Customer facing organizations across all industries today make use of multiple channels to reach out to their customers - it has become a necessity. The norm today is to run marketing across a plethora of channels, from paid to organic social media & other channels. For example, you’d see companies run paid campaigns on Facebook, LinkedIn, Twitter, TikTok, Google Ads, Bing Ads, YouTube, Pintrest, Snapchat and more.
This is apart from email campaigns, messenger / chat related promotional activities and the whole host of organic social media activities as well. As time goes on, we will see more and more channel fragmentation due to the nature of the industry.
The real issue here, apart from the amount of fragmentation of marketing activities, is the problem of measurement. I’m yet to see organizations get a good handle of their marketing measurement across channels.
Each channel produces its own set of metrics, APIs and their own custom data models. The differences in the data are accentuated by the fact that companies have completely different teams for handling different channels, and often these teams aren’t really talking to each other. There would be different target metrics, and even different terminology (for example a definition of conversion could differ from channel to channel & even from campaign to campaign). This, then leads to the primary problem:
Marketing performance measurement is broken - and it is fundamentally a data problem.
I feel like a lot of companies are missing the bus when it comes to building a good data strategy for marketing measurement. The approach that most organizations take in managing their marketing data is as follows:
Use multiple point solutions that help extract data from different social & digital channels. These would be setup independently by smaller groups within the marketing team as well. Many of these point solutions claim to provide a single, consistent data model across channels, but if you peel the layers and dig into the data, you’ll notice a lot of missing pieces.
These raw tables & views from these solutions are then often dumped into a data warehouse.
At this point, without much consideration to the big picture, the data & analytics team already gets busy in creating reports to satisfy business stakeholders without really going into a data consolidation exercise.
This is also the point where the marketing analyst throws in the towel and gets back to 3 hours of excel vlookups & pivots a day to build his / her reports.
One of the reasons why having a consistent data strategy feels like a waste of time is also because the entire infrastructure used for marketing measurement provides limited ways in extracting & reporting on data, which makes the whole exercise of developing a unified data strategy pointless.
The real impact of fragmentation of marketing metrics
The real impact of fragmentation goes beyond the additional time analysts spend on pulling out reports. It has more far-reaching consequences, such as significant cost leakages due to a lack of an overall view of spend efficiency across channels. Fragmentation also makes it difficult to spot trends emerging across channels and teams, including identifying emerging products, problems with campaign targeting strategies, and interesting customer segments that provide outsized returns. Furthermore, even basic marketing attribution becomes unclear and confusing because of missing data and connections.
The case for adopting the modern data stack for marketing performance measurement.
The way to solve this problem then, is to treat marketing measurement like other data fragmentation problems today. A significant proportion of these problems can be solved by making use of the principles of the modern data stack: modern ELT tooling, with CI/CD, version control and the ability for users to easily do data discovery. The most important part - a unified metric layer for marketing.
There are patterns in marketing data that can be exploited to build a good, foundational semantic layer that can go a long way in reducing the pain of marketing measurement. Here are a few examples:
Shared concepts: Ad platforms have data concepts that are shared - Accounts, campaign groups, campaigns, ad groups / ad sets, ads and creatives. Depending on which platform is used, they all have some variation of this.
Unique concepts: Then there are other concepts that are shared on some Ad platforms and not shared on others - for example, device / age / gender breakups are available on Facebook & Google Ads but not on LinkedIn
Search Ads also have commonalities - keywords, bidding strategy, campaign / ad type etc.
The metrics that all these platforms use have overlaps - impressions, clicks, conversions, engagements, cost, cpc, cpm etc.
Finally, web and mobile analytics tools like Google Analytics / Adobe Analytics / Mixpanel can be used for overall performance analysis across all sources of traffic
Based on these 5 concepts, we can already go about creating a robust set of views that will create v1 of the marketing data model. For paid social, these could be consolidated at an ad level, across all channels, with a pre-defined date granularity, followed by other types of views that combine demographics, conversions, campaigns vs forecasted and more.
The primary benefit of this approach is that the organization has full visibility across all its channels, and total control over building its data model in the way it runs its business - with custom definitions, and custom cuts of data.
In a subsequent post, we will discuss some principles of creating a robust semantic layer for marketing data, using either open-source or paid tools combined with modern analytics engineering practices.
Did you enjoy this article? Sign up for our newsletter and get articles every week on the practical applications of the modern data stack!
Comments