You don’t have a “data problem”
Five different marketing data issues companies face—from data quality to data activation—and how to tackle them.
First things first, forget about thinking of data as a boolean
Often, leads approach me to fix what they call a “data problem”. We marketers sometimes think of data as a boolean: either we have it, or we don’t. However, data problems come in many shapes and forms. And similarly, they need distinct solutions to be fixed. A data quality issue is solved very differently from a data accessibility problem.
In this article, I break down “data problems” into five distinct categories: measurement, quality, accessibility, literacy, and activation. I cover the symptoms of each and possible remedies.
I use this segmentation regularly with customers. It helps me understand if I’m the right person for the project or if they’re better off speaking to a data engineer or implementation specialist. I expect it can help you similarly and clarify what resources you need to actually get value from data.
Digging into the five unique “data problems”
Data Measurement
Common Symptom: data as an afterthought.
The data is not just there, and tracking is an afterthought. When marketers want to analyse a certain interaction, they realise it wasn’t being tracked even though the “new feature was shipped weeks ago”.
The problem can also be more minor. Some of it is tracked but there are still gaps. The gaps are closed ad-hoc so the team is never quite sure what tracking exists. It often means every investigation begins by Slacking the data team.
Possible Solution: engage marketing with engineering.
I haven’t seen this issue occur because the team didn’t have the right tooling. I’ve never had a client who didn’t have at the very least Google Analytics set up and most tech companies will also have the likes of Amplitude or Mixpanel implemented. This means it’s not a tooling issue, but a workflow issue.
A possible solution is to have guidelines for engineering on how to track new features. But from my experience, engineers lack crucial context on how the data will be used. So they don’t know exactly what should be tracked.
What’s the hypothesis of this feature? How should the data be segmented? With what data sets will it need to be joined with? Sometimes missing a certain variable can make the data 90% useless. Marketers are the ones who know the requirements of the data.
To ensure that data provides business value, involve those who rely on data (marketing analysts, PMs) in the measurement process.
Therefore, the most efficient way to ensure data is measured in a way that provides value is to involve the people who will use the data (marketing analysts, PMs) in the measurement process.
Data Quality
Common Symptom: stakeholders don’t trust the data.
You have data—maybe too much data—but people don’t trust it. When an analysis returns a result “marketers did not like,” the data team has to double-check events and pipelines.
Metrics don’t match, and new users mean different things for marketing and finance. Dashboards break, and because there’s no quality alerting, business users are the ones who notice it first. Yikes.
Possible Solution: move to a data engineering workflow.
Data quality issues need to be tackled by the team data team. It’s usually a combination of workflow and tooling. Tools like dbt can help, with engineering best practices like CI/CD, modular development and alerting. Migrating to a data engineering workflow is a big initiative. Still, it is a worthwhile endeavour if you want to use data to make decisions or to leverage it to improve campaigns, decrease churn or increase upselling opportunities. So, basically, if you plan on using data at all.
Data Accessibility
Common Symptom: marketers can only access data via UI tools or dashboards.
Some teams try to fix data quality by closing off data access. For example, by only enabling marketers to access data via data team-produced dashboards. If all data production is owned by a data team, then the chances of broken pipelines, metric discrepancies, and the like decrease. For sure, it does. Sharply.
But when domain experts don’t have full access to the data, the value you get from data also decreases. Sharply. Marketers need to be able to go down their own rabbit holes to get insights. They’re the ones that know the paths to take.
Possible Solution: enable marketers to participate in data transformation.
Similarly to the data measurement issue, a solution for this lies in workflow and better integration between marketing and data + engineering teams. Give marketers a sandbox of their own in your data warehouse, where they can write their own queries.
Data Literacy
Common Symptom: stakeholders don’t know how to use the data.
You’re tracking everything that’s valuable, there’s good governance in place, everyone trusts the data, AND marketers can easily access it. Hurrah!
But do marketers know how to model the data, interpret it, and take action based on it? Or does everything fall on the data team instead?
Domain experts are the best fit for uncovering insights and leveraging data for business results. But they need the right context and skillset to do that.
Possible solution: produce documentation and teach SQL.
Documentation on data lineage and transformations can be a big push to enable marketers. However, your team may not have the right skill set to leverage this.
Teaching SQL can go a long way, even if “everyone can use ChatGPT”. It’s not just a coding language, but a crash course on how to think of data, data sources and metric calculation. Airbnb had a successful data education program a few years ago which can serve as an inspiration. Also, you’d be surprised by how many marketers can’t even do a VLOOKUP.
Data Activation
Common Symptom: the data is only used for reporting or analysis.
We all love insights, but data provides value beyond learning. Data can be leveraged to customise CRM campaigns, build better target audiences for performance and predict (and avoid) customer churn. An effective data-driven organisation can leverage the data that’s used for reporting for all these additional (activation) use cases.
Possible solution: invest in marketing operations and perhaps a RETL.
To reach this stage, you should already have data measurement and data quality under control or things will snowball very quickly. Data activation usually lies within the realm of marketing operations / martech. A dedicated in-house or freelance stakeholder can help you settle the foundation. A Reverse ETL tool— like Segment or Hightouch—can also make the process centralised and simpler.
Time to fix your “data problems”?
I can’t help with all these problems, but I can, at the very least, assist you in identifying the symptoms and solutions for your organization. Head to my website and reach out if you want to talk about projects.
And don’t forget to subscribe to 021. I write about everything connecting the worlds of marketing and data in this newsletter: