Uncover Hidden Negative Keywords With These Three Advanced Strategies

PPC Samurai Sam Superhero

by Shaun Bond

August 25, 2020

Our top 3 techniques for uncovering hidden negative keyword opportunities

In their regular account optimization checks, most account managers include a scan of their search term report to find and block non-performing or irrelevant search terms. While this is a high value exercise, you could still be missing a critical opportunity to find hidden layers of negative keywords that are quietly burning through your client budgets.

Take your analysis to a deeper level with these three (advanced) techniques; so let’s get search term mining, PPC Samurai style!

Blue one in a circle

Bottom Up

Blue two in a circle

Top Down

Blue three in a circle

Don’t miss the skyscrapers

Blue one in a circle

Bottom Up

Blue two in a circle

Top Down

Blue three in a circle

Don’t miss the skyscrapers

#1: Bottom Up

Different campaigns will usually have different conversion rates and CPA/ROAS results, so begin by checking that terms are performing well relative to their peers in the same campaign (not looking at account level).

Start search term mining through the search term report, campaign by campaign. This ensures the metrics (eg Conv Rate, Cost/Conv and Conv Value/Cost) you’re evaluating for each search term are relevant within their cohort. You can use a pre-built workflow in PPC Samurai to do this, or kick it manual style.

Why campaign level and not account level?

If you look at the account level search term report, metrics that are reported against each search term lose relevance because they are being sourced from different campaigns with different settings and bid strategies.

For example, a search term that triggered from a bottom of funnel (or brand) campaign should have much better metrics than a search term that triggers from a top of funnel campaign, so comparing their metrics against each other in the search term report can be deceptive.

Get yourself set up with the right reports and data sets

  1. In the Google Ads interface, go to the campaigns view and set the time period for 90 days. Sort by campaign name.
  2. For the first campaign, calculate how many clicks should be required before an average search term produces a conversion:
Calculation Campaign

Example: If my 90-day (campaign) Conv Rate is 5%, I would expect an average search term to convert by the time it has 100/(0.05*100) = 20 clicks

    1. In a second tab, open the Google Ads interface and go to the search term report. Set the time period in this tab to All Time.

First test:

  • Sort your search term report by Clicks so you can check any terms that have more than double your calculated number of clicks to conversion.
  • If they have double the clicks and zero conversions, they could be good candidates for adding as negative keywords. In the above example, any search terms that have zero conversions but more than 40 clicks should be looked at as potential negatives.

Second test:

  • Sort your search term report by Cost so you can check any terms that have spent more than 2x the 90-day (campaign) CPA but have no conversions.

Third test:

For CPA based campaigns
  • If the search terms have some conversions but their CPA is more than two to three times the 90-day (campaign) CPA, they should be looked at carefully as potential negatives.
For ROAS based campaigns:
  • If the search terms have conversions but their ROAS (Conv Value/Cost) is less than half the campaign average, they should be looked at carefully as potential negatives.

#2: Top Down

We have assessed our Search Terms from the bottom up, but now it’s time to flip that and look from the top down.

At PPC Samurai we’ve helped thousands of agencies design and build their standard operating procedures, so we know that a vastly underutilised strategy across agencies is a search term strategy we call the “top-down aggregated scan”.

The goal of this strategy is to aggregate the performance of each search term across campaigns with similar CPA goals, and then test its aggregated performance against a reasonable performance benchmark. It follows that this strategy works best for lead-gen clients where your desired CPA is relatively constant.

You’ll rarely need to run this scan across brand campaigns because branded keywords generally tend to match closely with the searches they trigger. So, filter out display and brand campaigns to define an accurate and reasonable CPA and Conv Rate benchmark for non-brand search campaigns.

Get yourself set up with the right reports and data sets

  1. In the Google Ads interface, go to the search term view (for the full account) and set the time period to 90 days.
  2. Filter out brand & display campaigns.
  3. Take note of the CPA & ROAS for the “Total: Filtered search terms”.
  4. Calculate how many clicks you would expect to see before an average search term produces a conversion
Calculation Aggregated
  1. Change the time period to All Time, and ensure that your visible columns include Cost, Conversions & Clicks. Download the search term report.

