Audience Match Rate Optimization and Data Enrichment

Profit Optimization

Audience Match Rate and Data Enrichment

Recent and ongoing changes to privacy policies and privacy technologies have had a wide ranging impact on digital advertising, especially those on social media platforms.


Advertising platforms like those offered by Facebook, Google and TikTok work best when there is a high quality "data-feedback loop" (DFL) between a person clicking on an ad, and the subsequent actions that person takes after they click on that ad. Examples of this include:

Click on Ad > View website content

Click on Ad > Add an item to a cart

Click on Ad > Initiate a checkout

Click on Ad > Make a purchase

Without this feedback loop, the advertising platform doesn't know what actions a person takes after they click on an ad, and therefore doesn't know if the ad is effective. For example if 100 people click on an ad, and the advertising platform doesn't know if any of them made a purchase, then the advertising platform doesn't know if it is showing the ad to the best people in the first place.

Optimizing Ad Performance


When there is a highly functioning data-feedback loop, advertising platforms can optimize the performance of an ad by showing it to the people most likely to take a given action, eg a purchase. That's because they know which people clicked the ad and which of those then makes a purchase, and they use their internal data to show ads to more people like that.


But if the data-feedback loop is broken or of poor quality, then ad optimization is significantly reduced and advertising costs increase.


Since the introduction of the Apple iOS 14.5+ update, and the move away from cookies in browsers, this data-feedback loop has been interrupted and advertising platforms are less able to optimize ad performance.


This has resulted in higher advertising costs for business owners.

Optimizing Audiences


Another benefit of a high functioning data-feedback loop is the optimization of 'audiences'.  Two important examples of this are:

  1. 1
    Retargeting (custom) Audiences
  2. 2
    Lookalike Audiences

Let's look at each one in turn.


Retargeting (custom) Audiences


A retargeting custom audience is built from the actions a person takes after they click on an ad.


For example, if a person clicks on an ad and views some content on a website, then the advertising platform can include that person in an audience of people that viewed that content, and then advertisers can show specific ads to that custom audience.


This only works if the data-feedback loop exists between the advertising platform and the website.  


If there is no data-feedback loop, then the people who view the content on a website can't be added to a retargeting audience.


Since the changes with iOS 14.5+ the data-feedback loop has been interrupted, and it is much harder for advertising platforms to build retargeting audiences. This means that retargeting audiences have become significantly reduced in size, ad performance has dropped, and advertising costs have increased.


For example, instead of having 1,000 people in a retargeting audience, now there might be only 250.


So, not only are there less people to advertise to, and therefore less revenue coming from retargeting sales, but these smaller audiences tend to see ads more frequently, and so advertisers have to refresh their ads more frequently.


Lookalike Audiences


A lookalike audience is an audience that advertising platforms creates that is similar to an existing audience.


For example, an advertiser may have an audience of people who have purchased a certain product, and then uses that audience as a 'seed audience' to build a 'lookalike' audience of people with a similar behavioural profile.


The idea is to show ads to people who are similar to the original group of people that have already taken the action of interest.


The problem with new privacy policies is that the custom audiences that were used as 'seed audiences' to build new 'lookalike audiences' are now much smaller, and so the quality of the lookalike audiences built from them is also reduced.


The result has been that lookalike audiences perform poorly and the cost of advertising to them has increased.

Summary of the Problem


With the introduction of new privacy technologies and policies, the data-feedback loop used by advertising platforms to optimize ads and audiences has been interrupted. The direct result has been a reduction in ad performance and an increase in advertising costs.


The specific reasons for this are:


  • Less data with which to optimize ad performance due to a broken data-feedback loop
  • Reduced size of retargeting audiences due to poor audience match rate
  • Poor quality lookalike audiences due to the reduced size of custom audiences

Together, these three issues have made advertising more expensive and have caused advertisers to reduce their advertising budgets, and in many cases, pull back from advertising altogether.


This is a major problem impacting business owners globally, and there has been an urgent need to find effective solutions that are compliant with new policies.


A real solution to these problems would be highly valuable, especially if it leads directly to improved ad performance and reduced advertising costs.

So, what are the solutions?


