Using Attribution Theory to Improve Attribution Models

Using Attribution Theory to Improve Attribution Models

There has been a lot of talk about Attribution models for advertising and marketing efforts, but there’s a lot more to the picture than another metric to measure. There is no doubt that an accurate Attribution model is the best way to understand and improve ROI, targeting your marketing message; however, most of Google Attribution models demonstrate issues for B2C businesses and have even more issues for B2B companies that have a complex sales cycle. Typically, for business models with this type of sales cycle, there isn’t an “Add to Cart” button. There needs to be an in-depth process to track how the person contacted and what led to the sale (conversion).

If you want to learn more on how to use Multi-Channel Attribution Modeling, I would suggest reading Avinash Kaushik’s Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models. The post is several years old and some of the screen shots for Google Analytics are dated, but the logic and concepts are still valid.

Ok, this is great but what is Attribution anyway? It’s most certainly so much more than the ‘first, last, linear, position or data’ label. It is complex, yet simple. It’s actually a psychological theory – that’s right – it’s core ideation is not about marketing or advertising at all. Nor is it new; it’s been around for some time. One of the best references I found is from and that was published in 2012. Basically, Attribution Theory is concerned with how and why ordinary people explain events as they do.

  1. Internal Attribution is dispositional; meaning a person has an internal trait or reason for acting in a particular way. When you are creating an ad, what does your demographic look like? Should your message or visuals reflect a belief of your ideal customer? Are they buying because of a motive or a personality trait?
  2. External Attribution is situational; meaning assigning the cause to some situation or event outside the person’s control. Has there been an event like a gas shortage that forced people to purchase more fuel efficient cars? Your personality may be thinking big SUV, but fuel costs may be an outside force, which is why you drive a Prius. During the bank crisis, many people were leaving big banks or banks in general and opting for Credit Unions. Their reasoning was that they were ‘safer’ and more trust worthy.
  3. Covariation Model focuses on conditions that lead a person to attribute a cause to an environmental situation causing their action.
    • Distinctiveness of the entity – the behavior only occurs when the entity is present (HIGH DISTINCTIVENESS). Fred is laughing at the movie, does he laugh at all movies (LOW DISTINCTIVENESS) or only Mike Myers’ movies? The more specific the behavior is to this one entity, the less it tells us about Fred.
    • Consensus – do most others respond similarly? If most others respond the same way to this entity (laugh), then there is HIGH CONSENSUS, if most others do not respond this way to this entity there is LOW CONSENSUS
    • Consistency – does the person act this way in the presence of this entity most of the time (HIGH CONSISTENCY) or only some of the time (LOW CONSISTENCY)High Distinctiveness, High Consensus, and High Consistency leads to an External attribution; Low distinctiveness, Low Consensus, Low Consistency leads to an internal attribution
    Information Combination Predicted Attribution Example
    High Consistency
    Low Consensus
    Low Distinctiveness
    Personal Cause Situation: Mike Thomas expresses dissatisfaction every time he is given a price and delivery date. Other purchasers are not dissatisfied when given a price and delivery date. Mike expresses dissatisfaction on just about price or delivery date he is given.

    Perceived cause: Most likely cause of Mike’s dissatisfaction is something to do with his personal characteristics (e.g., he is a chronic complainer)

    High Consistency
    High Consensus
    High Distinctiveness
    Environmental Cause Situation: Mike Thomas expresses dissatisfaction every time he is given a cost of a product. Other purchasers are also dissatisfied when when given a price. Mike doesn’t express dissatisfaction other product prices or delivery dates he is given.

    Perceived cause: Most likely cause of Mikes’s dissatisfaction is something to do with the task itself (e.g., The product doesn’t have the value compared to the price).

    Low Consistency
    Low Consensus
    High Distinctiveness
    Circumstantial Cause Situation: Mike Thomas expresses dissatisfaction every time he is given a price and delivery date. Other purchasers are not dissatisfied when given a price and delivery date. Mike has not expressed dissatisfaction about price or delivery date he is given in the past

    Perceived cause: Most likely cause of Mike’s dissatisfaction is something related to the momentary circumstances surrounding his price and delivery date (e.g., Your product may not be important right now, his supervisor may want him to cut cost, having a busy and bad day, etc.)

  4. Plan your model don’t just look at simple reporting on your ads. You can get create an attribution model that will help you make informed business decisions. Consider this, is your product or service being purchased for internal or external reasons? That may determine how you position your advertising. Although this is counterintuitive, don’t always think in terms of ROI, rather invest in the research. Develop a hypothesis of what your customer is thinking, create some tests and use the data for future campaigns that ROI is the main concern. Look for a deeper meaning behind your attribution reportings.Always keep in mind the “why” is more important than the “what”. If you create four ads and your Attribution model tells you one ad appears to have more effect you need to understand why so you can further develop and improve future campaigns.
  5. For B2B or a business with a complex buying cycle, it can be next to impossible to attribute any of the touch points that lead to a sale. With a complex sale it might not be any one ad or feature or benefit of the product or service. It could be the customer – Maybe he really liked the sales person and trusted him. Additionally, you might not ever have enough data to connect all the dots during the sales cycle. Even when you get a lead from a White Paper on a landing page that was sent via an email drip is not really relevant. Why? You can never really determine why or how the sale was started or completed.

How can you make an Attribution model work for you? Look at your current customers, then profile them. Try not to create a persona, but a real profile. In a spreadsheet, create columns with quantifiable metrics; consistency will be important. Then document those characteristics for each customer. Step back and see if you can create patterns or cycles that they go through. If possible, you could interview them to get more information. From there, you can determine how they fit. Are they internally or externally driven? What is that driving force?

Attribution is a powerful tool to the digital marketer but there is much more then calculating ROA. You can start to improve your ROA and reach new customers that you may have been missing. Developing and understanding an Attribution model that works requires a true commitment from the organization. To learn more download the eMarketer “Marketing Attribution 2017 Five Best Practices” report.

Exigo Digital Marketing will work with you to create an attribution model and testing platform to improve your ROA. Contact us, to learn how we can help.

Kelley’s Attribution Theory

Get eMarketer’s
Marketing Attribution 2017 Five Best Practices Report
Sent Directly to Your Inbox

By | 2018-11-19T06:43:50-04:00 November 7th, 2017|Attribution, Digital Marketing|