Try searching for
The number of touchpoints between prospect and brand before conversion or sale can vary from anything between 5-50. With so many touchpoints, mapping journeys of consumers who constantly bounce between channels and devices, online and offline, is more complicated than ever.
The difficulty understanding this complex path to conversion makes attribution ever more difficult for marketers. In fact, according to research by eMarketer, cross-device attribution is the second biggest (42%) challenge for media professionals in 2021.
How then can you, the marketer, know which touchpoints and journeys deliver the best performance so you can make informed budget allocation decisions?
Enter multi-touch attribution or MTA.
MTA assigns credit across touchpoints according to their impact. Makes total sense right? So why is it that MTA has been, for years now, such a pain point for marketers and suffers from relatively low adoption rates?
In this blog, we’ll answer these questions plus examine the various multi-touch attribution models, their merits and their limitations. We will also look into how multi-touch can deliver insights to help shape your marketing strategy, and how to choose the model that’s right for your business.
Let’s get started.
Multi-touch attribution is a marketing measurement method that takes multiple online and offline touchpoints along the customer journey into account, and then assigns credit to each based on varying logic per business.
Offline touchpoints include television (excluding smart TV), radio, print (e.g. billboards, coupons, and direct mail), in-store, call centers, and sales calls.
Online touchpoints are divided into paid, owned, and earned media across digital properties.
Paid media includes search, display, social etc, while owned media refers to websites, emails, demos, content marketing such as blogs, and your brand’s social media accounts. 3rd party social media and blog mentions are examples of earned media, as is press coverage.
In multi-touch attribution, touchpoints can be weighted equally or proportionally, depending on the model used (more on this below).
Marketing plan template to increase ROI
Single-touch can refer to both first or last-touch attribution.
First-touch assigns full credit to the first known point of contact with your brand in the consumer journey, whereas last-touch gives all the credit to the last known touchpoint.
Of course neither attribute across the entire user journey, so why would a marketer choose to use them?
If you are focused on introducing your brand and widening the top of your funnel then you may opt for a first-touch attribution model. That is because it’s relatively easy to implement and delivers insights on how customers discover your brand.
Conversely, last-touch attribution focuses on the bottom of the funnel and what factors helped convert the user, which is after all, the essence of performance marketing.
Multi-touch takes a more holistic view and covers the entire funnel, delivering insights across the user journey. Let’s explore this in more detail.
Clearly, both first and last-touch attribution ignore the reality of a modern consumer journey (5-20 touchpoints on average, remember?); That’s where multi-touch attribution comes in. It takes a more sophisticated approach to mapping the journey, and offers a number of other benefits.
We mentioned above that the customer journey is complicated by multiple devices, channels, and touchpoints. MTA helps to overcome some of these complexities by reflecting, more accurately, the actual user journey.
The better the representation of reality, the better the understanding of where influence lies and which combination of touchpoints are performing the best.
Having a more accurate picture of the user journey and the touchpoints along the way will help you better customize your messaging, meeting consumers on the right channel(s) at the right time.
Allocating credit along the length and breadth of user journey means you are not placing undue importance on the first or the last touchpoint. Instead you are able to allocate credit using data to support any argument over the influence of a touchpoint in the path to conversion.
For example: let’s say a consumer first sees one of your paid ads on Facebook, searches for your brand on Google, you remarket them on a news site, and then they go back to Google to search for your specific product and purchase.
Who’s to say that the paid touchpoint (the first) should get credit over search (the last touchpoint)? Or vice versa? That’s where an MTA model can save the day by making sure each channel gets its fair share of the success.
Multi-touch places all of the data in one model so there is no need to jump between reports trying to identify the value of a touchpoint. The result of having increased agility is the ability to adapt and pivot your strategy according to the insights gained.
For example, say a large chunk of your marketing budget was dedicated to a bottom-of-the-funnel promotional email campaign. However, your MTA model showed that a social media campaign towards the top of the funnel is proving to be very successful at getting prospective customers to click through to your mobile site and purchase.
With these insights you were able to see the impact of the social media campaign at increasing conversions. As a result you pivot your strategy and place more emphasis and spend on the social media channel, and less on the email campaign, to maximize your Return on Investment (ROI).
