Ultimate Attribution Guide: Marketing Measurement for 2024 and Beyond

Updated August 2024 

There has been a lot, and I mean a LOT, of discussion in the advertising industry over the past three or four years about tracking and measurement. Specifically, how marketing mix modelling and multi-touch attribution fare in 2024. Due to increased customer privacy concerns from global governments and policy changes coming from large tech organizations (looking at you Apple), many of the digital tracking methods we’ve all come to love and loathe are becoming obsolete. These changes are forcing marketers and ad agencies to re-think their evaluation of media and how they attribute their efforts.

The industry is currently is evaluating two approaches.

• Stay the course and find an alternative tracking method that allows advertisers to continue to market on a one-to-one basis.
• Dust-off tried and true ‘old-school’ marketing measurement methods and look at them with a new eye.

Let’s look at each of these approaches in a bit more detail.

First Approach: Keeping it one-to-one

One of the main reasons digital advertising channels were embraced so quickly by advertisers, particularly social media and programmatically bought display/video, was it allowed marketers to do what they had long been dreaming of; reach the right user, at the right time, with the right message… then track them from site to site and back to their own until they converted to a sale. And for a while the dream was alive and well. The level of data that publishers, ad tech providers, brands and ad agencies were able to collect from online users was unlike anything they’d seen before. Performance marketing thrived (often at the expense of branding), cookies were dropped into browsers at will and mobile SDKs were stuffed with location trackers.

However online users and lawmakers started to realize how much online activity was being tracked and monitored leading to tough privacy laws like GDPR and CCPA being introduced. Some browsers stopped allowing third-party cookies and Apple introduced its App Tracking Transparency update which required iPhone and iPad users to expressly consent being tracked by app makers. All these changes have been a major disruption to the ad industry and seem to signal the end of an era for user tracking, but the ad industry isn’t ready to give just yet. The industry has gotten used to having user level insights and that’s hard to give up. So, in the face of losing third-party cookies, there has been a major push to come up with an alternative user tracking method.

In the quest for user level data, several identity-based tracking solutions have emerged. These solutions include:

1. The Trade Desk’s UID 2.0. At a high level, UID 2.0 works by changing an online user's email address into an alphanumeric identifier (a universal ID) which can then be used to connect their activities online as they go from site to site.
   • The program is reliant on having a neutral organization host the program. It looked like IAB Tech Lab would take it on but they have recently declined.
2. LiveRamp’s ATS. LiveRamp’s approach is to create a user identity graph that utilizes a mix of user identifiers such as email,  phone #, username etc. to create an encrypted identifier that can be used to target online users as they surf the web.

These solutions are promising, and are getting quite a lot of adoption but as we see companies like Apple giving its users the option to hide their email address when registering with a website or making a purchase, it brings up questions as to how sustainable an email-based identity tracking program could be.

Seems like a lot of work is going into this. Why is having user identity important?

Two reasons. First, user ids are the foundation of programmatically bought digital advertising. To show an ad to the right person, at the right time with the right frequency… well you get it, you need to know who that person is and what they are uniquely interested in. This requires identity tracking in some shape or form.

The second major use case is multi-touch attribution (MTA).

MTA is the process of giving each touchpoint someone has with your brand credit for the ultimate sale or conversion. Even if it’s only a fraction of the credit. There are many different types of models a brand may use (linear, time decay, U or W shaped, etc.) but the goal is generally the same, understand which online marketing efforts the user was exposed to before converting.

For the last decade ad agencies have looked to MTA to help gain better insights into the online user’s path to conversion and provide guidance when optimizing campaigns. Unfortunately, the data used in MTA is incomplete and can lead to incorrect assumptions about which channels are generating conversions.

What Makes Multi-Touch Attribution Incomplete/Inaccurate? 

• MTA only looks at digital marketing efforts. If a brand is running on radio, buys an OOH billboard on the DVP or sends out flyers/coupons, it’s not included in the model.
• MTA models don’t factor in external factors that affect sales in the real world. Things like seasonality, competitor activity, temporary price reductions, etc.
• Without 3rd party cookies, cross channel tracking will be limited. As a result, most MTA models will only work on touch points that occur within the same walled garden.
• If sales are occurring offline at a retail location it is hard to attribute to any given channel in an online only analysis.
• Online conversion events must remove invalid bot activity from real humans but there is an inherent bias for media vendors to claim more conversions versus less..

So, while MTA is still useful for providing media buyers with direction on which channels to buy or targeting methods to deploy, they hardly give a complete look at the user journey. For that, marketers would need to turn to a marketing measurement that encompasses all the factors that got into a purchase. It’s time to go old school but with an updated twist.

Second Approach: Updating Tried and True Methods

To truly understand how each marketing tactic a brand has in market is contributing to its overall sales, advertisers need to utilize a measurement methodology that includes both online and offline media but also considers the impact of non-media factors like COVID, weather, sales, competitive activity, etc. This is where Marketing Mix Modelling comes in.

What is Market Mix Modelling? 

Marketing Mix Models have been around since the 60’s and its main purpose has been to help advertisers understand how different factors in their marketing mix impact sales. Technically speaking, MMMs are “a statistical analysis that uses multivariate regressions on sales and marketing time series data to estimate the impact of various marketing tactics on sales then forecast the impact of future sales on a set of tactics” (thanks Wikipedia!). In other words, MMMs look at how various media channels and external factors have affected sales for that specific brand or product in the past. Then, it uses those insights to forecast future revenue.

