Multi-Touch Attribution Models: Get it Right

Introduction

Do you measure the impact of each marketing channel on your sales? Have you tried any of the multi-touch attribution models?

I used to think that all my traffic sources were equal. Then I discovered multi-touch attribution models and learned how to measure the impact of each traffic source.

Now I know which traffic sources are worth investing in my business. Do you?

There are many different attribution models, but multi-touch attribution is among the most popular. 

Multi-touch attribution lets you see how each touchpoint in the customer journey contributes to the conversion. It helps you understand which marketing channels are working and which aren’t. It also enables you to understand how people interact with your brand before they make a purchase. 

Multi-touch attribution can be beneficial for both small businesses and large enterprises. That is the beauty of it — It’s scalable.

If you are focused on knowing how your customers convert and how to improve your conversion, then please read further. 

Before we delve into the multi-touch attribution models, let me briefly describe the terminologies we will use in this article.

Terminologies

Conversion 

Conversion occurs when someone performs the final committing action that you desire, like subscribing to your service or buying your product.

Touchpoints

Touchpoints are a series of interactions or points of influence between the customer and your product/service that finally lead to the conversion. They can be ad clicks, keyword searches, visits, impressions, messaging, and sequences leading up to a  conversion. 

Channels (marketing channels)

In marketing, companies use channels to reach their target market. There are three main channels: direct, indirect, and digital. This article will refer to digital channels like paid search, email, social media, YouTube videos, etc.

Customer journey

A customer’s journey in marketing is the process a customer goes through when considering a purchase. The journey usually starts with an awareness of a need or want and the different options to find the best product or service to meet that need or desire. This journey goes further to purchasing a product or service and recommending the service or product to others.

Difference between multi-touch and multi-channel attribution

Multi-touch and multi-channel attribution are used interchangeably. 

However, there are notable differences:

Multi-channel attribution weighs attribution credit by channel. Examples: social, paid, organic, etc. It doesn’t include or factor in specific touchpoints like messaging of copy or sequences. 
Multi-touch attribution is more granular. It focuses on specific ads, including the messaging, the channels, and the sequence of interactions.

The different types of attribution models

Let us have a quick rundown of the different types of attribution models, as they are essential to understanding the multi-touch attribution model better.

There are two (2) types of attribution models:

1. Single-touch attribution models, and
2. Multi-touch attribution models

We will be using Google’s example of attribution modeling to demonstrate single-touch and multi-touch attribution models:

A customer finds your site by clicking one of your Google Ads ads. She returns one week later by clicking over from a social network. That same day, she comes back a third time via one of your email campaigns, and a few hours later, she returns directly and makes a purchase.

Single-touch attribution models

A single-touch attribution model is a model that attributes all of the credit for a conversion to a single touchpoint. 

Below are the different single-touch attribution models:

First-touch attribution model

In the first-touch attribution model, a credit for a sale or conversion is given to the first marketing touchpoint with which a customer interacts. It is also referred to as the first-click or first-interaction attribution. 100% of the credit is assigned to a single source for each conversion. 

In the example above, the paid search channel receives 100% of the credit for the sale.

Last-touch attribution model 

The last-touch attribution model gives credit to the final marketing touchpoint that led to a conversion. The last thing a customer does before converting takes all credit for driving that conversion. 

In the example above, the Direct channel receives 100% of the credit for the sale.

Last-touch attribution vs. first-touch attribution

Last-touch attribution falls victim to the same pitfalls of first-touch attribution. The only difference is that it gives 100% credit to the last interaction.

Last non-direct touch attribution model

The last non-direct click attribution model ignores all direct traffic. 100% of the credit for the conversion (like sale, subscription, or form sign-up) goes to the last channel before the customer converts.

In the example above, the email channel gets the credit.

Last Google Ads click attribution model

In the example above, paid search channel takes 100% of the credit for the conversion or sale. It is the only Google Ads click in the example.

