasons to use actual data:

• Your CTR might be higher or lower than the average predicted by Google
• You might have more credibility/a better offer than competitors on some keywords
• Your conversion rate might be higher or lower on specific keywords.
• The quality of leads or customers might be higher or lower on some keywords
• The intent differs for some searches and you might need different landing pages/goals
• There are many keywords you couldn’t have thought of before
• Keyword tool data is only a limited sample and might be inaccurate/out of date

The details.

So, we start with several obvious root keywords like “hotels” on a broad-match type in order to mine new keywords.

It will match to lots of different phrases we never would have thought of before and when those keywords meet a certain volume requirement, say 50+ clicks, or 3+ conversions, we would split it out in its own ad group (usually just as exact-match).

PRO TIP: In order to facilitate the building of these campaigns at scale, it’s essential to use a template. This will use concatenation in Excel/Google Sheets to automatically generate variations of each keyword and insert the keyword into the ad copy. Using a template cuts down the mistakes from building these campaigns and enforces best practice. It also makes maintenance and testing easier, because you can expect uniform structure and ads.

You can also go back and prune an account to make it simpler to manage.

If you review ad groups on a quarterly basis (or after large budget shifts) and find some keywords now qualify to be split out (because we’re spending more, or deserve to be cut back because we’re spending less, or performance has dipped), you can fold it back into the old ad groups.

Usually, I sort the keywords into buckets so I can see the progression between them and report on them separately. Test keywords are new, historic keywords that we aren’t sure are working anymore, or ideas from the keyword tool that looks worth it.

Then we kill the keywords that spend enough to hit the click limit but don’t hit the conversion limit. If a keyword doesn’t have enough data (less than 30 clicks) we push the bids up to get more clicks, or we keep it running at the same level for longer.

Test => Kill => Push => BAU => Top

Once we look into the keywords, we either kill or we move to BAU (Business As Usual) – which is our portfolio of keywords we can make effective bid optimizations about.

Top keywords are the terms that are responsible for the majority of our conversions: for these we build custom landing pages, write more custom (rather than templated) ads and watch performance very closely (daily vs weekly).

This method isn’t possible with the maximize conversions automatic bid-strategy enabledas you don’t have the fine-grained controls of manual bidding, but Google in some ways replicates this by pushing certain keywords over others based on historic conversion rates. In this scenario you can employ the same logic and just skip the pushstep, making sure you kill anything that has spent enough to drive 30+ clicks but hasn’t met the conversion threshold and is underperforming.

All in all, scaling single keyword ad groups will ultimately mean using them in conjunction with broader ad groups. This will maximize the effectiveness of SKAGs for only your best search queries while ensuring you’re covering broader intent… which is much easier to manage in large Google Ads accounts.

Interested in having the experts take a look at your Google Ads account? Talk to a strategist to take a step in the right direction.