Though many companies have some process in place to measure marketing ROI, few are doing it right. That's because today's consumers connect with brands through more channels and devices than ever before.
So what are some best-practices now? What do you need to know to get marketing ROI right?
Top-Down Approach: Marketing Mix Modeling
If you're more of a statistician, marketing mix modeling (MMM) may be at the top of your mind. It is often referred to as a top-down approach, and it's great for providing a holistic view of all marketing investment, including both digital and nondigital marketing activities and sales. Nonmedia factors such as promotions, competitor activities and economic conditions can also be included in the model.
However, in statistical modeling, it seems a lot of the models can get the job done; and, when you talk to different data scientists or vendors, their approaches can be quite different. Which one to choose? It takes more than a PhD student to do it right, because MMM is not a straightforward linear regression model. The model needs to take into account the uniqueness of marketing problems, such as integrated marketing synergy, adstock, lag effects, and diminishing returns in consumer response.
What are the pros and cons of MMM? Many marketers use MMM for high-level budget planning, and it is often done with little granularity and updated only once in a while, which means it is not enough for optimizing marketing execution.
However, MMM is extremely helpful when you don't have user-level data that is required for a bottom-up approach, such as when the majority of your marketing activities are offline (such as TV, print, outdoor billboard, events) and branding-based; in addition, you can create a what-if scenario simulator that allows marketing managers to better understand the impact of any changes they are planning to make.
Bottom-Up Approach: Attribution
Attribution drills down to the most granular level—e.g., a placement, an ad size, a creative, a keyword, a tweet—at the user level.
If you are more of a Web analytics person, you are probably an expert of the different attribution models—last-touch, first-touch, even-weighting, time decay, and so on—because most of the Web analytics vendors, such as Adobe Analytics and Google Analytics, have free attribution solutions built in their tools (and if you don't have a Web analytics vendor, you're in trouble).
If you have historically been a digitally heavy marketing organization, regardless of whether your organization has MMM or algorithmic attribution in place, you should set up rule-based attribution, because it can be easily used with other data within the analytics tool for all kinds of other analysis. You can slice and dice the data however you want, and share the results across the entire organization in nearly real-time. That's especially important in the types of big organizations where many marketing managers are analytics-savvy and optimize their own campaign execution regularly. In addition, if the goal of your offline campaigns is to drive audience to your website, Web analytics vendors like Adobe Analytics and Google Analytics have them covered too (and, yes, I am talking about vanity URLs).
How to use the different rule-based attribution models? Last-touch attribution is the default view, and it gives you which marketing channels drive sales and conversions. It usually makes search teams look really good, but even search teams should look beyond last-touch attribution to make sure their content strategy and keyword planning covers all stages of customers' decision journey.
If you're a social media manager, you should absolutely explore other attribution models because social media are usually where upper-funnel activities happen. Similarly, marketers who are focused on demand generation, for example, may find first-touch attribution useful because it highlights the campaigns that first introduced a customer to your brand. Even-weighing, time decay, position-based, and custom attribution models give you more options on allocating different weights to more touch points throughout the customer journey. For example, you can assign significant credit to higher-value campaigns.
Here are a few tips that you may not know. Last-touch attribution, for example, means tracing the orders that happen not only within the same visit but also within 30 days of the last interaction. If you're not customizing this cookie expiration period, you're missing out. In addition, big organizations usually have internal campaigns. For example, a brand has a corporate website and multiple online stores; a Black Friday sales banner is placed on the homepage of the corporate site in order to bring people to one of the online stores to shop. If you don't set up the internal and external campaigns correctly (for example, use the same eVar (in Adobe Analytics) to track external as well as internal campaigns), the orders generated by the external campaigns could be overridden by the internal campaigns. And if you want to make the tracking code creation and validation process more automated, vendors like Tracking First can help.
Advanced Algorithmic Attribution
Although many companies are having trouble setting up rule-based attribution models, many argue that it is still not completely data-driven. Therefore, companies that are above the curve simultaneously build advanced algorithmic attribution. This approach takes the guesswork out and lets the data show you the right answer.
Both MMM and algorithmic attribution are sophisticated models. But different from marketing mix modeling, algorithmic attribution also requires stitching together each user's cross-channel and cross-device journey. Some companies, like Facebook, Google, and telecom companies, have unique advantages in doing so. If you're one of them, attribution is the way to go. If you're not one of them, many probabilistic and deterministic solutions are out there to solve this problem for you for omni-channel attribution or personalization, such as Liveramp, Tapad, Drawbridge, and Adobe Marketing Cloud Device Co-op. Addressable TV also makes it easier.
However, to identify that the same person who saw your brand's TV commercial last week and then saw your brand's outdoor billboard yesterday... is no easy task. If you can't or you don't want to invest in ID stitching, you can start with marketing mix modeling, or combine attribution and MMM (a unified approach) to get more comprehensive insights.
Ad Fraud and Attribution Fraud
If you are gaining extremely low ROI for your programmatic ad buy, one thing you should think about is ad fraud, which has been a real—and expensive issue—especially in programmatic advertising.
According to the Internet Advertising Bureau (IAB) ad fraud has accounted for a large proportion of the digital ecosystem. How bad it is depends on where and how you buy, and also varies by country/region. A few vendors, such as Integral Ad Science and DoubleVerify, are specialized in measuring digital media quality and can eliminate ad fraud before your media bid.
It may be hard to believe that today there are still companies that are not measuring their marketing ROI and only looking at CPM (cost per thousand impressions). Most likely they have no idea what's going on and continue to pay for the fraud. However, the bad players have already started to game the attribution models (AKA attribution fraud): for example, allocating most of the budget to unnecessary retargeting campaigns to steal all the credit from the prospecting campaigns. Simple attribution models, especially impression-only and last-touch, are easier to game.
* * *
Many companies have something in place to measure marketing ROI, but few are doing it right. Are you one of them?
You may like these other MarketingProfs articles related to Metrics & ROI:
- Google Analytics 4 Is Almost Here—It's Time to Test and Prepare
- Measuring the Immeasurable: Customer Loyalty Metrics
- B2B E-Commerce: Six Common Return-on-Ad-Spend Measurement Mistakes
- Why Your Customer Experience Metrics Are Lying to You
- Six KPIs Marketers Should Be Tracking [Infographic]
- The History and Future of Web Analytics [Infographic]