The attribution models above are rule-based single-touch attribution models, which differ from the algorithmic or data-driven ones we will deal with later in this article.

Multi-touch attribution models

As discussed in the introduction, we are going to venture deeper into multi-touch attribution models: what they are, their benefits, how to implement them, and some challenges that you may encounter.

What is a multi-touch attribution model? 

Multi-touch attribution models are used to understand how people interact with a product, service, or brand. They help businesses see which marketing channels in the customers’ journey are most effective at driving conversions.

The detailed view of the customer’s journey makes this model very relevant for marketers trying to assess the effectiveness of each marketing channel.

What types of multi-touch attribution models can you use?

There are many types of multi-touch attribution models we can use, and these types are classified into

• Rule-based multi-touch attribution, and
• Algorithmic or data-driven attribution model

To illustrate each model clearly, we will still use the same example we used in the single-touch attribution models above:

A customer finds your site by clicking one of your Google Ads ads. She returns one week later by clicking over from a social network. That same day, she comes back a third time via one of your email campaigns, and a few hours later, she returns directly and makes a purchase.

Rule-based multi-touch attribution models

There are three (3) rule-based multi-touch attribution models:

1. Linear attribution model
2. Time decay attribution model
3. Position-based attribution model

Linear attribution model

The linear attribution model helps you understand the influence of your entire marketing campaign across multiple channels. The downside, the credits are equally spread among all the touchpoints. This is rarely the case, though, for every interaction in driving conversion. 

This attribution model is often used when the buying cycle is more complex and has several channels along the conversion journey.

The linear attribution model is also known as the even weighting attribution model.

In the example above, each channel (paid search, social network, email, and direct) shares an equal credit of 25% for the sale.

Time decay attribution model

The Time Decay Attribution model is a way of attributing value to different parts of the customer journey. It assigns a higher value to the most recent touchpoints and a lower value to touchpoints further back in time. This is because the most recent touchpoints are typically more influential in leading to a conversion.

The example above shows that direct and email channels receive the most credit. Social network channels receive less credit, while paid search interaction receives the least because it occurred one week earlier.

Position-based attribution model

Under this attribution model, we have two types:

• U-shaped multi-touch attribution model
• W-shaped multi-touch attribution model

U-shaped multi-touch attribution model

Position-based attribution models assume the two most important interactions in the customer’s journey before the conversion: the very first and the very last. 40% credit is assigned to each interaction: first and last interaction. The middle interactions receive the remaining 20% credit evenly.

In this example, paid search and direct channels receive 40% credit each, while social network and email channels receive 10% credit each. 

W-shaped multi-touch attribution model

The W-shaped model places a higher weight on the initial contact, lead creation or conversion touch, and opportunity or deal generation touchpoints. These three touchpoints receive 30% of the credit, while the remaining touchpoints between the initial contact and deal generation or creation evenly receive the remaining 10%.

To illustrate this model clearly, we will still use the same example above with a slight modification.

A customer finds your site by clicking one of your Google Ads ads. She returns one week later by clicking over from a social network. That same day, she signs up on your newsletter form. The next day, she comes back via one of your email campaigns and books a call. A few hours later, she returns directly and makes a purchase.

The W-shaped attribution model’s advantage highlights the top three (3) interactions or transitions in the customer journey: initial contact, lead creation or conversion touch, and opportunity or deal generation.

Algorithmic or data-driven attribution model

Algorithmic attribution models are also called data-driven attribution models. It is called DDA in Google Analytics. 

The algorithmic attribution model uses machine learning to dynamically assign (fractional) credit across multiple transitions or interactions that lead to a conversion.

While algorithmic attribution models take the guesswork out in knowing what fits the needs of your particular business, they need to be more transparent than their rules-based counterparts.

Here are the advantages and disadvantages of using algorithmic or data-driven attribution models:

The advantages of algorithmic or data-driven attribution models

• DDA attribution models take the guesswork out.
• DDA attribution models provide the most reliable predictive value.
• DDA attribution models can apply to various business models and strategies. It does not matter if your business has a long sales funnel; the model will adjust to reflect this landscape. 

The disadvantages of algorithmic or data-driven attribution models

• We cannot thoroughly understand how machine learning works and assume that machine learning does a better job in computing than humans.
• This attribution model depends on data volume and quality. Machine learning works best with a high volume of interactions (transitions) and conversions. Google Analytics has a minimum requirement of at least 600 conversions per month. 
• This capability often comes with a price tag. In Google Analytics multi-channel funnel (MCF) reporting, data-driven modeling is only available with a 360 (paid) license. However, GA4 does offer free attribution modeling, excluding direct traffic. 

Rule-based versus algorithmic attribution models

Traditional rule-based attribution models have biases in allocating value to channels or touchpoints that make them ineffective. 

The more complex the customer’s journey is, the more accurate insights in the attribution models are needed to determine the channels’ value and resource allocation. 

A subtle difference in insights may help a business win the market competition.

The algorithmic or data-driven attribution model can detect touchpoint or channel changes quickly and adjust the weight values more accurately.

Full path multi-touch attribution model

A full path multi-touch attribution model considers all touchpoints in the customer’s journey or experience that lead to the point of purchase. It is more than just the conversion event, like a sign-up form.

A full path multi-touch attribution model gives 22.5% credit to the first touch, lead creation or conversion touch, opportunity or deal generation touch, and closed-won touch. All middle touchpoints receive the remaining 10% evenly.

This model has all the benefits of the linear and the w-shaped models. 

Custom multi-touch attribution model

As the term implies, you may design your multi-touch attribution model if you need a different model than the preceding models presented. 

Creating a custom multi-touch attribution model means you already fully grasp the various marketing channels and how your prospects or customers interact with the multiple touchpoints.

You should use one of the preceding attribution models presented above and apply customization after some time as the need arises instead of jumping straight to customization.

What are the benefits of multi-touch attribution models?

Let’s get straight to the point on the benefits:

• Allows for a comprehensive perspective of the customer journey on what drives conversion compared to the single-touch attribution models, which are usually the first or last touchpoint. 
• Determine the most influential touchpoints that convert. 
• Optimize the channels which are responsible for higher conversions.
• Adjust those that are underperforming.
• Allocate most of the budget to the touchpoints or channels that provide the highest return on investment (ROI).
• Offer flexibility and agility for marketing teams to quickly implement what is and isn’t working when things change or evolve, as the market usually does.

The key benefit of using a multi-touch attribution model in marketing is greater insights into all interactions that lead to a conversion. Regardless of your attribution model, the goal in marketing is to increase conversions, while the goal in business is to increase profitability.

Mult-touch attribution model versus marketing mix modeling 

Marketing mix modeling (MMM) and multi-touch attribution (MTA) models are two well-known approaches used to quantify the impact of marketing and advertising activities on modeled KPIs (such as sales, subscriptions, or other types of conversions).

Though we won’t deal much with marketing mix modeling in this article, let us compare key differences between MMM and MTA:

Length of modeled period

• MMM needs to measure about 2-3 years of weekly data
• MTA relies on a much shorter period of 1 week to a month of a customer’s digital journey.

Impact of advertising

• MMM quantifies the impact of advertising above the benchmark or baseline. (The base level of modeled sales or other KPI built up over time.) In this regard, MMM reasonably reflects what is effective only when they have the potential to raise the KPI level above its baseline. 
• MTA does not differentiate between the base and incremental conversions.

Evaluation of contributing factors

• MMM’s key strength lies in its holistic approach and the ability to quantify the contribution of all marketing as well as external factors.
• MTA often needs to consider offline and other external factors not captured in the customer’s digital journey.

Agility

•  MTA is quick and more frequent to implement as it does not require long data collection periods.
• For MMM this is not the case.

Businesses that market through both online and offline channels can benefit from implementing both approaches.

