There’s this thing I’ve learned to call keyword bleed.
It’s when the search queries you mean to target with a specific keyword end up somewhere else — in a different ad group or campaign.
Unless you pay attention to your search query reports, you won’t even notice. But it’s happening.
It’s contaminating your data, and it might be costing you money.
Google provides a surprising amount of detail on how this works. They explain how keywords get prioritized when two or more match a user’s search query.
Ultimately, only one keyword can enter the auction and trigger your ad to show in search results.
For example, let’s pretend someone searches for best optometrists in des moines iowa.
Inside your Google Ads account you’re targeting these keywords:
+iowa +optometristeye doctor+optometrist near me
All of these keywords match the search. Which one does Google pick to trigger your ad?
Here’s an abridged explanation from the previous link.
- A keyword that’s identical to the search term is prioritized first.
- Second, when the keywords are identical to the search term, an exact match keyword is given preference.
- If neither of these situations occurs, they’ll select the keyword with the highest ad rank.
None of these keywords are identical to the search term. None of them use exact match as a match type either. That means Google will pick whichever keyword has the highest ad rank.
Keep in mind nobody knows the exact formula that calculates ad rank (well, somebody knows, but they’re not talking).
We only know the ingredients: your auction-time ad quality, the ad rank thresholds, the context of the search, any impact from ad extensions — oh, and your bid.
So if your bid on +iowa +optometrist is higher than your bid on +optometrist near me, then +iowa +optometrist is probably going to have a higher ad rank, and you might pay more for that search term than you meant to.
This isn’t my opinion on a hypothetical situation. Google describes a cheaper keyword with a higher quality score and ad rank as a “rare occasion”.
Meaning, even though they also said “Having multiple keywords that could match the same search term shouldn’t increase your costs in any way.”
I just explained how it does.
But let’s say Google isn’t possibly (probably) charging you more than you mean to pay per click.
The fact remains that, unless you explicitly tell Google how to handle when a search matches multiple keywords, they’ll decide for you, and this laissez-faire approach creates dirty data and paints an inaccurate picture of campaign performance.
The solution is to use negative keywords and force Google to follow your instructions instead of their own. They even recommend doing this.
It’s an easy thing to do. It’s also an easy thing to forget.
I see it all the time when auditing Google Ads accounts (Bing Ads too — it works the same way). It’s one of the first thing marketers forget when building a new campaign or optimizing an old one.
If I had to point out the most common mistakes related to keyword bleed, they’d look something like this.
1. Forgetting to Add Brand Keywords as Negative Keywords to Non-Brand / Generic Campaigns
It’s a great idea to target searches that contain brand terms separately from generic searches.
People who search for a brand perform differently than those who search using generic words. They’re in different stages of the funnel. As a marketer, you likely have separate expectations and goals for each group.
You’ll see some form of this segmentation in nearly every agency-built or enterprise-level account — campaign naming conventions that indicate a clear difference in their chosen keywords.
Let’s use Onitsuka Tiger as an example (they’re my favorite shoe brand).
They could have a couple of campaigns that look something like this.
Brand | Men's Shoes | Mobile
Non-Brand | Men's Shoes | Mobile
The campaign Brand | Men's Shoes | Mobile contains the keyword +onitsuka +tiger +mens +shoes.
The campaign Non-Brand | Men’s Shoes | Mobile contains the keyword +mens +shoes.
The first campaign is meant to target people who are searching for men’s shoes specifically from Onitsuka Tiger. The non-brand campaign is much broader — it’s targeting all kinds of men’s shoes. They could be nike mens running shoes, or mens casual slip on shoes, or luxury mens dress shoes — all because of how broad match modified keywords work.
Or, they could even be onitsuka tiger mexico 66 mens shoes in black.
So even though the difference in intent is very clear, Google’s logic allows brand searches to bleed over into the non-brand campaign unless you add brand negative keywords.
It’s easy and efficient to create a negative keyword list of brand keywords and apply it to all non-brand campaigns. You can also add them to each campaign individually, or at the ad group level if necessary.
2. Forgetting to Add Geo-Modifiers as Negative Keywords to Non-Local / Generic Campaigns
Geo-modified keywords are another popular tactic, especially for local lead generation.