There are SIX key solutions to the above problems.


  1. Integrate compliant digital fingerprinting (DFP), zero party data (ZPD), first-party data (FPD), and third-party data (3PD) to restore the data-feedback loop for ad optimization
  2. Improve audience and event match rate quality for retargeting and lookalike audiences
  3. Improved top of funnel ad performance through 3rd party data
  4. Improved top of funnel ad performance through creative and ad copy
  5. Improved middle of funnel ad performance through strategic creative and ad copy
  6. Diversify traffic sources and include alternative advertising platforms
  7. Direct email outreach through the compliant resolution of site visitors identity (Identity Resolution)

Let's go over each of these in a little more detail.

1. Integrate compliant digital fingerprinting (DFP), zero party data (ZPD), first party data (FPD), and third party data (3PD) to restore the data-feedback loop for ad optimization.


Digital fingerprinting (DFP) is a collection of anonymous fragments of data that users leave behind when then visit your website. These fragments include the users operating system, computer or device type, font, browser ID, and User Agent.


While none of these fragments can identify a person, they go some way to identifying a behaviour that is carried out on that computer or device. This information can be used in isolation or in combination with other types of compliant data to improve the data-feedback loop.


Zero party data (ZPD) is data that a user volunteers in addition to their name and email.


ZPD doesn't identify a person, but does provide information about them. For example, if a user chooses a Large item of clothing instead of a Medium, then that is ZPD. If a user fills in an application form, then that is also ZPD. If a user interacts with a chatbot, or sends a message, then that is also ZPD.


Like digital fingerprinting, ZPD is collected on the website by the website, and can be combined with other forms of compliant data to build a picture of the user and included in the data-feedback loop.


First party data (FPD) is collected by the website and is Personally Identifiable Information (PII).


This includes information such as first and last name, email, and phone number. FPD is collected when a user fills out an online form or survey, provides their name or email in a chat bot, or completes a purchase. 


FPD is owned by the website and the website requests permission to save this information. For example, on this site you can read our privacy policy and cookie policy to see how and when we collect FPD, and when visiting this site users are presented with a popup about the information we collect, giving them the opportunity to decline.


Third party data (3PD) is collected by a third party, or seperate business, and is often sold or leased to another business to use for marketing purposes. Third party data providers must show that they have collected the data with permission and that they are sharing it with permission.


As an example of this in use, the images below show a Purchase Event being recorded in the Facebook ads platform.  The images show a high Event Match Quality and all of the information that is sent along with that Purchase Event. This makes purchase attribution far more accurate and enables Facebook to better optimize ads for the Purchase Event.

Image1: Facebook Purchase Event.  Purchase events typically have good Event Match Quality because Personally Identifiable Information is usually included with the event, such as email and first name.  Often, however, there is not enough data sent back to Facebook and reporting accuracy is poor, which means ad optimization is poor. With our Digital Fingerprinting (DPF) and First Party Data (FPD) solution, we improve Event Match Quality by sending enriched data to Facebook in a compliant format (SHA-256).

Image 2: Facebook Purchase Event Deduplication.  When sending enriched data to Facebook through a server-side API, it is essential to deduplicate that data from events recorded through the standard Facebook browser pixel. If this is not setup, then Purchase Events will be recorded twice, inflating results. In this image you can see that all the information was received with a 100% deduplication success rate.

2. Improve audience and event match rate quality for retargeting and lookalike audiences


The combination of digital fingerprinting and third-party data can be used to resolve the identity of site visitors in a process called Data Enrichment. Enriched data can be used to improve the data-feedback loop and can also be used in compliant direct outreach.


By improving the data-feedback loop, both audience and event match rate quality improve dramatically. This means that retargeting and lookalike audiences improve in size and quality, and we typically see a reduction in the cost per action when using data-enriched audiences.


As the images below demonstrate, our data solution improves Event Match Quality by significantly increasing the percentage of events that included anonymised (hashed) data that helps Facebook match an event to a Facebook user.