What to consider when choosing between a multi-touch attribution model or marketing mix modeling?

The following drives what approach to use:

• Presence: online (website) or offline (brick-and-mortar stores)?
• Budget
• Access to data: user level or an aggregated level?

Recommendation

Marketing decisions can benefit from the combination of both MMM and MTA. The recommendation is to start with the approach that brings immediate benefits for optimization. You may use the two approaches in tandem. 

While MMM provides broader insights and recommendations on achieving a more significant impact on the bottom line, MTA is the key to optimizing the customer’s digital journey effectively and efficiently. 

Consistent success based on MTA results will beget a more substantial above-the-line impact quantified within MMM.

Steps to implement multi-touch attribution models

Step 1: Know your goal

The goal is the result you desire to achieve. KPIs are not goals. 

To achieve a long-term goal, we must designate short-term goals geared toward the long-term goal. 

In my years of experience as a digital marketer, I have learned this from partners, colleagues, bosses, mentors, and even clients. 

Each goal comes with a set of KPIs and targets. Processes are breakdown to achieve these short-term goals with the direction of accomplishing that long-term goal.

This simplifies things, makes them more achievable in a short period, keeps the team effort on track, and creates a healthy, emotional, and psychological environment for the team.

Step 2: Establish your KPIs

From the very start, establish your key performance indicators (KPIs). Identify which metrics matter for your business to understand whether your efforts are supporting you in reaching your goal.

Selecting the right KPIs can determine the success or failure of the business. Only use KPIs that reflect on how to drive your efforts to your goal.

Step 3: Set a target

Your target is the level or benchmark you aim to achieve for your KPIs.  

Step 4: Choose an attribution model

Start with any attribution models you can work with to achieve your short-term goal. Modify and scale your attribution model as you move forward. This calls for simplicity and scalability when selecting your attribution model.

These are the basic steps to consider when choosing your first multi-touch attribution model:

• How you collect the data.
• How to combine and connect the data.
• How to visualize data.

How you collect the data

Collecting the data is the first step. Data is the oil that lubricates your marketing engine. 

You basically need to know the following:

• Who visits your site?
• How they got there? 
• How do they convert, or what is their behavior? 

There are three ways you can collect data, and they can work in tandem:

JavaScript  

You add a code to your web pages to understand who interacts with your site and how. The calls include the following:

• Page records when a customer views the page;
• Track records of what the customer does on the page;
• The segment Identify ties that behavior to other traits you know about them; and
• Inbound determines where the customer comes from. Ex. email or organic social.

UTM parameters 

These are snippets that populate at the end of URLs. They provide data about where the customer came from, including information like the following:

• Source: This is where your traffic came from. Ex. Google, Newsletter, Facebook, Twitter, Yahoo, Bing. 
• Medium: This is the marketing medium. Ex. CPC, banner, email.
• Campaign: This can be the slogan, promotion, or product/service. Ex. thanksgiving promo.

Here is a sample UTM parameter: https://www.twibiagency.com/?utm_source=newsletter&utm_medium=email&utm_campaign=thanksgiving_promo

The above UTM parameter shows that the traffic came from a newsletter, using email as a communication medium, and the promotion is about Thanksgiving.

I used to include the UTM parameters for term and content. However, since GA4 does not currently report utm_term and utm_content in Google Analytics 4 properties, I make sure to use the utm_campaign parameter as descriptive yet as short as possible. More on GA4 parameters here.

If you are not familiar with UTM parameters, there are many tools you can use. Google Analytics Campaign URL Builder is one tool used to create UTM parameters. 

Note that you can integrate the URL parameters with the JavaScript calls to get a clearer, more complete picture of your users.

APIs

It is another way of integrating with your CRM, advertising platform, and other proprietary ways of identifying your customers.

How to combine and connect the data

The next phase is to connect your data in one place to make sense of it. This is tricky for small-medium businesses as most platforms or solutions lean toward enterprise-level businesses. 