When I used to work at an automotive advertising agency we would run geo and non-geo campaigns. They were almost the exact same campaigns. The only difference was that keywords in the geo campaigns had been appended with the applicable city or state (or state abbreviation).
It makes sense. Somebody searching for 2018 Silverado could be some 14-year-old kid looking for pictures of his dream truck. His dad who searched for 2018 silverado in cedar rapids iowa? He’s ready to buy.
This type of keyword and campaign structure functions the same as the aforementioned brand vs. non-brand campaigns.
Search terms with the city and state in the query will bleed into non-geo campaigns unless you add those words as negative keywords.
So if you’re targeting that dad searching for a 2018 Silverado to buy for his son, and you have a couple of campaigns that look something like this:
Chevy | Geo | 50 Mile Radius | All Devices
Chevy | Non-Geo | 50 Mile Radius | All Devices
Be sure to add cedar rapids and iowa as negative keywords to the second (non-geo) campaign.
3. Forgetting to Add Alternative Match Types to Match Type-Specific Campaigns
I’ll occasionally see accounts using a match type-specific campaign structure.
In these accounts, the same campaign has been duplicated three times. All three campaigns use the same keywords with different match types: One campaign uses all exact match keywords, another uses all phrase match keywords, and the last one uses all broad match modified keywords.
The idea here is that exact match keywords regularly perform better and should be prioritized first, followed by phrase match keywords, then broad match modified keywords.
In order to avoid keyword bleed, alternative match types for each keyword need to be added as negative keywords at the ad group level. Not all alternative match types need to be added though. This chart explains which keyword match types require which negative match type alternatives.
By doing this you prevent any searches from triggering a keyword with a less specific match type. In other words, if a search could trigger a phrase match keyword or broad match modified keyword, it will trigger the phrase match keyword.
Audience Overflow
All of this logic — this idea that, even though you intend to target people one way, Google may decide to target them another if you allow them — it applies to audiences too.
Remarketing is a great example of this.
Unless you follow the same thought process for remarketing campaigns as I just explained for keywords, your remarketing efforts might not be running how you think they are.
Here’s what I mean.
1. Forgetting to Exclude Remarketing Audiences From Non-Remarketing Campaigns
Google Ads has two targeting settings for audiences: observation and targeting.
When you add a remarketing audience with the observation targeting setting, Google continues to target users as usual but observes if they belong in your remarketing list. If they do, you can adjust bids for them (e.g. increase bids 25% for previous website visitors).
Audiences added with the targeting… targeting setting only target users in that audience. This is good if you want to show different ads to your remarketing lists than you do to in your regular search campaigns.
Neither targeting setting is right or wrong. I tend to use observation first — add remarketing as an audience to regular search campaigns to see what’s going on. If the data justifies breaking it out, I’ll build a separate remarketing campaign using targeting as the targeting setting.
However, targeting remarketing audiences exclusively in their own campaigns doesn’t mean you’re not targeting those same users in your regular campaigns.
Remarketing audiences can overflow into your non-remarketing campaigns unless you add them as exclusions, so that’s what you need to do. Add them as exclusions, just like you would with negative keywords.
2. Forgetting to Exclude Converters From Remarketing Campaigns
This practice varies by market.
The majority of my experience with Google Ads is with local lead generation.
Most lead gen accounts should exclude converters from their remarketing campaigns — they already converted, so why waste money retargeting them?
Other markets, such as ecommerce, are different. Previous purchasers are likely to purchase again — maybe not the exact same product, but they’re worth targeting.
You might want to target the two groups separately. In one audience you could have remarketing for previous website visitors who haven’t made a purchase. In the other audience, you could have remarketing for previous converters (people who have made a purchase).
Which brings me to my last point…
3. Forgetting to Exclude Deep-Funnel Remarketing Audiences From Upper-Funnel Remarketing Campaigns
Onitsuka Tiger sells shoes online.
Logically, all users who have made an online purchase have visited their website. This means everyone in their “converters” audience is also in their non-converters (website visitors) remarketing audience.
Unless you add the converters audience as an exclusion to the non-converters campaign, it’s possible someone who has made a purchase to see an ad meant for users who haven’t.
You can take this entire philosophy as far as you want.
I think that, beyond the examples mentioned here, you’ll start to see diminishing returns. But if you implement these your data will be clean, and you might save some money too.