Big data, in and of itself, is nothing special. Everyone and anyone can collect and aggregate data. But that doesn’t mean it’s providing value. The challenge is twofold: first, ensuring sufficient data integrity and second, successfully gleaning the insights to fuel smarter marketing.
The truth is it’s harder than ever to keep track of the modern consumer. Brands need to manage a seemingly endless number of touchpoints across different devices and marketing channels, from digital and social content to real-world interactions and purchases. It can be daunting to stay apace of where consumers are interacting with you, much less understand what drives and influences their behavior. For example, if you’re using a digital attribution model, you might not be able to account for the impact of traditional media. And if you can’t tell if that TV ad influenced a customer to visit your website, for instance, how can you have a true picture of performance and return on ad spend?
Then there’s the issue of data quality. “Garbage-in, garbage-out” applies here. If you’re putting bad information into your system, then no matter how precise your data model, you’re getting bad analytics in return.
You can see how data projects can falter. Consider some of these common reasons consumer behavior gets lost in data translation:
Multiple systems contributing multiple log files
It’s not unusual for marketing departments to pull data from disparate cookie-based tracking systems and point solutions. One tracks website performance, another logs social interactions, and mobile, while other data comes in from point of sale systems and loyalty cards. Each output has different sets of attributes, at different cadences, ingested in different ways. The effort to match all these IDs can be herculean and nearly impossible. As a result, marketers are double, triple, and quadruple counting touchpoints, leading to misappropriation of data and misleading or incoherent insights. Offline conversions with no touchpoints, online conversions with no clicks, clicks with no impressions: It’s virtually meaningless and will ultimately misallocate marketing credit.
Ingesting deceptive data
How trustworthy is the data you’re putting into your system? Consider the issue of location data. For in-app impressions, app publishers may pass GPS location data (in the form of latitude and longitude) in the bid request sent to an ad exchange. However, what happens if the user blocks that app publisher from access to the real GPS? The publisher may use a proxy, such as a zip code associated with a device owner or a location associated with an IP address—both notoriously inaccurate. Or the publisher might even fabricate the data, injecting a random or arbitrary GPS value into the bid request. Either way, fraudulent, inaccurate or stale data can create incorrect results.
Using insufficient tracking technology
Are you getting an accurate picture of how your users are moving from device to device or app to app? For example, connecting mobile app to mobile web environments is a critical part of understanding the consumer path to purchase. People who click on in-app ads are frequently sent to the browser for the brand’s website. This transition changes that phone’s external identifier from a device ID to a cookie. Incorrectly tracking transitions from app to web and back again, at scale, is one of the most common ways that data becomes garbage-y.
To overcome these challenges, an increasing number of companies are now looking to adopt a unified measurement approach, what we call a “single system of record” that relies on a strong marketing tech stack that tracks users across their devices, is scalable and utilizes machine learning to translate the vast amounts of disparate data into actionable conclusions.
A recent Forrester Wave study on marketing measurement and optimization solutions pointed out this growing demand for a unified measurement approach as consumers straddle digital and traditional media. This evolution is increasingly critical for advertisers and agencies seeking to understand ROI and optimize spend.
Still, the quality of attribution is only as high as the quality of the data inputs and the modeling process. As people move between online and offline, from location to location and from device to device, having the right data is more critical than ever. Having a system that can marry different data types, different taxonomies and different cadences while maintaining consistency is fundamental to making that happen.
