LinkedIn Impression Attribution

Updated by Mikkel Settnes

Improve touch-based attribution by including effects from LinkedIn ad impressions on account level.

Standard touch-based attribution models cannot handle both ad clicks and ad impressions at the same time without resulting in misleading conclusions, due to the quantity nature of the different types of the data.

This leaves a big gap in the understanding of campaign effectiveness for impression based ads.

Dreamdata’s impression attribution bridges this gap by providing a specialized algorithm to add the effect of LinkedIn ad impressions, to touch based multi touch attribution models, blending the effect of clicks and impressions.

Giving you a more holistic view of attribution.

How it works:

  • Estimates the impact of impressions: utilizing Dreamdata's native Partner integration with LinkedIn, Dreamdata analyzes account-level impressions on LinkedIn, highlighting the effect of your LinkedIn ads on your funnel.
  • Combined with touch-based models: it seamlessly integrates with existing touch-based attribution models available on the Dreamdata platform, providing a unified and more accurate view of all contributing factors.
  • AI-powered for simplicity: combining the effect of LinkedIn impression data with Dreamdata’s AI driven attribution models leverages AI to automatically learn and adapt, giving you valuable insights without the hassle.
Example:

LinkedIn provides ad impressions on an account level, allowing you to identify when a particular company was exposed to a particular ad campaign. 

Both ad impressions and ad video views count as impressions in this context.

Dreamdata automatically adds the events linkedin_ad_impression and linkedin_ad_video_view to your data when you enable the LinkedIn integration

Enabling impression attribution on any multi touch attribution model will assign attribution credit to the account based impressions in the journey increasing the attribution to the associated campaigns.

The attribution assigned to impressions in a journey will depend on several factors like:

  • The type of campaign - different campaigns have different effect.
  • The amount of exposure present for each campaign in a journey - very few impressions will not get credit.
  • The recency of the exposure compared to the chosen stage - this includes time lag effects.

Read more below.

Benefits:

  • Get a clearer picture of ad effectiveness: Understand how both clicks and impression exposure contribute to reaching funnel stages, leading to more effective campaign optimization.
  • Make data-driven decisions: utilizing LinkedIn's account-based impression data, offers unmatched insights into the performance of LinkedIn ads allowing you to allocate marketing resources more efficiently. 

Ready to get started?

To enable impression attribution within your Dreamdata platform

  1. Go to Data Platform -> Settings -> Attribution Models
  2. Select your attribution model (available for all multi-touch attribution models)
  3. Set how important impression based media is for your business. 

The last step will define the basic assumption of the model with respect to how much (total) effect can be credited to ad impressions, giving a high-level control of the model’s behavior.

The credit will still be determined by type of impression, recency etc. but up to a maximum defined by the importance level

  1. None: 0%
  2. Low: 0-2.5%
  3. Moderate: 0-5%
  4. Important: 0-10% 

If impression attribution is disabled (setting 'None' ), the impressions will still be visible in the customer journey and count towards influenced metrics, but will not be eligible for attribution.

Different attribution models can have different setups

Only multi touch models can enable impression attribution

Details

When measuring the effect of ads, there is a fundamental difference between something a person did (a touch) and something a person was exposed to (impressions). 

Exposure like impressions or video views should not be treated similar to clicks in a standard touch-based attribution model. Treating impressions and clicks similarly will result in incorrect conclusions. This is because the amount of impressions compared to the number of clicks will skew these models.

Dreamdata’s impression attribution is designed to bridge this gap for LinkedIn ad exposure. It estimates the effect of LinkedIn account level ad impressions and video views, and combines it with the standard touch-based attribution model.

The click of an ad will be measured using a touch-based model and the effect of the impressions are measured using the specially designed exposure attribution(see below for How does the impression attribution algorithm work?). 

The total attribution is the effect of clicks and exposure (like impression and video views) combined.

What are LinkedIn Account based impressions?

Dreamdata’s native partner integration with LinkedIn will add account based impressions and video views to the customer journey. This automatically adds the events linkedin_ad_impression and linkedin_ad_video_view.

Account based impressions and video views are ad exposure to a specific business account (= domain), but where the individual person shown the ad is not known. 

