A novel approach to PPC account optimisation: Keyword Type Analysis
Keyword type analysis can have many uses across different aspects of paid search account management such as keyword research, campaign structure, and bid optimisation.
This post will walk through what keyword type analysis is and how it differs from how we normally analyse keyword data. It can provide insight that individual keyword analysis may not be able to provide and this post will outline the methods for achieving insightful data.
What is a Keyword Type?
To best describe what a keyword type is, looking at a few example phrases is a good place to start:
- Buy lady gaga tickets
- Buy Beyonce tickets
- Buy green day tickets
We can see that while they are different, the structure of these keywords share a fixed component, and are unique in that they each specify the name of a different band / artist. In this example the fixed component of these keywords is "Buy tickets", and the three artists specified are lady gaga, Beyonce and green day. The fixed component can be described as a keyword type, the three artists can be described as variables.
So in the following example keywords:
- Opticians in Camden
- Opticians in Islington
- Opticians in Finchley
The keyword type is "Opticians in" and the variables are three different areas of London.
What is Keyword Type Analysis?
In the majority of paid search accounts, a large number of unique keywords within them can be viewed as a smaller number of shared keyword types. Many accounts will repeat a finite set of keyword types across a number of product names, locations or themes, and often these keyword types behave similarly across multiple keywords. This is where keyword type analysis comes in.
In instances where you have performance data for keywords, stripping the variable components from these keywords allows you to group unique keywords by keyword type and then isolate keyword type performance as an aggregate. This data can then be used to make meaningful account decisions, which can ultimately improve efficiency, volume and return on investment. The same is true of search query performance data as search queries can follow a similar pattern of fixed and variable components, instead of analysing keyword type performance, we analyse search query type performance.
How do you do it?
In order to transform keywords into keyword types, you will need to strip out every variable you have identified from a list of keywords and replace them with placeholder text. The Find and Replace function might work but with larger lists you will require a more time effective method. With lists of individual products potentially running into the thousands, automating the process is required. Using a list of variables and a list of keywords there are ways to strip out occurrences of the list of variables in the list of keywords by using Excel, which should provide you with an output of the keyword type associated with the keyword.
What are the Benefits?
By downloading a search query report and aggregating the search queries into search query types you can easily identify new keyword types that may match to your Phrase or Broad match terms but may not exist in the account activity at present.
You can also use search query type analysis to identify new variables which don't exist in the account activity that you might want to add into the account as either positive or negative keywords, any search queries which don't contain your placeholder text are matching to one of your keywords but don't contain any of the variables you have identified.
In the below example we can see the performance of a section of search query types for a campaign with keywords relating to nurseries in various locations. We can see that best child care in "x" has a high conversion rate, if I am not bidding on exact match versions of this keyword type, this might be a useful addition to my keywords. We can also see that "nurseries in croydon" has performed well, even though Croydon has not been identified as being in my list of variables, therefore I may want to add a set of "Croydon" keywords if this location is relevant to my account.
When constructing a campaign build for a new product or service, you won't have the benefit of past performance data to guide the ad group structure. But if you have the performance data of other, similar products where search user behaviour might be similar, you can use the aggregated keyword type performance data as a surrogate to inform your new campaign build. This can be especially useful where the life cycle of an individual product is short, in which an efficient campaign structure can make or break end performance.
When launching a new campaign, the average CPC for a keyword type can be used to determine your initial bids for a keyword that falls within that type.
In another example below, we can see the top ranking exact match keyword types by conversions, if I am building a campaign for a new location (say, croydon), I would want to make sure these keyword types are included in the build. In addition I can use the keyword type average CPCs as benchmark to set initial Max CPCs for my new keywords. If I know that the average difference between my bids and my actual CPCs is around 15%, I can adjust the keyword type average CPCs upwards.
There are circumstances in which keyword type performance data can be used to make optimisation decisions. However, in most cases the individual keyword performance will serve this purpose better. The keyword type performance may not necessarily be a good indicator of performance for the individual keywords that belong to that type, since they will still be different and we expect a natural variation.
That said, here we see some less common terms, if we look at keyword type private nurseries in "x" the combined spend for these terms is over 75, and none of them have converted, this may be the impetus we need to lower the bids, or pause these terms, as they don't seem to be converting. If we look at the keyword type angels nursery in "x", one of the keywords of this type has converted, but the combined CPA is high, it may be that we can expect a higher conversion rate for that term at 1 location, but in other locations our offering is not attractive; it may be beneficial to limit use of this keyword type to 1 location.
Search User Behaviour
Keyword types offer an additional method of analysing changes in search user behaviour over time. By recording the traffic or performance of keyword types over different time periods, you can recognise the seasonality, growth or decline of different keyword types, which can inform account strategy.
As with keywords and search queries, elements of ad copy are often repeated across multiple products. By aggregating individual ad performance to ad copy type, you can test new ad copy types with the goal of increasing CTR across multiple product ads.
Below we have tested three different headline types, across multiple products, with the same description lines 1 and 2. Where we may not have enough data to make rotation decisions for each of the ad groups which share these ad types, on aggregate we can say that type 2 is the headline most likely to get a higher click through rate.
The larger the account, the more valuable the time saving ideas become. Isolating data by Keyword type is a method of simplifying data in circumstances where we can find a similarity between keywords.
It is hoped that the user behaviour across these keywords is likely to be similar, and that the decisions made using this data are beneficial to many aspects of a paid search account and may dictate future strategical decisions.