Image 3: Facebook PageView Events.  PageView events typically have lower Event Match Quality because there is usually not enough data sent back to Facebook with the event for Facebook to match the event to a person. With our Digital Fingerprinting (DPF) and First Party Data (FPD) solution, we are able to send this enriched data to Facebook in a compliant format (SHA-256) and Event Match Quality improves.

Image 4: Facebook PageView Events.  This detailed image shows that we are able to send key information to Facebook with the PageView Event so that Facebook can match the event to a person. This image shows we are able to send SHA-256 (Hashed) Email, IP Address, user Agent, External ID, Browser ID (fbp), State, Country, City, Last Name, First Name, ZIP Code and Click ID (fbc)

Image 5: Facebook PageView Events.  This image shows around 12.7 PageView Events tracked with 95% receiving enriched data through the Advanced Matching setup. With clients we take on, we typically see Advanced Matching at around 5% - 20% before implementing our data solution.

3. Improved top of funnel ad performance through 3rd party data


An audience built with third party data is collected and provided by a separate business. These audiences have been curated and verified in specific categories and can be used as a top of funnel audience for advertising and marketing. While there is no guarantee that these audiences will perform better than standard interest or keyword targeting, there is every chance that they will perform better, and at the least it is important to test.


If a 3rd party audience performs better at the top of funnel than a standard audience, then this is a success. If the audience does not perform better, then it gets written up as a worthwhile test.

4. Improved top of funnel ad performance through creative and ad copy


There is little point having great tracking and data enrichment if the ads themselves are suboptimal. As is always the case with marketing, it needs to be the right message to the right people at the right time.


If the message isn't right, then no amount of audience match rate optimization or data enrichment will help.


Ad creative and ad copy needs to be written and developed according to the highest principles of behavioural psychology and neuroscience. By combining superior ads with superior data solution, the advertiser has a compelling and testable competitive advantage.

5. Improved middle of funnel ad performance through strategic creative and ad copy


Every experienced advertiser knows the benefits of retargeting people who are in the middle of the funnel. However, many people execute this step poorly.


Recent research in over 30,000 online shoppers across a broad range of products and services has shown how important the middle of the funnel is for profitability and life time value.


However, middle of funnel retargeting is more than simply reminding a user to purchase or consume content. In order to attract shoppers to a new brand, there are at least 6 psychological triggers that must be included in advertising messages. When these 6 psychological triggers are included, the buying decision is shifted by up to 75%.


If these 6 psychological triggers are also combined with improved retargeting audiences through an enriched data-feedback loop, then the advertiser has a distinct advantage.

6. Diversify traffic sources and include alternative advertising platforms


Many advertisers start advertising on one platform, and then allocate the majority of their advertising budget to that platform. This is partly due to the learning curve needed when shifting to an alternative ads platform.


An example is users of Facebook that are thinking about shifting some of their ads budget to TikTok, which has a similar advertising interface, but is a completely different social advertising experience.

7. Direct email outreach through the compliant resolution of site visitors identity (Identity Resolution)


'Identity Resolution' is a relatively new technology that enables website owners to decode the name and email of site visitors for use in marketing personalization and direct outreach, such as with email marketing. 


Whereas most website visitors are lost to the business, Identity Resolution enables website owners to reach out to their site visitors with an offer to help or assist them.


Identity resolution can be a highly valuable source of leads, and typical site resolution rates are expected to be in the range of 40-60%. This means that for every 1000 people who visit a website, 400-600 of them will be identified and added to a complaint list of leads for email followup.


If 5% of 400 site visitors were to convert into a sale, that would be an additional 20 sales. With a typical ecommerce conversion rate of 2%, this could potentially double the number of sales.

Summary


Advertising is a crucial and key activity for most businesses, and yet new privacy policies and technologies have significantly reduced ad performance and increased ad cost across the major platforms.


The solution to this critical problem is multi-faceted and includes (1) improved ad copy and creative, (2) improving the data-feedback loop to improve ad performance and audience optimization, as well as (3) adding other sources of revenue through alternative ad platforms and email outreach to site visitors through identity resolution.


Each of these solutions requires a premium investment with a data-science partner, and suits businesses investing over $100,000 a month in advertising.


To talk with us about your unique situation, you can fill out an online application form and schedule an appointment here.