This week we have another in depth PPC Clinic review from Shane in response to a query that Julie had when she wrote in to ask about how granular she should go when it comes to organising keywords for PPC campaigns:

“Hi Docs.
 
When creating a PPC campaign for a retail website, what is the best way to organise the various permutations and combinations of keywords?
 
Take the following ‘widget’ example – all of the following ‘widget’ related keywords should be bid on, as tightly within the account structure as possible:
 
‘widget’
‘buy widget’
‘cheap widget’
‘blue widget’
 
That would be logical I’m guessing – an ad group for each.
 
But how do you account for the combinations e.g. ‘buy cheap blue widget(s)’?
 
Then, as is often the case, how do you factor in synonyms of ‘widget’ which people will also be searching for?
 
I know the importance of negative keywords to tighten each ad group in this situation, but without having an optimal account structure in the first place it kind of feels like wasted effort.
 
I know this is a problem because the keywords that get the most impressions and clicks in my campaigns tend to be broad and phrase matches rather than exact.  That said, I do get a good CTR on the broad and phrase, and nearly always have a QS of 10 (a few 7’s but really not many, and none below 7).
 
This feels like the missing piece in my PPC puzzle so any help much appreciated!
 
Thanks.
 
Julie.”

Here’s what Shane had to say …

Shane

Hi Julie.

This is a good question and one which crops up even in well set up and well managed accounts, basically just how granular do you go, at what stage and what’s the best way to do it?

Split into Themed Adgroups

We would definitely split the account down into themed adgroups, so in your example, yes, we would make those separate adgroups for sure: 

‘widget’
‘buy widget’
‘cheap widget’
‘blue widget’
etc …

For combinations of multiple theme phrases like “Buy cheap blue widget” then there are three options:

1) Create a ‘fishing’ test adgroup and load all the terms (but link them to the most relevant landing page for each term via the individual keyword url option). See which of these terms gets enough volume to create other smaller adgroups from them, and then either move those keywords to the most relevant existing adgroups on a keyword by keyword basis, or to newer adgroups if enough of a common theme with volume came up.

Or

2) Add the terms straight to the most relevant adgroup (isn’t always clear cut, as some could go in two or more different adgroups i.e “cheap” or “blue”)

Or

3) Create lots of small and very targeted adgroups knowing most of the work will be a waste of time.

The problem in large scale PPC is to know how granular to go from the outset; you want to cover the main terms, but avoid wasted time and maximise the number of sales in a short timescale whilst you are working on the campaigns. If you think about it, you’ll see the data is there for you to figure out in most cases.

Taking “blue widget” as an example, let’s say this has 1000 searches per month, what’s the odds that “buy blue widget” is more than this!? … very little! In fact, with every variable added, you’ll find the search volume usually drops significantly, so you know this will be a lower volume term. Taking it one step further, “buy cheap blue widget” will be even less often searched, so you can quickly figure out the order of priority when breaking the keywords down by looking to the search volume of the next most popular phrase up.

“Buy cheap blue widget” – A Case in Point

So Julie in your “buy cheap blue widget” example: depending on how big the volume is in the first place for “widget” and “buy widget” etc., you could end up with the following combined variables:

action, price, colour, item etc.

… then you could figure out all the “action” words such as:

buy, find, compare etc.

… then the “price” ones such as:

cheap, cheapest, discount etc.

… then the “colours”:

blue, red, yellow etc.

… and finally the item and variations, synonyms like:

“thingybobs”, “thingymajigs” etc.

Now imagine the scenario in doing this by creating an “optimal account structure” from the outset. When combined into two, three and four keyword combinations, that’s around 350 adgroups just to cover these four variable themes.  In most cases, many won’t result in any impressions let alone clicks or sales (we used to do it this way years ago when 2000 keywords per adgroup was a target, not a redundant feature as is the case now).

The solution is in which way to tackle the workload, we hate messy structures too, but optimal account structure is what you strive FOR at a point in time; usually always ahead as the landscape is constantly changing, but at the outset you can have “an” optimal account structure for now that will do to get going, and then quickly refine and expand as you go.

The main point I’m trying to get across is that you can’t always start out with it as optimum unless you are willing to waste many hours (in some cases hundreds of hours).  Once you step away from the main terms and add variables to keywords to other variables, most of what you do from guess work won’t actually get many impressions in most cases. So you can either sit down and create huge campaigns with every possible variation of multiple keywords, tens of campaigns, hundreds of adgroups and tens of thousands of keywords, OR you can do the volume terms first, then get those live and break it down based on volume into smaller adgroups. Once you’ve done that, you can keep re-refining those adgroups to smaller ones as new keywords bed in etc..

In your “buy cheap blue widget” example, I’d be tempted to do just a couple of shorter keyword adgroups first to see what traffic is out there:

“cheap coloured widget”
“buy coloured widget”

Basically, we would see which of the new terms gets a viable volume of searches, and then create groups out of those starting with the most searched and clicked terms. So if “price blah blah widget” gets more traffic over “coloured blah blah widget”, create your adgroup to be predominantly price based and put the lower volume coloured yet priced keywords in there too (but link to the coloured landing page by each keyword url). Do this until such time as there are enough keywords and/or traffic in them to earn the right for an adgroup of their own.

Broad, Phrase and Exact Match

When you say you get most clicks on broad and phrase, move your mindset to think like this:

Every broad match click you get from now on is a phrase match you missed that you could be getting a better CTR for in a tighter adgroup with relevant adcopy going to a targeted landing page (*some will be clicks you’d wish you had on negative matches too!) and paying less for overall.

Every phrase match click you get from now on is a click you missed adding as an exact match that could be getting a better CTR in a tighter adgroup with targeted adcopy going to a targeted landing page (*some will be clicks you’d wish you had on negative matches too!) and paying less for overall.

The higher the CTR, the lower the cost per click as google earns more, so even though you may have a QS of 10, you can still lower that click cost with more highly targeted matching. This will benefit the overall health of the campaign, as CTR of the url itself is taken into account with QS in conjunction with the health of adgroups, campaigns and the account as a whole.

Synonyms

As for synonyms, use the search query reports to see exactly which terms are driving traffic. Search google for synonyms etc. and then do adgroups for these based in order of the search and/or click volume on the main “widget” variations you already have. You’ll find you have to stop after a few adgroups as the volume is too low. There’s no real point doing every adgroup exactly copied from “widgets” campaigns if the main synonym for it is “thingamabob”. If it only gets 10% of the volume as a keyword as widgets does to begin with, you’ll just bloat your account up with useless adgroups which is pointless work.

On the plus side Google is beta testing a new matching option called modified broad match. I blogged about this recently at http://www.revenueaddict.com/broad-match-gets-cleverererer-good-news-finally.html. This should help in the targeting as you could go ‘+cheap widget’ or ‘+cheap blue widget’ and catch all variations on a broad but more targeted match, as the term after the “+” must be in the search term, but the rest is still broad match. However, it’s no excuse for not running frequent search query reports to identify the keywords you need to phrase and exact or negative match.

Hope that helps!