My new company makes a fairly significant deal out of multiple word order combinations when building out keyword lists, even for broad match. Like making sure to have both “brown shoes” and “shoes brown”. How have you found it to matter? Is there a formula in excel that is useful when doing this?

I found scant information when searching for “word order” and “keyword order” on all major PPC sites…

I’m sure Brad will give you the exact answer, but I’ll chime in with my experience.

This is just for broad and modified broad. For exact and phrase, this discussion is obviously different. I believe your question is about broad matching, right?

My experience is google’s matching algorithm is far from perfect and hence it can behave differently when matching the (modified) broad version of ‘shoes brown’ and ‘brown shoes’. You probably will have different QS or CPC for this keyword in my experience. I can’t explain the technical reason why but I am guessing it’s because google treat them as two different keywords in its inventory since the info google holds about these two keywords is different (due to what other bidders have been doing).

I used to use different word order bidding in the past in the hope I could exploit any imperfection in the algorithm if I could. However, I was faced with the following challenge:

Assume ‘brown shoes’ is not performing well due to any number of reasons and you decide to reduce the bid. Adwords can now start matching to the other keywords instead and you can’t reliably measure the effect of any changes you make to either keywords. This is true for bid changes, Ad text changes, etc. You can’t use negatives here to minimise the overlap as this is broad.

As I needed to accurately measure the effect of my optimization I decided to avoid keyword overlap as much as possible.

Maybe there is a solution out there. The above is just my experience.

Thank you for sharing your experience, that is helpful. As I understand you, basically when you reduce the bid of one word combination on broad match – it could start showing for another variation you are bidding on that has a higher adrank / higher bid, etc.

Yes, correct. And that is a challenge as you never get to see the exact effect your bid management or Ad testing is having.

What I’ve seen is that when I raise my bids, I show for more variations, and when I lower then I show for less. So, usually when I lower them I don’t show for different variations in most cases, just a smaller subset of what I was showing for.

With broad match, trying to measure the true effectiveness of all those keywords variations at various bid levels is about impossible. You could do it if you have waaaaay to much time on your hands. I think you’re better off adding the exact/phrase/modified combinations so you have more control over the display and setting your bids at those levels.

As far as needing: brown shoes, shoes brown, shoe brown, brown shoe all in broad match – that goes to the heart of AdWords – how much time do you have?

If you really want to show for something – then add it. However, it gets to a point where you are adding so much stuff that there’s no reasonable way to measure and bid on items without losing site of something else you’re working on (like testing).

Usually, I don’t bother with every possible combination of broad match. I usually start with one combination; and then as I find the actual queries, I add those as keywords so I can set bids based upon the query and then let broad match find new words for me.