THE RKGBLOG

Google Household Income Targets Nice, Not Quite What Advertisers Need

Google offers demographic geo-targets for the United States based on average household income (HHI) for advertisers to use in setting bid modifiers based on the wealth of the area users are searching from. At RKG, we’ve fully explored how these income level targets behave, how much traffic is attributed to them in different set ups, and their usefulness in the future of geo-bidding.

Here, we discuss our findings on these levers, as well as offer up suggestions for how Google can make them more useful for advertisers.

HHI Targets Exist Among Geo Targets

HHI targets are comprised of zip code level geographic locations, and Google has organized these targets to exist among the other geographic locations a campaign is targeting. This means that bids for traffic that hit an income level geo-target will only be adjusted using the modifier for that income target, as opposed to hitting a geo modifier and then the income level modifier. In other words, HHI modifiers and standard geographic location modifiers are not stackable.

For campaigns targeting the United States and HHI percentiles, Google is only able to assign between 50-60% of traffic to income brackets, with the rest of the traffic rolled up to the national level:

Google HHI Attribution

We’ve confirmed with Google that this is because they are currently unable to assign a zip code level location to all clicks, and because not all zip codes are assigned an HHI. A disproportionate amount of the traffic that is not assigned to an HHI is likely from mobile, as mobile makes up 60% of traffic that is recorded at the less precise state and national levels in Google’s API click performance reports, compared to just 20% among traffic recorded at the city level.

The Location-Income Bracket Hierarchy

When we have added city, state or zip code locations to a campaign with HHI targets, both the income targets and national level location targets of the campaign are ignored in favor of the more granular location targets. Bids will then be modified based solely on the value of the modifiers for these more granular locations.

Thus, if HHI targets were added at the United States level, and then more granular targets such as city, state or zip code were added, the HHI bid adjustments would be largely ignored for traffic coming from the more granular locations added.

Because Google appears to completely respect city and zip code location targets, advertisers would then theoretically be better served in setting up zip code level locations for as much of the country as possible and baking income level data into the values of the modifiers placed on each of these locations. This would allow you to include additional location factors that differ in importance by advertiser in calculating the value of the geo modifiers.

For example, if you’re selling pool supplies, temperature is likely another important aspect of a geographic location that would impact performance. Thus, you may want to include temperature data along with income in calculating the value of each geo modifier.

BUT…

The most granular locations Google passes via API in its click performance reports are at the city level. In order to automate geo bid adjustments using third party conversion data, then, advertisers have to use city level locations, not zip code.

Income, however, varies much more by neighborhood than by city, and performance varies significantly by income.

To the first point, the charts below demonstrate this by pitting the percentage of traffic attributed to each HHI target in Google’s data against the amount of traffic that can be attributed to those same brackets using city level locations data from Google’s click performance reports (Google’s current help page on the subject omits the income values of the 41-50% bracket as a result of privacy concerns and imprecise data, making it difficult for a comparison. However, earlier Google reports of the income bracket values including all HHI percentiles were published by bloggers, which I use here):

Clicks by Google HHI Target

Clicks by Location ID HHI

As you can see, 19.5% of traffic is attributed to income brackets of $96K+ using Google’s zip code level income targets, while only 5% of traffic can be attributed as such using city level income data. This gulf exists despite the fact that 47.2% of Google traffic is attributed to the national location, compared to just 11.9% of the traffic passed in the click performance report being reported at the state or national level only.  Using city level data via the click performance report, 83% of traffic is attributed to incomes between $0K-$96K, compared to just a third of traffic being attributed as such within Google’s zip code based HHI targets.

All of this is due to the fact that city average incomes are much more, well, average than zip code incomes.

As for how much performance can vary by income, here’s a look at overall account wide conversion rate by income bracket for one advertiser.

CR by HHI Target

As you can see, conversion rate increases with each step up in income level, and clicks from the top 30% in income all convert at least 50% more often than clicks from the bottom 50% in income. Using city level location targets to adjust bids on those cities with incomes in the top 30% would impact just 1.9% of total traffic, compared to over 11% using the zip code level data Google uses for it’s HHI targets.

Needed Changes

This demonstrates how important it is to advertisers that Google begin to pass zip code level geographic information in its click performance reports, in order to properly refine bids at the most granular level. Google must also increase their current limit of 10,000 geographic modifiers per campaign, as launching 10,000 zip codes will only allow advertisers to make adjustments for 25%-30% of traffic (or less, depending on which zips you launch) for campaigns targeting the entire United States.

Income also isn’t the only geographic attribute that could vary significantly in impact more by zip code than city. For example, advertisers’ brick and mortar locations and those of their competitors are likely to have an impact on online performance. But if there is only one physical location in a large city, the impact of that location is likely different for those who live right next door compared to those who have to hop on the highway to get there. Advertisers need zip code level location data for this reason.

One less attractive but still helpful possibility would be for Google to re-configure how HHI targets and other geographic targets interact. If Google were to make the modifiers for actual geographic locations and those for HHI brackets stackable, as well as pass advertisers HHI reports to enable automation in the calculation of HHI target modifiers, the two could be used together for optimization. This is still not as desirable as receiving zip code level data in click performance reports, however, as advertisers would have to split out the impact of income on geographies in calculating modifiers for the standard geographic location targets.

Give Advertisers the Power

While HHI targets may provide advertisers levers with which to optimize bids based on income, advanced marketers want to do much more with geo modifiers than the current setup allows. Current temperature, proximity to water, brick and mortar location, shipping center location, proximity to universities; all of these performance changing factors and more can be used in calculating granular location modifiers that could provide real benefit to those who can leverage them properly.

Before any of that can happen, however, Google has to hand over the keys to their most granular geo data. Anything less diminishes the huge potential geo modifiers have within Enhanced Campaigns.

  • Andy Taylor
    Andy Taylor is a Senior Research Analyst at RKG.
  • Comments
    5 Responses to “Google Household Income Targets Nice, Not Quite What Advertisers Need”
    1. Nice writeup. This is what we have been finding in using location intelligence to optimize bidding strategies. Furthermore, using a more granular indexing system allows you to define and target “geographies of opportunity” that cross the boundaries of ZIP codes.

    2. Andy Taylor Andy Taylor says:

      Thanks Chad. Absolutely agree that properly bidding geographic modifiers hinges on the ability to properly assign and analyze the impact of performance defining attributes across geos. Much like long tail keywords, most granular locations will not have significant data in the recent past with which to make bid adjustments, and so must be grouped into buckets of shared attributes.

    3. Brian says:

      I am a bit confused, it seems that you can pull zip code performance data in the dimensions section, and you can input zip code level modifiers, is this something new?

    4. Andy Taylor Andy Taylor says:

      Hi Brian,

      If you’re using Google’s conversion tracking for bid optimization and manually handling all of the calculations and implementations of your geographic modifiers, the issues outlined in this post likely won’t be a big deal to you and the data you speak of is perfectly fine to use and has existed for a while http://www.rimmkaufman.com/blog/google-adwords-tracking-zip-code/02062014/. But, in order for advertisers using bidding platforms to automate geo bid adjustments using third party conversion data, we have to use city level locations instead of zip codes since the location IDs passed through the API only go down to the city level. Google could solve this issue by making the zip code data it already has (that you’re seeing in the dimensions tab) available through the API for those who need it for automation.

    5. Or better yet, Google could move to radii which can be much more accurate than zips. This is where Bing is actually doing better than Google.