The difference between analysis with aggregate vs. disaggregate data
One of the most practical and universally applicable ways to deal with marketing measurement is to combine disaggregate tracking insights with aggregate analysis. To aggregate data is to compile and summarize data while to disaggregate data is to break down aggregated data into component parts or smaller units. For those of you familiar with this space, I’m talking about using Marketing Mix Modeling (MMM) to provide a holistic view of incrementality (efficiency), coupled with either the analysis of disaggregate data or a full-blown Multi-Touch Attribution Model (MTA), to gain detailed reporting for specific activities that warrant it.
To cover our bases, Marketing Mix Modeling (MMM) is a top-down or aggregated data form of modeling. Popularized in the 1990s, MMM is a reliable way to measure marketing effectiveness and understand the impact of marketing activities on business performance. Identifying the marketing drivers behind performance helps companies understand which drivers are responsible for changes in contributions to sales over time. These learnings when applied to optimizing media budgets help drive long-term performance. But it is also beneficial to couple your modeling with detailed marketing tracking whenever possible.
Multi-Touch Attribution (MTA) provides you with a more granular, person-level view than Marketing Mix Modeling’s aggregate method can ever hope to achieve. It works by looking at individual channels within your marketing mix and collects user-level data to help you determine the impact of your individual touchpoints at the customer level. This type of data allows you to determine the success of interactions across the entirety of your consumer’s journey – at least that’s the theory.
MTA has rather specific data requirements. In some cases, either the data that is available or the cost of the analysis makes it impractical. Often, proper analysis of disaggregate data can serve as an effective surrogate, filling the gap between MMM and person/spot level detail. This approach allows brands to benefit from the common sources of detailed digital data without sacrificing the strategic benefits of aggregate modeling.
If you are interested in learning more about marketing effectiveness through modeling and how you can bring this process in-house, you can check out our case study, “How to build an in-house Marketing Mix Modeling Capability" here.
Blending Aggregate and Disaggregate Data to Make a Unified Model
Taken in isolation, both MMM and MTA can provide valuable insights. However, by blending these two types of insight, you can avoid some of the shortcomings that they each have. For example, Marketing Mix Modeling offers insights into overall consumer trends over the long-term, but it’s not always possible to use this older data to inform and influence current campaigns nor may results be granular enough to make tactical suggestions on a micro-transaction level. Multi-Touch Attribution will allow you to gauge the impact of each of the individual digital channels across your marketing mix but often ignores certain elements of the marketing mix and rarely allows you to determine customer trends, account for external factors, or changes like you might be able to see with Marketing Mix Modeling.
By combining both user-level data MTA and aggregate data MMM, you can create a unified model and leverage the strengths of both techniques. Using this blended approach can help you to do things like forecast, estimate your marketing ROI (Return on Investment), and get a more comprehensive view of your overall marketing effectiveness at both the aggregate level and the disaggregate level.
Forrester coined the term “Unified Measurement” in 2016 to describe this process. You aggregate data to a level where you can compare all relevant factors, you use aggregation to unify data so that you can objectively compare all your options, and finally, reveal a holistic picture of what drives your business.
Unified Modeling in Action
As an example, let’s say you sell computers directly via your website and through one national retailer. You also support your brand through Online Display Ads, Paid Search, Paid Social Media, and Cable TV, but additionally, you run price promotions and purchase in-store displays (in brick-and-mortar). You know that external factors such as the economy, the time of year, and even competitive actions impact your sales. You don’t know who actually buys your computers from the brick-and-mortar retailer, and even for your online sales, you don’t know everything that each customer was exposed to. With this in mind, you have two real goals, each coming from a different internal audience. First, you need to build a data-driven case for spending to take to your Private Equity owners, keeping in mind that it needs to tie back to the overall financials of the business and be strategic, or you’ll be laughed out of the room. And second, your marketers, especially the social media team, need more detailed data to optimize the impact of that particular budget.
With these goals in mind, your best solution is to use Marketing Mix Modeling to estimate the sales impact from each relevant factor – the overarching business drivers. If you refer to the picture below, you’ll see that your MMM work shows you which tactics are most efficient and the changes that will drive growth. In the example all spends change, but for now, let’s focus on Display vs. Search. The analysis shows you that you can reduce spend on Display and suffer only small declines in Profit generated. If you move some of that money to Search it will increase overall profit because the marginal response is higher.
Now you have a great case for strategic planning and for your senior audience, but you also need to support real-time decision-making for the marketing team. You don’t have “perfect” data at this lower level, but you do have details for several of your key spends. This is easy to access as API connections are populating the marketing dashboarding system. It’s the same system that houses and prepared your data for the Marketing Mix Modeling work, it's just more detailed.
While the incremental ROI of each overall tactic, one of which being Paid Search, is being reliably estimated in the Marketing Mix Model, additional details useful in execution come from the detailed reporting. This lets you see specifics such as campaign, ad group, and even creative, and evaluate them via interim KPIs (Key Performance Indicators) such as web traffic, social interaction, promo tracking, and period comparison trends, to help you compare success rates. For example, converting an ad view into a website visit, or increasing the time of engagement with an ad, tells you that you’re meeting your “within tactic” goals, even though they don’t, by themselves, represent your strategic goal of driving sales and profit.
Now, let’s look at the Google Ads campaign performance table below. The brand budget optimization is recommending an extra $1.2M for Search. One of your strategic goals is to gain market share, so you are trying to increase search targeted at competitors. The information below lets you see which competitive search terms have the best Click Through Rates (CTR) and lowest Cost Per Acquisition (CPA). With this knowledge you will increase spend (i.e., more aggressive bidding) for the search campaign called Competitive TV Video. And since these dashboards update daily and/or weekly you can track progress based on these “within tactic” KPIs and continue to make changes in real-time.
Breaking down by campaign and/or ad group can pinpoint successful, or unsuccessful messaging. This works by first calculating Click Through Rate (CTR), Customer Acquisition Cost (CAC) and then plotting them via time series with promotional campaigns marked on the timeline. Creative can be used to take this one level deeper. Clicks, Click Through Rate, and Cost Per Acquisition can also be broken down by geographic region or by aggregate demographics such as age groups, sex, with/without kids, to assess performance based on preference and/or common demographic trends.
If you expect to keep up with the ever-changing world of marketing, you must learn how to leverage several types of data at the right time in the right places. By using the blended or unified approach, you can do things like forecast, estimate your marketing ROI, and get a more comprehensive picture of your marketing effectiveness. That being said, it’s false to suggest all data analysis must be conducted using both aggregate data and disaggregated data. Both Marketing Mix Modeling and analysis of disaggregate data have their time and place.
At ScanmarQED, we’re constantly striving to help you in getting actionable insights that allow you to make smarter marketing decisions. If you have any questions related to marketing measurement and modeling, please contact us by clicking the button below, and we’ll be happy to discuss how we can help you and your organization.