The impressions are tied to a specific campaign group and campaign. 

* for privacy reasons LinkedIn only report account based impressions if more than 3 are served over a given period.

 

When should I use impression attribution ?

Enable impression attribution on an attribution model to include the effect of LinkedIn ad impressions and video views to the attribution towards any supported stage goal. 

For example, if you are using LinkedIn ads without a direct measurable call to action like a form submit, adding impression attribution will provide you with a more accurate attribution picture.

 

How does impression attribution work?

Impression attribution utilizes the account based impressions within each customer journey to estimate the effect of impressions on reaching a stage (see limitations below for which stage models are supported)

For each campaign the effect per impression is estimated including saturation and carry over effect. This effect is time dependent - meaning it can change month-to-month. 

In this way, impression attribution gives more weight to impressions from high performing campaigns, compared to the same number of impressions for low performing campaigns.

The effect per impressions is first scaled to include saturation effects (see Saturation details below). In this way it incorporates the natural diminishing return of exposure based media.

Afterwards, we add carry-over effects describing the time lag effects for ad exposure (see Carry Over details below).

Generally, journeys with a lot of impressions from campaigns with a high effect per impression will give more credit to the impressions and their associated campaign. Journeys with only a small amount of impressions or where all impressions are from campaigns with low effect per impression, will give less weight to ad impressions. Naturally handling that not all types of impressions are equal.

The total attribution of a model with enabled impression attribution is
  1. Credit to the account-based impressions using set importance level
  2. The remaining credit is assigned by the chosen touch-based model

This ensures that attribution to account based ad impressions blends seamlessly with each multi touch attribution model.  

Example
A W-Shaped attribution model has enabled impression attribution.

So the algorithm estimates the effect of impressions on each journey.

The remaining credit will be attributed according to the W-Shaped model.

The model automatically handles both carry over and saturation effects when determining the effectiveness of ad exposure for a particular campaign. 

Below follows a more detailed explanation of these effects.

 

Saturation Effects

Impression attribution will automatically incorporate saturation effects independently for each campaign. 

This ensures that campaigns which naturally lead to many impressions will have a different saturation compared to campaigns with fewer impressions.

Saturation effects based on a Hill transformation function

The Dreamdata impression attribution incorporates saturation effects using the Hill function for each campaign [1] with parameters dynamically determined based on the data input. 

This approach ensures that the model dynamically adapts saturation levels as it learns more about the campaigns you are running.

Saturation effects are important to get the correct impact of exposure based media, as effects naturally depend on the amount of impressions. 

  • A certain number of impressions are needed to obtain an impact. This ensures that a very small amount of impressions only get a low incremental effect. 
  • As an account is exposed to more and more impressions the effect saturates ie. increasing the number of impressions will only have a small incremental effect.

Time lag effects (Carry Over) 

The impact of an impression is expected to have a lasting effect ie. not only influence things happening on the same day.

This lasting effect is usually referred to as carry over effects and ensures that impressions can affect outcomes happening within a certain period after the ad exposure. [1]

This is essential for getting a trustworthy estimate of the effect, especially when the customer journeys are long. 

The model will automatically start to decrease the effect of impressions happening more than 2 months before reaching a stage, thereby giving more weight to recent exposure, while still allowing compounding effect of ad exposure. The total attribution window of impressions are 180 days.

The time lag effects are applied independently to each stage model ie. an impression can be recent for the MQL stage but have decreased effect for the Closed Won stage.

Limitations of impression attribution

The impression attribution algorithm is made to operate on account based impressions, like the ones supplied by LinkedIn. Therefore, it only works for attribution towards account based stage models. 

The impressions are purely account based so attribution cannot be made for journeys that only consist of touchpoints of specific contacts. 

Exposure attribution therefore does not add attribution to stage goals that are:

  • Contact based: the journey only includes touchpoint of a known contact
  • Opportunity role based: the journey only includes touchpoints on contacts added to the opportunity or deal

Read more about the journey type for stage goals and how to set it up here

[1]  https://research.google/pubs/bayesian-methods-for-media-mix-modeling-with-carryover-and-shape-effects/


How did we do?