The risk of misunderstanding your customer journey means that you may be blind to which combination of touchpoints are delivering the highest ROI, and equally important which are underperforming.
Multi-touch attribution offers the insights needed to know when to reduce your marketing spend on the channels that aren’t delivering, and to leave more budget for doubling-down on the ones that are.
Despite multi-touch giving a knock-out blow against single-touch, it’s not as widely adopted as one would expect. One reason for this may be the fact that for multi-touch attribution to become an industry standard it would require buy-in from across the ecosystem to agree that specific models be used in specific circumstances.
In the table below we can see that even in the US, which is the most advanced market, adoption of MTA is still only at 65% as of 2021. Furthermore, the percentage increase year-over-year is minimal, meaning that opinions on using MTA are not changing.
Last-touch attribution remains the industry standard when it comes to billing — the mechanism through which attribution companies assign credit to media companies and essentially determining which party gets paid for driving the conversion.
For example, mobile apps pay ad networks for installs based on the last-touch attribution model. The last media source to touch the user before an install gets the full payment for delivering that user. However, for marketers looking to gain insights on the whole marketing funnel it is important to use MTA internally and in parallel. Understanding the value of each touchpoint will help allocate credit going forward.
So what’s the bottom line?
Even if conversions are paid according to last-touch, they should be evaluated within an MTA prism for smarter decision-making in the future.
There are several different MTA models, each with its own logic, and its own pros and cons.
The linear model applies credit equally to each touchpoint along the user journey. The linear model is good because it doesn’t apply bias and gives a picture of the entire funnel.
However, it fails to recognize if one touchpoint was more influential than another. For example, under the linear model, a paid first touchpoint is just as valuable as an organic search second touchpoint and a direct last touchpoint. But maybe the customer wouldn’t have ever known about your company without that first touchpoint meaning it was essential and should probably have more weight than, say, the last direct touchpoint.
The linear model oversimplifies the user journey and is considered the least sophisticated form of MTA.
When to use?
The linear model is best to use when you want to know which touchpoints exist but not necessarily the impact of each.
The time-decay model assumes that the closer a touchpoint is to the time of conversion, the greater the influence. The credit is therefore weighted accordingly, with the last touch receiving the highest percentage.
When to use?
Time-decay is a good model for companies focusing their strategy on the lower end of the funnel. Marketers use this model when there is a short consideration stage, such as a promotional campaign.
However, the time-decay model neglects to analyze the nature of the touchpoint. For example, in the case above there was a banner ad and an influencer video. If the banner appeared after the influencer video, according to the time-decay model it would receive more credit.
It is unlikely that a banner was more impactful in the decision-making process than an influencer talking about your product. So, to assign more credit to the banner just because it occurred after the video is often flawed.
The U-shaped model places its focus on two key touchpoints, the first and the last, and assigns 40% of the credit to each. The remaining 20% is spread across the intermediary touchpoints.
The U-shaped model is useful if you are looking to understand what kicks off and closes the customer journey. It is also good because it still acknowledges the intermediate touchpoints while assigning them less credit.
While this is a first good step at recognizing that not all touchpoints deliver the same value, it is still a relatively simplistic approach to carving up the credit.
When to use?
U-shape models are good for short sales cycles, like buying a low-cost product on an e-commerce app. The app marketer wants to place the most value on touchpoints that introduced the brand, and touchpoints that converted the user.
Taking the U-shaped model one step further is the W-shaped model. Again this credits the first touchpoint, but also the lead creation and opportunity creation touchpoints.
In the W-shaped model, each main touchpoint is assigned 30% and the remaining 10% is assigned among the intermediary touchpoints.
For example, a prospect is browsing for a holiday package. Their search offered them a number of ads relating to different destinations. They click on a booking.com ad. They like the idea of a holiday on the Mediterranean so they search the mobile website for some more information. They register to the site and add the destination to their wishlist, which is considered the “lead creation”, but don’t book the holiday.
A few days later they get an email offering a 10% discount for a holiday in Cyprus for users who install the app. They click on the link in the email, which leads them to the App Store, download the booking.com app and complete the purchase. The discount code is automatically entered through clever deep-linking technology. The email is considered the “opportunity creation”.
The W-shaped model is good because it has these identifiable stages, however it’s not always so simple to differentiate the stages and so it can over-complicate the calculation.
