When we talk about the consumer journey (also known as the customer journey), we have long followed a linear model, but the reality is that this journey takes place over several sessions, sites, devices, media channels and physical stores.
Even relatively low-cost products are investigated, discovered, compared and purchased at multiple points of contact. There are literally dozens of channels that can influence the sale.
Decision-making is a cognitive process that involves both the reason and emotion of an individual. In addition, it is a process that requires time, takes previous experiences into account and is based on several analyses.
Source: Measurement and Attribution Plans (IAB Brazil)
Looking at the scenario above, is it possible to say with confidence which channel/device we should attribute to the purchase? Probably not, right?
In the digital world it is common to give all credit to the last contact made by the user, which is called last-click attribution, in accordance with the scheme below:
From this we can see that different models of media attribution are necessary, as we need to determine the role that each channel plays in the consumer’s journey. All the models show that vehicles do not work independently. All interactions are unique to a situation, to a product and to a brand. Each channel has a specific function in achieving the goal.
What are the pillars for ensuring good attribution?
Below are the main pillars for the attribution of media within companies.
Data
Users leave digital trails (via cookies) and these traces are essential for identifying the user and mapping all of their contact points before the conversion.
However, even though Google Analytics knows how to identify the same user who visited a site several times, it will not know where that user came from if the campaigns are not parameterized.
The parameterization of campaigns is the inclusion in URLs of parameters that will help you identify the source of visits to your site.
The entire process needs to be aligned for an attribution model project to work. Consideration must be given at the time of the briefing, in the media plan, AdInfo, Adserver, reporting, etc. to ensure that all data is collected correctly.
Models
A lot of thought must be given when choosing which model of attribution to use.
In reality, there is no perfect model. Every time that you choose a new model you must run a series of tests, applications, measurements, changes and more tests, until you find the one that is most suitable for your goals.
Within the same company it is possible to use more than one model. The choice may change according to the business, media strategy, campaign or the consumer response to your communication.
Having said that, we can talk about the maturity of the models:
- Assisted Conversions
Assisted conversion shows the contributions of each channel to the final sale. This model allows us to identify the channels that contribute most to the conversion, without being the actual converters. It also allows us to identify the converting channels.
This model uses the Index of Assistance (assisted conversions divided by last-click conversions or direct conversions), which calculates the ratio that each channel generated assists against its direct conversions. It also allows us to see how far the channel appears to be from the conversion when we compare the channels.
Source: Think with Google
2. Rule-Based Attribution
Rule-Based models are predefined attribution types.
a) Last Click: The last channel before conversion gets the full conversion credit. It is the default model of leading media buying tools and Digital Analytics.
b) First Click: The first channel of the journey receives full conversion credit. This template is appropriate if the goal of the campaign is to create awareness.
c) Linear: The conversion credit is equally distributed among all channels, that is, if a user went through four channels to reach the conversion, each will receive 25% of the conversion credit.
d) Time Decay: Credits are assigned progressively to the channels that are closest to the conversion according to the distance in days from the interaction.
e) Position-based: The first and last points of contact receive greater weight by conversion (40% each). The remainder is distributed among the other channels/devices.
3. Algorithmic/Data-Driven Attribution
In recent years, there have been advances in more efficient algorithm-based attribution models. By applying machine learning, algorithms and statistics we can infer the individual contributions of each channel based on data.
Scenario 1:
Scenario 2:
The use of advanced statistical models and inferences allows for a continuous and constant process of customization of the result-based model. The model is statistically adjusted to optimize the journey according to the similarity of conversions that have already occurred.
Measurement
After the data is collected and organized and the models are applied, an important part of the process of media attribution is the measurement of results. Analysis outputs are not only the most accurate results, but lead to more refined strategies.
We must have a good definition of the KPIs and support metrics of our business. The model may change, but media goals must remain the same.
Now it’s time to put it into practice!
It is essential to support your assumptions in the attributed data and “risk” your optimizations. Bring the culture of attribution models into your company and the campaign development steps closer together. Bring data-driven culture to your own marketing decisions.
Remember: the model used and the strategic possibilities should be reviewed periodically, and you should always test!
Profile of the author: Angelica Oliveira | A Marketing graduate of University of São Paulo she is currently an analyst at DP6, a marketing intelligence and performance consulting company. She has 2 years of experience in the digital market, working with Digital Analytics, Social Intelligence and on SEO and DataViz projects.
