In this piece, we’ll share advanced ways to assess your Google Ads accounts and improve performance. 

These are all methods used by our proprietary PPC audit tool. It’s designed for internal use, so those methods rely on some degree of coding. However, we’ll also share broader techniques to assess your performance, break down your PPC data and take action.

For more basic auditing techniques, read our PPC audit guide and checklist here.

  • We’ll discuss the top ways to audit account features including how to use NGrams to review keyword opportunities
  • Break out performance clusters to improve your bidding
  • Check for keyword overlaps at scale
  • Review opportunities to improve quality score
  • Slice and dice as many segments and metrics as possible

About our PPC audit tool

Our audit tool is only available to our own PPC consultants, but all diagrams here are outputs from it. The tool performs both standard PPC checks and new to give innovative insights into accounts. With the aim of inspiring both new and advanced PPC specialists. 

For the coders: The tool, built in Python, uses a combination of simple packages. It integrates with data direct from Google Ads and uses:

  • pandas and sklearn for the data manipulation and machine learning requests
  • BeautifulSoup for scraping and parsing
  • Seaborn and matplotlib for the visualisations. 

Every single audit and ad account is different, so its robust design allows our team to add new modules. It also outputs files with new segments. This allows our analysts to take away the data and dive deeper where necessary.

1. Reviewing keyword opportunities with NGrams

An NGram is a phrase, made up of N number of words, that can be found within the text. 

For example, “dog barked loudly” within “The dog barked loudly at the postman”, would be a TriGram, whilst “at the”, or “the postman” would both be examples of BiGrams. 

Break down the Conversion Rate (CVR), Return on Ad Spend (ROAS) and Cost per Conversion (CPC) of these phrases to understand key performance trends.

For example, OneGrams help you to understand the impact of words such as “free” or “cheap”, which might feature across many hundreds of keywords. On their own, the keywords might not make up much spend. But, grouped together, you may find that makes up a large chunk of your costs. 

These are particularly useful for shopping campaigns. Here, a large proportion of unique search terms will have one or two impressions.

An output sheet like the below lets you filter down to phrases that have 0 conversions. This helps to identify poor performers that you might want to negative. Or it lets you find the highest ROAS phrases. Here, you might want to expand your coverage or work the phrases into ad copy.

NGrams showing near me

These need to be broken out programmatically, which is where our auditor tool comes in handy. 

2. Improve bidding by breaking out performance clusters

“Performance clusters'' mean groups of keywords that perform similarly in terms of ROAS, CPA, and seasonality. Find and group these to improve bidding on the account.

First identify keyword themes. For example, “raincoats'' might form a performance cluster around a particular CVR. Sales across this cluster may all increase in the winter. Although, in the UK, the seasonality for these may well be all year round!

Our keyword grouping tool uses a machine learning technique called K-Means Clustering to identify groups. It finds the terms that look the most similar across several performance metrics.

The different colours in the chart below illustrate performance clusters the tool has identified in one account.

3D SQR Grouping

However you choose to do so, find these groups and you can do a number of useful things with your audit:

  • Rapidly understand a new account and what keyword groups work well
  • Check if bid strategies are set up the best they can be
  • See if the account needs a restructure.This way of looking at keywords is structure-blind. It looks at individual keyword performance, rather than the performance of campaigns imposed.

3. Check for keyword overlaps

Keyword overlap in an ad account is so important, as any PPC expert will need no telling.

You need to check that searches are going to the most specific keyword they can do, and do so at scale. 

There are a few ways to check this across an account:

  • Look at the number of unique ad groups triggered by your top searches
  • Measure the proportion of spend going through certain match types.

The below chart has columns for spend across brand and non-brand campaigns. The light and dark blue indicate the proportion of spend on brand and non-brand at the search term level. This shows if the search term actually contains what you consider to be a “brand” search.

Brand and Non-brand overlap

Next, zoom in on those search terms in the wrong column to learn why they are appearing in the wrong place.