When to use?
The W-shaped model should be used for longer sales cycles that often have extra touchpoints that are significant to the progression of the customer journey.
Full path accredits four main touchpoints; the three mentioned above in addition to the final sale/closing touchpoint. Each of the four are attributed with 22.5% of the credit and the remaining 10% is spread between the intermediary touchpoints.
Full path is a good option for sales teams as it offers insights beyond the opportunity stage and provides a more granular view of the funnel.
When to use?
Similar to the W-shaped model, full path is intended for companies with a long sales cycle and extended customer consideration stages, for example high-value purchases like a car.
A custom model lets you create a version that fits your business needs. None of the other models mentioned are neither good nor bad, but intended for different business use cases.
However, the custom model allows you to tweak and create a hybrid model, assigning a higher percentage to a specific stage in the funnel for some campaigns and lower for others.
A custom approach delivers flexibility and granular control to the marketer, but is more complex to set up and can offer some maintenance challenges.
When to use?
A custom model is able to fully address the complexities of the entire customer journey as they apply to each business specifically. Apps that require extensive touchpoint measurement for example, travel apps, will benefit from a customized model.
Markov chain gets its name from Andrey Markov, a Russian mathematician who developed a theory on stochastic processes in probability.
His theory described a sequence of possible events where the probability of each event depended on the “state” of the previous event.
And what has this got to do with multi-touch attribution you ask? Well, the same theory could be applied to advanced MTA modeling, creating a framework in which marketers can more accurately model the user journey.
Using a Markov chain model combined with data from a campaign, marketers can identify the probability of a prospect moving from one touchpoint to the next.
For example, a customer sees an ad on LinkedIn (first touchpoint), then browses the company’s website (transition), before clicking on a product video relating to the same product (second touchpoint).
The number of times customers transition between touchpoints is calculated into a probability. The complete chain depicts the most likely path to success and the importance of each state in the customer journey.
When it comes to multi-touch attribution, there is no one size fits all approach.
Ultimately, it will depend on the specific KPIs for your app and your campaign. For example, if you’re measuring the number of app downloads then the time decay model may work best. But if you’re measuring LTV you may want a model that covers more touchpoints even after conversion such as the full path model.
Before you decide which model to use it’s crucial to determine your KPIs and then decide which model is the best fit.
You should also test out different models and see which one aligns most closely with your strategy. If one model doesn’t work, try another until you get the insights you need.
Finally, compare the results between the different models and see where you can optimize for improved results.
We have discussed how the various multi-touch models allow marketers to assign credit in more sophisticated ways. However, for a truly holistic view, MTA should be combined with Marketing Mix Modelling (MMM).
MMM is traditionally used for budget planning and strategic purposes and takes a top-down approach to attribution modeling. Conversely to many digital attribution models that rely on user-level real-time data (or as close to real-time as possible), MMM insights are aggregated and referred to as part of quarterly or even annual strategic conversations.
MMM accounts for a number of other influencing commercial factors beyond marketing and advertising, ranging from macroeconomic conditions, to seasonality and even the weather.
For example, we saw in 2020 that app downloads climbed 33%, which was largely attributed to COVID-19 lockdown measures and people being asked to stay at home. Using MMM, which accounts for the pandemic in its calculations, helps marketers assess the impact of external influences and can be used for strategic planning.
Using both MTA and MMM together adds to the agility we mentioned earlier. Marketers are able to see both sides of the coin : the strategic side with MMM, and the tactical side with MTA.
For example, is a 10% discount code sent via a push notification on a taxi ride app better received in a rainstorm or during a heat wave? The data attained through MTA combined with the insights on external factors will help marketers make smarter decisions.
Implementing your multi-touch attribution strategy can be summarized into a 5-step process:
As we mentioned above, your KPIs will guide your strategy. If your primary goal is user acquisition then you may choose a different model than if you’re measuring LTV or the uninstall rate. Defining what you’re measuring will help guide which attribution model is best suited to your needs.
Today, most organizations store customer data in a CRM system, but if not maintained and cleaned regularly, your data could end up causing you problems.
Do a thorough audit of your data to ensure it meets your quality standards. Make sure you fix missing contact fields, remove duplicate data, ensure data is inputted into the correct field, and correct typo’s or other inaccuracies such as out of date job titles or companies.
