%%PLG_FT_BLOG%%

When we introduced single keyword ad groups in 2009 everyone thought we were nuts. Illustrated best by this quote from a competitor “Finch is splitting campaigns into atoms and it becomes so complex it is impossible to manage.” Why did we do it? We did it for a very simple reason: it enabled us to isolate the variables that impact performance. The purpose of isolating those variables was to build an algorithm and data model combination to dramatically increase the accuracy of setting the bid to gain more leverage and reduce risk.

The ad group level provides the most accurate and granular way of isolating variables such as match type, PLAs, and bid modifiers: device, audience, location, etc.  When combining all these it builds a click-profile in real-time as you enter your next click auction. When having only one keyword/PLA in the Ad Group you will gain very granular control of performance, which is key for optimizing.  

Having multiple keywords in an ad group poses a number of challenges that are worth considering:

A/B testing of ads: Let’s say you have 25 keywords and 5 ads in an Ad Group and you want to optimize the click-through-rate (CTR).  You change the 2nd line and enter a different call-to-action statement.  Three days later you come back to see how that change performed.  Maybe the CTR stayed the same when you look at it, but in reality maybe the change helped 4 keywords dramatically but hurt 6 others.  You will never be able to catch these scenarios unless you isolate the ads by keyword. 

Match type separation: I will use an example to illustrate why using Ad Groups to create match type separation works better. Let’s say you create a Broadmatch+ keyword in AdWords and insert it into your campaign. A few days later you take a look at how it is performing and then make adjustments based on the initial performance. But maybe not so fast, if you go look in the Keyword details report you will find that it was not always served up as a Broadmatch+, but rather i.e. ⅓ as Broadmatch+, ⅓ as Phrase and ⅓ as Exact. Google will use their discretion to serve up what they think is the best for you. The problem with that is of course that different match types perform very differently and lumping them into a group like this make the optimization very inaccurate. The same principle applies for PLAs in your Shopping campaigns. By separating keywords into a single Ad Group by match type you obtain max control of the performance, and you can keep track on the performance on an Ad Group level. The silver lining on this is that the bid modifiers also sit on the Ad Group level.  This strategy is forcing Google to pick the Ad Group vs the Keyword/PLA, and you gain the control you need to perform.

Bidding by the average: This is the result of lumping things together instead of creating a granular data structure for your campaigns. Many advertisers set bids at the Ad Group level, and when you have a lot of keywords in the Ad Group the bidding is based on averages - this leads to average results. Even if you set the bid on a keyword level inside the Ad Group, you will run into the problems described above.

This is a very complex strategy to execute and there are a lot of details that needs to be thought through. It is however possible to fully automate how this is managed and a key part to approaching AdWords Search and Shopping in a Programmatic approach. Figuring out how to identify process and obsessively automate the execution is what we do at Finch. Does it work? Take a peek at this to see examples on before/after performance when getting granular and automated.