The equations used in MMMs go far beyond a few simple ROI calculations. MMMs include things like ad stock, impression volume, ad saturation/diminishing return levels, ad frequency and more. MMMs can also be used to compare the return on ad spend (ROAS) of each media channel at a high level or break it down by targeting method, creative execution or by “walled-garden”’.

Arima’s Marketing Mix Model displays ROAS per channel 

MMMs are also naturally privacy compliant since no unique customer data is required to create one. Also, if an advertiser or agency doesn’t have sales data success proxies such as brand lift, website traffic, footfall traffic or even search volume can be used.

Not Your Grandfather’s MMM

Marketers have used marketing mix models for decades, but they have changed a lot in the last 5-8 years. In the past, MMMs were expensive and only gave a snapshot of one point in time. This meant only big brands could afford them, and their insights quickly became outdated. With better technology and APIs, companies can now create MMMs that provide ROAS insights almost in real-time and at a much lower cost. Arima’s Self-Directed Marketing Mix Model (SDMMM) goes even further by giving the user control. Arima’s SDMMM is easy to use, so you can create an MMM with your sales and marketing data in just 5-10 minutes.

Functions like sliders help users visualize the impact of different marketing budgets.

Sounds great, right? But there are a couple things to note:

1. Marketing mix modelling is a backward-looking analysis so if an advertiser is looking to add a new media channel to the campaign mix it won’t be able to predict its impact on sales. If a brand wants to see how a new channel will impact sales, they will need to run a campaign on it first then (generally for a month or so) to test its effectiveness.
2. MMMs can tell advertisers a lot about the impact each media channel is having on its bottom line but additional factors such as editorial context, brand alignment and tactical adjustments need more detail than MMM can provide.
   
   

MMM and MTA are not mutually exclusive MTA, they can be combined. Using the power of both approaches, brands can create a unified marketing strategy with greater insight into the most effective ways to reach their customers.

Hybrid Approach: Google’s Viewpoint

Google has been pushing a combined hybrid approach where MTA and MMM can co-exist. Their stance is that MMM, being occasional, feeds the attribution scheme but does not receive feedback from it. In this view, MMM acts as a periodic input that supports the continuous process of MTA without the bidirectional influence.

Google demonstrated how MMM, Incrementality Experiments, and Attribution work together. 

A hybrid approach demands the user choose the overriding scheme since both MMM and MTA will report a number for the effectiveness of Paid Search and other channels. The dilemma is which number to trust?

Arima’s POV: Actionable Hybrid Marketing Mix Models 

Since MMM’s scope is all media, we believe the integration of tactical optimization needs to happen through the use of Geography instead of cookies and IDs which can only address digital media. It makes sense from  Google’s perspective to focus on digital media and MTA to propose a hybrid model, however the reality is that all media can be tested and optimized through tactical spend allocations. Further, all media has synergistic effects on each other, traditional broadcast media pushes enormous search intent volume and cannot be ignored despite its lack of 1:1 ID data for MTA. For this reason, we believe digital and analog channel ROAS numbers in MMM prevail over anything reported in MTA. 

A hybrid structure which includes geography enables measurement and testing in all media pushing a focus on the planning and buying tools to accommodate the advertiser's needs.

The results of using these methods.

Consider the following approach:

• Use MMM to guide strategic decisions and understand the synergistic impact of your media mix.
• Use MTA for tactical, short-term optimizations within digital channels.
• Use GEO to design and perform market tests
• Use Media Planning and Buying tools which integrate with MMM and enable multiple market measurement.

Using the Arima approach, you create a balanced strategy that leverages the strengths of all  the available data, ultimately driving more informed and effective marketing decisions.

What attribution method works best in 2024? 

Although both MMM and MTA approaches have their merits and drawbacks, we see Marketing Mix Modelling is the frontrunner for several key reasons. For example, Arima’s Self-Directed Marketing Mix Model has the following benefits: 

• Comprehensive View: Unlike MTA, which only focuses on digital attribution, Marketing mix models provide a complete picture of how all marketing activities interact and influence outcomes.
• Strategic Decision-Making: By analyzing media spend along with the effectiveness of other marketing levers like pricing and promotions, MMMs empower strategic resource allocation across the entire marketing landscape.
• Sales Forecasting: Leveraging historical data, MMMs can forecast the future impact of different marketing strategies. This foresight allows for proactive planning and informed resource allocation decisions.
• Simple and Fast: Arima's platform empowers any team member to handle attribution. The SDMMM takes just minutes to set up and is designed to be easy to use and understand. 

In Summary

Weathering the many challenges that come with a ‘digital only’ approach to campaign analysis (ad blocking, fraud, a patchwork of local and global privacy legislations, third-party cookie loss and a series of social media measurement inconsistencies, etc.) has left many marketers looking for an alternative, future-proof way to measure the returns of their marketing efforts.

As we move forward we are likely to see advertisers using modern, actionable Marketing Mix Models as the foundation of their marketing analysis and campaign planning.

If you're ready to stop guessing and start knowing, investing in marketing mix models could be the game-changer you need. Reach out to us today to learn more.