Poor data can leave holes in your analysis and lead to incorrect assumptions and ultimately wasted budgets.
With the introduction of iOS 14 and other privacy measures, the quality of your data and the safeguards you implement around it have never been more important.
In fact, data quality was cited as the biggest barrier to revenue attribution success (43%) globally, so making your data clean early on in your attribution process will pay off in the long run.
Use an analytics software that can identify the role of each touchpoint to analyze the data once it has been collected. Doing so will help make it easier to identify any insights and address app campaigns that require optimization.
Continually review the metrics and assess if any channels are underperforming. If you find this to be the case then adjust your campaigns accordingly. Your attribution should always reflect your business goals. If your goals change then you may need to optimize or regroup and try a different attribution model.
Once you have identified trends and understand the data patterns, apply these insights to future marketing efforts to improve the performance of campaigns and channels.
Remember this is an ongoing process, so when you acquire new insights be sure to apply them and improve your results.
Multi-touch attribution, for all of its benefits, is not without its challenges. Let’s discuss a few of them here.
As mentioned above, despite the fact that last touch falls short, it is the industry standard for billing.
Why? Because multi-touch is complex and requires buy-in from multiple parties across the ecosystem.
For example, you elect to use the W-model. That would mean the three main touchpoints had agreed to take 30%, instead of the last click receiving the full 100%.
All the partners, advertisers and ad networks need to be on the same page, which requires collaboration across the industry — something that has just not materialized in a meaningful way.
multi-touch models are both challenging to implement and difficult to analyze.
Not all companies have the talent required to deploy the models or the ability to glean the insights they deliver. In fact, data shows that as few as 9% believe that their organizations have a “excellent” understanding of multi-touch attribution.
The path some attribution models take to reach their conclusions remains a mystery to many marketers, making it difficult to validate the results. For example, attribution models will always credit an action to a touchpoint, even though marketers know some actions happen organically.
Marketers can use incrementality testing to identify which conversions were related to a specific marketing campaign, and which would have happened on their own, giving greater clarity to the attribution results.
Incrementality is not a new concept and has been largely overlooked due to its complexity in both execution and analysis. However, it has recently gained traction as a useful tool post iOS 14 to help fill the gap with data-driven insights.
Multi-touch attribution models are intended to factor in multiple channels and devices. These channels should in theory include offline channels such as TV, radio, and print.
However, it is very difficult to attribute these channels and there is a severe lack of data relating to them. Together, this creates a challenge as to how to aggregate any data you do acquire into your chosen model.
None of the models is a perfect solution.
Each campaign will have different requirements and goals. It is therefore important to assess what these are before starting your campaign and apply the model that is the best fit in this instance.
Remember, you can always change the model if you see you aren’t getting the insights you need. Sometimes it’s simply a case of trial and error.
The limitations on 3rd party cookies and the depreciation of Apple’s IDFA mobile ad identifier will make attribution more challenging because it will be harder to identify customers across channels. Instead, aggregated level attribution and measurement will take center stage through methods like incrementality testing.
Relying on 1st party data will also become much more important, for example, by directing users to a brand’s mobile website and from there to the brand’s mobile app (in such a scenario there’s no need to collect the IDFA).
Multi-touch attribution gives you the ability to analyze the impact of each touchpoint along the customer journey and identify which are delivering the highest value.
To understand which model best fits your need you should remember:
Multi-touch attribution assigns credit for a conversion across all customer journey touchpoints, helping marketers understand which touchpoints are most effective and aiding budget allocation.
Common multi-touch models include linear, time-decay, U-shaped, W-shaped, and full path, each with different ways of assigning credit across touchpoints depending on the journey’s length and complexity.
MTA enables marketers to allocate budget efficiently, customize messaging, and gain agility by understanding which touchpoints are driving conversions. It takes into account the complexity of the customer journey while eliminating biases.
MTA is complex and there’s no one-size-fits-all approach. It requires cross-industry cooperation and advanced analytical skills, and its lack of standardization makes adoption and validation difficult for many companies. Privacy changes add further complexity.
With increasing privacy regulations, relying on first-party data and methods like incrementality testing is key for accurate attribution while respecting user privacy.