Prepare your spreadsheet

  1. In the spreadsheet, remove the rows above the metric column headers and the total rows at the bottom of the sheet.
  1. Click on Insert > Pivot Table and select the options radio buttons below, then click OK.
  1. This should open a new tab in your spreadsheet, and that tab should contain a blank pivot table.
(If you don’t see the “PivotTable Fields” pick box on the right, go to PivotTable Analyze > Field List and select the Field List.)
  1. Next, click and drag “Search Term” to the “Row” box, and “Clicks”, “Cost”, and “Conversions” to the “Values” box.
  1. Ensure all three items in “Values” are all set to “Sum of”.
  1. Your pivot table should look something like this now:
  1. Now sort the table by clicks, largest to smallest

First test:

Now it’s time to look at the conversion column, checking for any terms that have zero conversions but more than double the number of expected “clicks to conversion” you worked out before.

In this client example, the “clicks to conversion” worked out to 6.8%, meaning the average non-brand search term should convert after 15 clicks (as worked out above in the ‘Get yourself set-up section’). If we double that to a threshold of 30 clicks, you can see 3 terms that have had more than 30 clicks but no conversions. Time to add some negatives!


Second test:

Next, move on to a test that looks for “zero conversions but spent >2x average CPA”.

To do this:

  • Sort by the cost column (large-small)
  • Find any terms with zero conversions but a cost that is greater than 2x my average CPA.

For this example, the average CPA for the non-branded cohort is $65. Given that, three search terms surface as good candidates for adding as negatives; they have spent more than $130 (being 2x the CPA) but have no conversions. In this case they happen to be the same three culprits as the first Top Down test, but they can be different, so run both tests.

Why is it important to do a top down analysis?

Consider the following scenario:

Search Term Listing

If you only ran a bottom up analysis, none of these “guitar lessons” searches would have triggered for review as, individually, each row was below the campaign level #clicks to conversion and 2xCPA tests.

By aggregating the data, we see a very different story. “Guitar Lessons” has triggered in multiple different campaigns and adgroups, effectively spending $333 with no conversions; yep, time for a review!

#3: Don’t miss the skyscrapers

In this new brave world where Google is promoting automated bid strategies combined with loose matching between keyword and search terms (aka intent matching), it’s becoming more common to see the occasional VERY expensive search term triggering in your account.

One of our PPC Samurai users recently started running the skyscraper workflow just to see what it would do, and they were shocked to find one search term that was costing them $793 per click (vs an average CPC of $80)!

Horror Face

Um…Why hadn’t they spotted that? The search term was hiding underneath a modified broad match keyword that had high click volume, so the effects of this huge account anomaly were masked by the overall cost of the keyword.

So how do you find these anomalies easily?

  1. Navigate to your search term report.
  2. Set the time period to 7 days and create a filter to display only those terms with clicks >=2.
  3. Sort by Avg. CPC
  4. Now click on each campaign, one after another, and check if the average CPC for your searches exceeds 10x the Total: Campaign average CPC.
  5. Once you have checked all campaigns for 7 day anomalies, move the time period out to 14 days and repeat the process through all campaigns individually.
  6. When the 14 day checks have been done, move to 30 days and repeat.

We like to do this campaign by campaign to ensure the search term CPC’s are being compared to a relative cohort and are being treated under the same set of parameters (bid strategies, audiences etc).

Hopefully you don’t have any searches that pop out, but if you do, it’s an opportunity to either add them as a negative keyword OR potentially add them as a single keyword adgroup (exact match) and add sculpting negatives to push traffic on that term to the exact keyword. This allows you to control the bids on that search term much more rigidly.


Performing these scans regularly will undoubtedly uncover hidden opportunities for you to save your clients money, or at the very least, flag search terms that are silently undermining your CPA or ROAS performance so you can take action on them.

We’re big advocates of the awesome BMM/Exact campaign strategy, as you can easily incorporate these checks into your daily routine; adding further value to your accounts and moving you one step closer to being indispensable to your clients. Ultimately, isn’t that what we want for them and for us?

If you’re NOT already using BMM/Exact, you can check out our free training course here!

Subscribe to the newsletter

If you enjoyed the article, sign up for our newsletter to get these regularly sent straight to your inbox!

Was that helpful? Here are some other articles