RKG Logo

At RKG we sometimes pine for the good ole days of Yahoo/Overture and the flat account structure. Each keyword had it’s own title and copy block, we didn’t have to play games with breaking up ad groups to change copy on different collections of keywords…ah simplicity. Granted, that structure is the ONLY thing we missed about the pre-Panama platform… but I digress.

The adgroup structure makes a great deal of sense for the engines: much less data to store, much more efficient performance, and it even carries the side benefit of helping retailers “organize” their campaign. In fact, in the content advertising realm, adgroup structure is important to performance and we recommend paying attention to the Google/Yahoo recommendations.

The problem is rigid hierarchical schemes don’t work very well in search advertising. It turns out that keywords often have many different attributes. Classifications might run along the lines of product category and subcategory, but also manufacturer, material-type, holiday association, sku specific vs general, gender, etc. Forcing all terms into neat campaigns and AdGroups limits flexibility.

Instead, we parse groups by the whole collection of attributes, grouping only those terms that share all attributes, not just two of them as campaign - adgroup structure would suggest.

Why is that an advantage? It allows us to make copy changes across any dimension or combination of dimensions without having to break up adgroups and recombine them later. Having a big sale on “Metal widgets”? Great, done. Oh, it’s only on “Acme Metal Widgets?”, no problem done. Need all the “Apex Automotive Seat Cover” terms going to a separate landing page, no worries. Because each combination is clustered as atomically as we could ever want we spend less time creating and combining groups.

What’s the drawback? For us the only drawbacks are that periodically the engines tell our clients that we “don’t use adgroups correctly”. Our clients occasionally complain that they can’t find things easily in our accounts because there might be thirty different ad groups related to “Women’s shorts” and they don’t know which one is hiding a particular keyword.

Those drawbacks don’t really impact us in our management. The combination of APIs and Adwords Editor make the Engine User Interfaces almost irrelevant, and as we provide full service management, while clients are welcomed to poke around in the account, they certainly shouldn’t need to do so.

More importantly, from an analysis perspective, campaigns and adgroups play no critical role in our system. We database everything on our side, so analyzing data by product category, manufacturer, color, gender or some combination thereof is a breeze.

We’d make the case that tying the classification schemes to the campaigns and adgroups is problematic on many levels, particularly if performance analysis — and heaven forbid bidding — are tied to that structure. What happens to these analyses and bids if you have to split or combine adgroups to get promotional copy tied to the right terms? Bad business.

While campaigns and adgroups can provide small programs with some sort of organizational framework, a large, sophisticated program needs to be much more flexible and much less brittle than the two-tiered hierarchy allows. The data: costs, sales, keyword classifications, etc have to live on your servers to get the maximum analytical benefit.

Having to explain what we’re up to to each new engine account rep is annoying, but the performance gains are worth it.

George

If you like this post, consider subscribing to our RSS feed. You can also have new posts sent to you via email.

Share this post (via email, Digg, Delicious, etc)

Possibly Similar Posts

Trackback

http://www.rimmkaufman.com/rkgblog/2008/01/25/adgroup-headaches/trackback/

Comments

  1. Alan Rimm-Kaufman, January 25, 2008:

    The only drawbacks are that periodically the engines tell our clients that we “don’t use adgroups correctly”.

    Different visiting Google reps give different (contradicting!) advice on best practices for adgrouping, at times even contradicting advice in the Adwords online docs. If we had consistent advice from Google, we’d follow it of course; as it isn’t consistent, that’s tougher!

    As George points out, these nuances aren’t critical to us for core (non-content) campaigns, as our own database schema is richer than the three tiered Campaign >> Adgroup >> Keyword, and our algorithm groups for high QS.

    Sigh.

  2. George Michie, January 25, 2008:

    Okay, a wag might ask: why are you having this discussion on the blog instead of walking down the hall and talking to each other? I dunno.

    I respectfully disagree with my friend, collegue and…boss!

    He writes: “If we had consistent advice from Google, we’d follow it of course”. I’d say instead: if doing it differently helped performance we’d do it differently.

    It’s important to note that as much as we respect our collegues at Google and Yahoo, these folks haven’t seen much backend data and don’t really know much about retail marketing. Google told us for years that our way of doing copy tests was wrong, now they recommend doing copy tests the way we do them.

    It’s critically important to recognize that the engines’ interest and their advertisers’ interests are different. We’ll doggedly pursue our clients’ interests, but part of doing that is taking the engine’s advice with a grain of salt.

  3. Shelley Ellis, January 28, 2008:

    George, I do not long for those old days of Yahoo although I understand your logic. ;)

    And I SO agree about the engine’s interest and the advertiser’s interest. I always do what is best for my client in spite of what the “best practices” page says. My theory is that if you do the same thing everyone else is doing (following the search engine’s advice) that you will get the same results that everyone else gets. I don’t want average results…I want success for all my clients!

Your Comment

Tags

RKG: ,
Technorati: ,

Email Updates

Categories

Recent Comments

  • Nick Stamoulis: Great tips during this crazy time. People are most likely going to pull much of their PPC advertising as a whole these days but the...
  • Namu: Thank you for the great tip. Now I can read my starred items on my iPod touch offline!
  • Ricardo Figueiredo: Thank you for the good advice Alan. It's intriguing to observe how there are some businesses, and individuals, feeling the...
  • George Michie: Chris, I wouldn't be surprised if that's a real number. Inc says they have 550 employees, so their income would have to be $50...
  • George Michie: Hi Christian, I suppose they take the same percentage hit off their commission that the retailer does. To my thinking it's the...
  • Chris Zaharias: I read the magazine on a flight Sunday and recall seeing iCrossing on there at ~~$100M in revenues, and thought the same thing of...
  • Alan Rimm-Kaufman: Christian -- I didn't mean to imply all retailers will face Q4 losses. But it is not improbable that many retailers will be...
  • Christian Little: Despite the economic crisis, how could most retailers be facing a Q4 loss? For most retail this is the best time of the year, you...
  • Christian Little: That's pretty remarkable...makes me want to build a coupon site lol. Don't coupon sites take a huge hit in commissions though...
  • Stephen Schramke: Sage advice. Thanks for sharing!
  • George Michie: Could be Neil. I have my doubts. My suspicion is that there just isn't much work being done, other than taking commission checks to...
  • Neilzb: Those numbers are pretty remarkable, but if I had to guess I would say that it’s possible that they are just 8 people 'outsourcing' full...
  • Jeff Cornejo: I disagree that a revenue/employee ratio shows ANY kind of profitability. If anything, a mostly-passthrough model, with high...
  • George Michie: Hi Dan, The IP address of the advertiser isn't a factor, anyone can run geo-targeted ads regardless of where their website resides....
  • dan shipe: Hey, me again. What about possible exploits to this system? Adwords must evaluate the geographic region based on the IP address of the...

Blog Stats

  • Posts: 758
  • Words: 336,078
  • Comments: 1,346

Administration

Close
  • Social Web
  • E-mail
Powered by ShareThis