There are more affordable platforms that can combine and connect most of your data into one place. Still, they need help to combine and connect all data points and the different channels.

If you know of an affordable solution, most of our readers would appreciate knowing. 

How to visualize them

A graph or chart can present numerous insights. To achieve such a feat, you need to be able to query your selected data into a visual report. You can use platforms like Google Analytics, Google Search Console, and Google Data Studio. Bing also has its own set of platforms. There are many others. 

Step 5: Build and nurture your team

Gather your core team members. They are vital in collecting, analyzing, and optimizing various data and touchpoints.

Step 6: Utilize marketing analytics software

Use analytics software that can support multi-touch attribution modeling even if you start with a single-touch attribution model. Again, scalability is an essential element in marketing. You can try GA4 to begin with; it is limited, but it may serve its purpose for a small-medium business.

Once you have collected your data, you can begin drilling for crucial insights. 

Here are some of the questions you may want to consider:

• How efficiently is a marketing channel moving prospects closer to conversion (achieving maximum productivity with minimum wasted effort or expense)?
• How effective is that marketing channel in converting? How successful is it in converting prospects to a sale?
• What are the poorly performing channels? Are they simply underperforming? If yes, why? What is blocking conversion? A pop-form? The payment method? The messaging? 
• Which channels collaborate to influence conversions?
• What is the average sales cycle?
• How to remove friction to convert prospects to customers easily?
• Are there abrupt changes in the touchpoints?

Step 7: Continue to test and rinse (optimize)

Consistency is another key to success. Evaluate regularly your multi-touch attribution data to refine and test your strategies. You and your team can establish optimal techniques and marketing sequences to reach your consumers at the right time.

Common challenges of multi-touch attribution models

While multi-touch attribution models offer more insights, they have flaws. 

Setup and tracking

A multi-touch attribution model is more challenging to set up and track. Data integration from various marketing channels and platforms is a must. Data integration takes work. 

For reliable Data Modeling, there are procedures to be followed: data cleansing, combining, standardizing, and merging (de-duplicating).

Attribution weightings

Determining how much weight to assign to each touchpoint is a concern, especially for rule-based multi-touch attribution models. Applying your attribution weightings might not be biased or prone to error.

External factors

The multi-touch attribution model primarily focuses on digital user-level insights. External factors like exposure to television or print ads affecting marketing efforts and conversions may be crucial. If aggregate data is not used, seasonality may also not be considered.

Data wrangling

As mentioned earlier, combining and connecting the data is a headache. None of the models may provide complete visibility into the customer journey. Therefore, multiple models may need to be employed to correlate the data from each for the most accurate insights. The time spent aggregating the data in a digestible way may not justify the meaningful insights derived. The need to use multiple attribution models worsens this challenge.

First-touch attribution versus multi-touch attribution

With the examples above, now we know what a first-touch and multi-touch attribution model is. 

The first-touch attribution model best shows how top-of-the-funnel marketing generates awareness of a product, service, or brand. With that said, the first-touch attribution model is typically best for niche situations as it only helps understand the touchpoints which are effective at the beginning of a customer’s journey. 

If you want to understand more about your customer’s journey, use a multi-touch attribution model.

Conclusion

There are three (3) key characteristics I hold dear based on my experience when it comes to implementing a marketing strategy:

• Simplicity
• Scalability
• Consistency

To implement your multi-touch attribution model, make sure it is simple to understand and implement. Make sure that it can scale. And apply it consistently.

With that said, use an attribution model which you can easily understand and implement. Make sure you can quickly scale it from one attribution model to another when needed. Consistently monitor your data.

If you’ve been trying to implement any of the multi-touch attribution models mentioned and you are struggling to measure the impact of your marketing campaigns, don’t fret.

While you continue to test and rinse, you’ll eventually get there. 

If you still feel lost and really have no idea which marketing channels bring you the most revenue, kindly contact us to book a consultation so we can help you track and measure the impact of every single touch point in your marketing funnel.