What you will find in the article:
- Differences between multi-touch attribution and marketing mix modelling
- Using both methods enhances marketing effectiveness
- MTA provides a detailed insight into digital campaigns
- MMM captures the influence of offline media and external factors
- Both methods have their strengths and weaknesses
Almost every month, I meet marketers who are faced with the same question. A supplier or consultant advises them to use either multi-touch attribution (MTA) or marketing mix modelling (MMM). It often sounds as though these are two competing methods and that one must be chosen.
But that’s the problem. Over the past two years, the view has become increasingly prevalent that MTA is on the way out due to cookie restrictions and that the future belongs to MMM. The whole debate is thus often reduced to the question of which method to choose. But in my view, that’s the wrong question. MTA and MMM do not address the same problem. Pitting them against each other is akin to asking whether a microscope or a telescope is better. Both serve to observe reality, but each from a different perspective. Once we understand what each method is designed for, it ceases to matter which one is ‘better’. What will be more important is how to use them together.
What attribution is used for
Multi-touch attribution usually replaces the highly misleading last-click model. MTA tracks individual customer interactions with the brand throughout the entire journey and distributes credit for the conversion amongst all relevant touchpoints. It does not rely solely on the last click, but seeks to show the role played by individual channels and campaigns.
It is precisely this level of detail that is its greatest benefit. Thanks to this, it is possible to determine, for example, that a customer first came across the brand via a video on Instagram, returned two weeks later via a branded search, and only made a purchase following a retargeting campaign. Such information enables campaigns to be optimised on an ongoing basis, whilst they are still running.
Furthermore, practical experience shows that marketers’ demands for the level of detail required to evaluate campaigns are gradually increasing. When I started at Roivenue in 2019, a large proportion of clients were satisfied with evaluations at the individual channel level. Today, campaign-level analysis is practically the bare minimum, and it is not uncommon for clients to need to track results right down to the level of specific creatives.
And it is precisely at this level that most day-to-day marketing decisions are made. A media specialist does not reallocate budgets once a quarter. They do so every week, sometimes every day, at the level of individual campaigns, target audiences or creatives. It is precisely here that MTA provides insights that more strategic methods are unable to offer.
Let’s imagine a customer who sees your advert on Facebook, clicks on it, but doesn’t make a purchase. The next day, they search for the brand themselves via Google and complete the purchase. The last-click model attributes the entire sale to organic or brand search. The Facebook advert that kicked off the whole process is practically non-existent in the report. It is precisely for this reason that campaigns focused on brand building – such as display advertising, video or brand campaigns on social media – tend to be undervalued in the long term. In data based solely on last-click attribution, their return on investment appears worse than it actually is. A comprehensive attribution model, which takes into account not only clicks but also impressions, can provide a fair comparison across all digital campaigns
The benefit of such a detailed view is clearly demonstrated, for example, when evaluating the effectiveness of individual display ad placements. Oneplay, for instance, advertises a wide range of its programmes and, thanks to this attribution-based approach, is able to assess how specific combinations of promoted programmes and ad placements perform. Whilst some yield very good results, in other cases the advertising does not pay off. This does not necessarily mean that the right target audience is not visiting those sites. The reason may simply be that a large proportion of these users already have a Oneplay subscription.
Where MTA reaches its limits and when MMM is appropriate
However, this does not mean that MTA is a one-size-fits-all solution. Its greatest strength – the ability to track individual customer journeys – is also its greatest weakness. By the very nature of the methodology, attribution provides a granular, tactical view; it does not work as well with incremental effects and is susceptible to the quality of the input data.
Furthermore, MTA is, by its very nature, primarily a digital method. It cannot reliably capture the impact of television advertising, outdoor advertising, print media, or the effect of a strong brand spread through customer recommendations. If a significant portion of the marketing budget is allocated to offline channels, it remains largely blind to them. It is precisely in these areas that marketing mix modelling demonstrates its greatest strength.
Unlike attribution models, MMM does not track individual customers or their purchase journeys. It works with aggregated data and uses statistical models to identify the relationship between marketing investments and business outcomes over time.
As a result, it is able to capture what MTA, by its very nature, cannot – the influence of offline media, long-term brand building, or external factors that affect marketing performance. It includes, for example, seasonality, the economic climate, competitor activity or other circumstances that are not directly linked to individual campaigns but nevertheless significantly influence results. Because MMM does not work with data at the individual user level, it is not reliant on cookies or the ability to track customer journeys. The data required for marketing mix modelling is, quite simply, more readily available.
However, this does not mean that one method is more accurate than the other. Each has its strengths where the other reaches its limits. MTA offers a detailed view of digital campaigns and enables their ongoing optimisation. MMM provides a strategic view of the entire marketing mix, including offline activities, but requires more than a year’s worth of historical data to function properly, and the outputs are almost never available ‘in real time’; on the contrary, they are often delayed by several months. The level of granularity is, at best, at the channel level, and sometimes even at the platform level. On its own, therefore, it cannot answer all the questions that marketing seeks to address.
The difference between MTA and MMM results is not a problem; it is information
Anyone who has ever compared MTA and MMM outputs knows that their conclusions often differ. A channel that looks excellent in the attribution model may perform worse according to MMM, and vice versa.
We recently saw just how crucial the correct interpretation of these differences can be with one of our clients, when we compared the results of their internal MMM with data from Roivenue. The measured effectiveness of remarketing itself was almost identical in both cases. However, the results for the other channels differed significantly, which fundamentally altered remarketing’s position in the overall comparison. Whilst one methodology ranked it amongst the most effective channels, the other showed it to be the least effective.
