Marketing Performance Measurement Is (Still) Broken, and It's a Data Problem

In short: most companies can't measure marketing performance across channels, because every channel has its own metrics, data model and team. This is a data problem. You fix it by bringing all marketing data into one place and building a shared "metric layer" on top, using modern data stack practices.

This article was first published in January 2023. The problem it describes hasn't gone away, and we've lightly refreshed the text.

Phone screen full of social media app icons

Customer-facing organisations in every industry now use many channels to reach their customers. It has become a necessity. The norm is to run marketing across a long list of channels, from paid to organic social media and more. Companies run paid campaigns on Facebook, LinkedIn, X (Twitter), TikTok, Google Ads, Bing Ads, YouTube, Pinterest, Snapchat and others.

That's on top of email campaigns, messaging and chat promotions, and a whole host of organic social media activity. Over time, we'll see even more channels, because of the way the industry works.

The real issue, beyond how scattered marketing activity has become, is measurement. I'm yet to see an organisation that has a good handle on its marketing measurement across channels.

Each channel produces its own metrics, APIs and data models. The differences get worse because companies often have separate teams for different channels, and those teams don't always talk to each other. They have different target metrics and even different terms. For example, what counts as a "conversion" can differ from channel to channel, and even from campaign to campaign. That leads to the core problem.

Marketing performance measurement is broken, and it's fundamentally a data problem

A lot of companies are missing the bus when it comes to a good data strategy for marketing measurement. This is how most organisations handle their marketing data today:

1. They use several separate tools to pull data from different social and digital channels. Smaller groups within the marketing team often set these up on their own. Many of these tools claim to offer one consistent data model across channels, but peel back the layers and you'll find a lot of missing pieces.
2. The raw tables and views from these tools are then dumped into a data warehouse.
3. Without much thought for the big picture, the data and analytics team gets busy building reports for business stakeholders, without first bringing the data together properly.
4. This is also the point where the marketing analyst gives up and goes back to three hours a day of Excel VLOOKUPs and pivot tables to build reports.

One reason a consistent data strategy can feel like a waste of time is that the tools used for marketing measurement offer only limited ways to extract and report on data. That makes building a unified data strategy seem pointless.

The real cost of scattered marketing metrics

The impact goes well beyond the extra time analysts spend pulling reports:

• Wasted spend: without an overall view of how efficiently money is spent across channels, budgets leak.
• Missed trends: it's hard to spot patterns across channels and teams, such as emerging products, problems with campaign targeting, or customer segments that bring outsized returns.
• Unclear attribution: even basic marketing attribution becomes confusing, because data and connections are missing.

The case for using the modern data stack for marketing measurement

The way to solve this is to treat marketing measurement like any other scattered-data problem. Much of it can be solved with the principles of the modern data stack: modern ELT tools, CI/CD, version control and easy data discovery for users. And, most importantly, a unified metric layer for marketing.

There are patterns in marketing data you can use to build a solid foundation, often called a semantic layer, that takes much of the pain out of measurement. Here are a few examples:

1. Shared concepts: ad platforms share the same basic structure: accounts, campaign groups, campaigns, ad groups or ad sets, ads and creatives. Each platform has its own variation of this.
2. Platform-specific concepts: some details exist on some platforms but not others. For example, breakdowns by device, age and gender are available on Facebook and Google Ads, but not on LinkedIn.
3. Search ads have their own shared ideas: keywords, bidding strategy, campaign and ad type, and so on.
4. Overlapping metrics: all these platforms use similar metrics: impressions, clicks, conversions, engagements, cost, CPC, CPM and more.
5. Web and app analytics: tools like Google Analytics, Adobe Analytics and Mixpanel give you overall performance across every traffic source.

With these five ideas, you can already build a solid set of views for the first version of your marketing data model. For paid social, you could combine data at the ad level, across all channels, at a set date granularity. You'd then add views that combine demographics, conversions, campaign results against forecasts, and more.

The main benefit is that your organisation gets full visibility across all its channels, and full control over its data model, shaped around how it runs its business, with its own definitions and its own cuts of the data.

FAQ

Why is cross-channel marketing measurement so hard? Every ad platform and analytics tool has its own metrics, definitions and data structure, and different teams often own different channels. Without one shared data model, the numbers don't line up.

What is a semantic layer for marketing data? It's a shared set of definitions, like "campaign", "conversion" or "cost per click", applied the same way to data from every channel. Everyone then reports on the same numbers.

Do we need expensive tools to fix this?