Einstein Attribution is a Pardot Einstein feature that aims to solve the gaps left in campaign attribution as a result of contacts not being associated with Salesforce opportunities and campaigns properly. Without the correct linkages between contact, opportunity, and campaign, marketers will never really know how much pipeline their campaigns contributed to.
With the tagline “leaves rules-based influence models in the rearview mirror”, Einstein Attribution introduces a new concept that goes beyond traditional Salesforce Campaign Influence.
You may remember I recently published an overview of Pardot Einstein, the set of AI-powered features for Pardot marketing automation, namely Einstein Behaviour Scoring, Einstein Lead Scoring, Einstein Campaign Insights, and the focus of this guide, Einstein Attribution. The overriding aim of these technologies is to find patterns buried deep in your engagement data, recognise themes, and take action, such as making recommendations to users.
When I first heard about Einstein Attribution a few months ago, my mind was blown. My research for this guide and the previous was heavily supported by Alon Shvo, Product Manager at Salesforce Pardot, who is working closely with the Einstein Attribution and Einstein Campaign Insights. This product is Alon’s baby – I could tell how proud he was to show me, after all, the product is achieving something pretty remarkable. He shed light on how these work under the hood – which I will now share in this deeper dive.
Overview of Einstein Attribution
| One-line Summary | Current Rules-based method* | Available? | Licenses Required? | Feature Prerequisites | Data Requirements? |
|---|---|---|---|---|---|
| “Leaves rules-based influence models in the rearview mirror” | Campaign Influence | Summer ‘20 release | Pardot Advanced Edition Salesforce Enterprise Edition (and above) | Pardot Connected Campaigns | 50 opportunity contact roles to get started! |
See the full table on Complete Overview of Pardot Einstein Features.
How Does Einstein Attribution Work?
If you are familiar with Campaign Influence, then I am sure you are also aware of opportunity contact roles. Opportunity contact roles are a key component for marketers, forming the bridge between opportunities and campaigns. Campaign Influence uses contacts and opportunity contact roles to determine which campaigns influenced an opportunity – unfortunately, opportunity contact roles are a pain to keep consistent.
Without the correct linkages between contact, opportunity, and campaign, marketers will never really know how much pipeline their campaigns contributed to.
Einstein Attribution aims to solve the gaps left in campaign attribution. The greatest takeaway about Pardot Einstein Attribution is that there’s no dependency on opportunity contact roles.
‘Virtual’ opportunity contact role records are created to patch up the gaps. These are not actual contact roles, but are created only to be leveraged in the back end.
More About Virtual Opportunity Contact Roles
As I said, ‘Virtual’ opportunity contact role records are created (not actual contact roles) but are only leveraged in the back end.
The data model crunching all the relevant data (leads, contacts, opportunities, activities etc.) that can help determine which individuals have a role in an opportunity. As a result, this creates an attribution graph using data extraction (eg. domain matching algorithms). The result is ‘Virtual’ opportunity contact roles that plug the gaps where real opportunity contact roles haven’t been created (but should be there). The team say that virtual opportunity contact roles have up to a 10x greater attribution coverage – think about how much opportunity credit that is discovering! The image below shows how this works.
What About Leads?
Most marketers will know that leads in Salesforce don’t get much of a look in on opportunity campaign influence, just the way that the data model is structured.
Besides Contacts, Einstein Attribution is able to leverage leads to accounts matching algorithms to associate prospects and leads directly with Opportunities that they are involved in – even if your sales team didn’t convert that lead, yet.
Campaign Influence Timelines
One thing that came to my mind was how timelines work with Einstein Attribution. With these data models doing magical things in the backend, how can we check that the time between a marketing engagement happening warrants ‘influence’ on a deal?
You can still configure the campaign influence timeframe (up to a max of 2 years). Plus, if an opportunity was created before the marketing touchpoint, then this marketing touchpoint won’t be included in the model. There’s no risk of cross-over in time periods.
This one element of rules-based influence models that gave marketers control, so you will be glad to hear you will still maintain visibility!
Data-driven Campaign Influence Model
‘Rules-based influence models’ are the first-touch/last-touch/even-touch Campaign Influence models you will be familiar with. These models rely on data relationships (opportunity contact roles) and timing (Campaign History at the time of lead conversion etc.), which makes this way of measurement less accurate overall.
What does Einstein Attribution aim to do? Essentially, Einstein Attribution will take you one step closer to the holy grail of accurate marketing attribution, by picking up gaps in attribution using a AI data-driven model that “attributes revenue share based on your actual customers, their engagement, and your successes”.
It’s an additional Campaign Influence model called ‘Data-Driven Model’ (and will appear in the Campaign Influence models list in Salesforce setup, alongside your other standard ones).
Einstein Attribution uses the Campaign Influence architecture, so the model’s output is actually campaign influence records.
The Data-driven model’ is a new influence model that will appear in the list in Setup beside first-touch, last-touch etc.
Multi-Touch Attribution Dashboard
The star of the show is exploring the Data-Driven Model in the B2B Marketing Analytics Multi-Touch Attribution dashboard:
As you may know, the Multi-Touch Attribution dashboard is one of the dashboards that comes out-of-the-box with B2B Marketing Analytics. You don’t have any additional setup effort here! Just as you would use the pre-made filter to switch between a first-touch and last-touch view, you do the same to show the Data-Driven Model:
Data Prerequisites
Usually AI-powered features require large amounts of data for the AI to ‘learn’ from and refine the data model/s. You will see this is the case with several of Einstein’s other products.
You will be pleased to hear that Einstein Attribution has minimal data prerequisites – by any org’s standards! You need at least 50 opportunity contact roles to ensure optimal model performance when you first start out.
My Thoughts on the Benefits
I think that the results that Einstein Attribution could potentially surface will be fascinating for marketers. The marketing teams that will gain the greatest value from Einstein Attribution will be those that find additional opportunity credit that should be attributed to marketing campaigns – the ‘lost’ credit that a lack of Contact Role discipline causes.
On the other hand, even if Einstein Attribution doesn’t surface a shocking amount of lost credit, it proves that your organisation’s Opportunity Contact Roles processes are sound. You can silence those nagging doubts over data accuracy.
The Multi-Touch Attribution Dashboard in B2B Marketing Analytics brings this all to life. You will be able to visually see how much additional opportunity credit Einstein Attribution has been able to pick up by switching the compare the dashboard filter to display different models (Einstein vs. rules-based).
Finally, Alon wants you to think a step deeper:
“Einstein Attribution will result in smarter and more accurate channel level attribution, since the influence weightings are based on actual conversion data, as opposed to rule-based models which are essentially a set of arbitrary pre-defined rules”
Does this go over your head? It might help to read up on how Einstein Behavior Score works, using conversion patterns to determine a score, as opposed to adding up individual marketing touchpoints over time.
Want More Information?
I recommend you watch this webinar recording to learn more.
The Original Algorithm
I will leave you with this, the original algorithm for Einstein Attribution. Don’t worry, I don’t follow it either – let’s be thankful for the Einstein team that made it a reality!
My sincere thanks goes to Alon Shvo, Product Manager at Salesforce Pardot, who contributed to this article, and was patient answering my questions as I wrapped my head around how the features worked!