Many marketing teams perceive this situation as a problem. They regard it as an error that needs to be resolved. In doing so, however, they are depriving themselves of the opportunity to gain a better understanding of their marketing. The difference between the results of the two methods is not in itself an error, nor is it a sign that one of the methods has failed. It is an indication of a certain degree of uncertainty. Each method works with different data, a different time horizon and a different level of detail. If their conclusions differ, this presents an opportunity to discuss why this is the case and what it tells us about the channel. Often, it is also a situation where significant improvement can be achieved.
Third perspective: incremental experiments
When MTA and MMM differ substantially on an important marketing decision, incremental experiments come into play. The principle behind them is relatively simple. The company deliberately reduces or, conversely, increases its marketing activity in only part of the market, region or target group, and then compares the results with a control group. This makes it possible to determine the actual impact of a specific campaign much more reliably. For marketers, this is, in practice, one of the most direct ways to verify the actual benefits of a specific marketing activity.
Of course, experiments cannot be carried out for every campaign. They are time-consuming and make sense primarily for significant investments or where individual methods yield differing results over the long term. This is precisely why they should not be seen as a substitute for MTA or MMM. Their role is to help make decisions where other approaches disagree, and subsequently to refine the interpretation of their results.
The problem with the modern ‘MMM and experiments are enough’ approach
Recently, there has been a growing view that, with various changes to user measurement capabilities and new restrictions, attribution models are losing their relevance. The argument is simple: tracking user journeys is becoming increasingly difficult due to privacy concerns, so the best approach is a combination of marketing mix modelling supplemented by experiments.
At first glance, this makes sense. MMM shows how to allocate the budget across individual channels, experiments verify that the effect is genuine, and the problem is solved. However, this is where a fundamental gap arises. MMM helps with strategic decision-making. It typically works with weekly data and is updated at best once a month. Experiments, on the other hand, require time for both preparation and evaluation, and it makes sense to use them only for more significant campaigns or investments.
Marketing, however, isn’t run on a quarterly basis. Every week—often every day—someone decides whether to adjust a specific campaign’s budget, change the targeting or swap out the creative. These operational decisions aren’t going to disappear; they’ll still be made based on whatever data is currently available. And when that data isn’t good, decisions will be made primarily on the basis of gut feeling.
If a company lacks a high-quality attribution view, the marketer will fall back on the data they have to hand – figures from Google Analytics or directly from advertising platforms. This data has well-known limitations. Each platform tends to attribute credit to itself; it fails to see part of the customer journey, and this leads to duplication of reported conversions. This creates a paradoxical situation. Strategic decisions are underpinned by a high-quality model, but day-to-day operations rely on metrics that we know provide a distorted picture of reality. Neither MMM nor experiments are therefore a substitute for MTA. They address a different level of decision-making and work best in conjunction with one another.
It is not a question of choosing one method
When we take a step back and look at the whole issue, it becomes clear that the question ‘MTA or MMM?’ doesn’t actually make sense. Today, mature marketing teams do not pit individual approaches against one another, but combine them.
MMM helps to decide on the long-term allocation of investment across the entire marketing mix and also includes offline channels that MTA cannot see. MTA, on the other hand, enables the ongoing management of digital campaigns and the making of day-to-day decisions at the level of individual activities. Experiments then serve as an independent check when the two methods contradict each other or when significant investments need to be validated. However, each of these approaches has its blind spots, which is precisely why they deliver the greatest value when used together.
What this means for businesses
Until recently, MMM was mainly associated with multinational brands that had huge budgets and years of historical data. Today, this is no longer the case. On the contrary, even medium-sized companies and online shops can benefit significantly from a combination of MTA, MMM and experiments. Often even more so than large corporations. Smaller firms are able to react more quickly, and knowledge of specific channels is concentrated among a smaller number of people. As a result, the actual preparation of models often takes much less time. The first step need not be a complex model or an extensive analytical project. All you need to do is stop evaluating individual channels in isolation based on their own reported performance. It is precisely this approach that leads companies to shift budgets towards channels that are easy to measure, rather than those that actually contribute to growth.
If someone asks you to choose between MTA and MMM, try asking a different question. It’s not about which method is better. It’s about what decision you need to make right now. If you’re working on a long-term strategy and the allocation of your marketing budget, MMM will play a key role. If you need to optimise digital campaigns on an ongoing basis, MTA is the right choice for you. And if their conclusions differ on a decision that really matters, it’s time to experiment.
That doesn’t mean, however, that companies have to implement all three approaches at once. It makes more sense to start with the method that addresses their biggest current problem, and add others as their needs evolve. For many of our clients, for example, Roivenue is the first specialised tool they use to evaluate marketing effectiveness. However, as measurement quality improves, new questions gradually arise that attribution alone cannot answer. We also encounter the opposite situation: brands that have been using MMM for several years begin to use Roivenue, but gradually come up against a lack of detail in their day-to-day campaign management, whilst at the same time not wanting to rely solely on the reporting from individual advertising platforms.
Today’s best marketing teams are not looking for a single universal method or a single perfect figure. They know that each method captures a different aspect of reality. They do not gain a real advantage by choosing just one of them, but by being able to link all three approaches and utilise their strengths at the right moment. If someone asks you to choose between a microscope and a telescope, it might be better to reply that a good analyst uses both.
Source: mediaguru.cz