The same applies for searches appearing across a range of match types. A more advanced audit may identify if search terms built out in exact are appearing across other match types. This indicates you need to create more negative keywords to make the search go through the keyword you want.

In a similar vein, the light blue in the below chart is search terms that built out as exact-match keywords.

As much of the light blue as possible should fall within the “Exact” column (as it is below). If it is not, you might be having an issue with your cross-match negativing.

Keyword spends by match type

4. Quality score opportunities

Read our guide to quality score here. This explains what it is, and how to improve it. Quality Score is generally a good indicator of account health.

An excellent indicator of an account's opportunities from quality score factors is a quality score matrix. The below chart shows an example of this. 

Impressions by Quality Score Factor

You should be trying to get as much spend as possible into the top right. Aka “Above average” on both. This gives you a good objective target to aim for, as well as a guide on where you need to improve.

The auditor tool actually goes more in depth on this to help us isolate exactly what ad groups we need to focus on. 

It scrapes the ad copy, keywords and landing page and produces two completely new metrics per ad group. 

  1. Match Score - The average proportion of word matches between the keyword and ad copy. Improve how your ad groups score on this and your “Ad relevance” will improve. 
  2. Landing Page Score - Measures the number of matches between the keyword and the landing page, weighted by SEO prominence. It also uses TF-IDF. This is a measurement of how frequently the word more generally appears. Downweighting “the”, for example.

This gives us areas where we can make targeted recommendations. For example that some ad groups need more granular landing pages, with X and Y keywords appearing on them.

The below boxplot visualises the impacts of Match Score.

Boxplot of Word Match Score by Ad Relevance

You can see the very clear relationship between the Match Score, and the Ad Relevance that Google applies. 

To be more precise, less than 25% of “Below Average” ad groups have a Match Score above 0.5. Whilst over 75% of the “Above Average” ad groups have this.

The real power this tool gives us is the ability to hone in on those ad groups and fix a factor we know to be fixable. 

By improving this tangible, fixable metric, we know it will have a big knock on effect on quality score. Start on those ad groups with a below average ad relevance and a low match score, and raise the bar across the account.

Use our free Page Insights Chrome Extension. Get topline SEO improvements for your PPC landing pages and increase quality scores.

5. Slice and dice other metrics

Opportunities will exist all around the account. Break down every segment you can to identify where CPA is high or ROAS is particularly low. 

This will let you hone in on the problem areas, and by diving in deeper you can understand how to fix that.

On top of performance by device, by search network, hour of week, performance over time, you can look at more advanced factors like performance by number of words in each query.

The latter has highlighted some unique opportunities for our accounts. For one specialist client, we found that targeting keywords of >3 words in shopping campaigns led to a 28% increase in ROAS.

Clicks by Hour of Week
Clicks by Search Query Length

For example, are CVRs poor through mobile? This might reveal an easy-to-fix issue with the checkout process on your m-dot site. Or, it could suggest that you need to exclude mobile entirely.

Make your process of looking at these as quick but as thorough as possible. 

This is where the auditor comes into its own. Applying these granular tests across every account. 

It provides overviews, attaching labels to areas that need a closer look. And ultimately gives our experts the insights they need to make impactful changes to PPC accounts. 

These are insights which can and do make impacts on spends running to the millions on large PPC accounts.

We’ve delved into some pretty advanced techniques here. Get in touch if you’re interested in learning more. Or if you have any questions, or are looking to run a PPC audit on your own campaigns. 

Our experts are always on hand if you’d like a more in-depth personal analysis or a bespoke strategy. Contact the team, we always love to talk about PPC.

Learn about our free, personalised insights club

More Paid Media

Forget ROAS – ecommerce brands should focus on POAS

21 June 2021

A complete guide to managing PPC bid strategies

02 June 2021

How to conduct a PPC audit

11 May 2021