Data-driven marketing decisions: Improve the quality and speed of decision-making

“Why do we bother with marketing reporting?”

That’s the basic question I ask marketing leaders when giving talks on marketing measurement strategy. Marketers do reporting to see how they’re doing against their goals and identify the best channels for investment. And if you’re an agency, you’re probably building dashboards and spreadsheets because your clients expect them.

While all of these are true, I’d argue that the ultimate goal of marketing reporting and marketing analytics should only be one thing—to make better decisions.

Better decisions lead to growth, better results, and, most importantly, happier customers.

In this article, I’ll discuss the art and science of data-driven marketing decision-making, including:

• Four types of marketing decisions
• Factors impacting marketing decisions
• Tools for effective data-driven decision-making

4 types of marketing decisions

Marketers make many different decisions, but we can classify them into four buckets:

• Optimization
• Tactical
• Operational
• Strategic

Optimization decisions: Improve campaign efficiency

In these cases, we’re dealing with relatively infrequent decisions that don’t require a lot of data to make them. They’re small changes that drive big impacts. For example, you can boost your conversion rate by changing the color of a button or using the results of your A/B tests to inform your creative strategy.

Any efficient brand is going to be doing brand tests and optimization activities. Even though you’re not going to run these kinds of experiments every single day, they periodically uncover growth opportunities and help you figure out where your message resonates and how you can connect to and convert your audience.

Tactical decisions: Take immediate marketing actions

You need to make tactical decisions on a relatively frequent basis. They’re time-sensitive and need quick execution but generally don’t require too much data to make. 

For example, budget pacing for a paid media campaign. You want to check your campaigns day over day to make sure you’re not blowing through the budget while also ensuring you’re using enough of your allocated budget. Additionally, you need to monitor reach and frequency to ensure you’re reaching enough people without overloading them with the same message. 

Since you’re doing these things daily, you probably want to be doing these things fairly autonomously as well.

Operational decisions: Navigate sudden changes

Operational decisions focus on ensuring marketing activities run smoothly and efficiently but will require large amounts of data. Let’s take outlier detection as an example. If your conversion rate crashes overnight, you may have some problems with your website. Maybe there’s a form fill that’s not going through. Maybe there’s a competitor that launched a product that is at a significantly lower price, and your offering is no longer competitive. 

You need a large amount of data to determine whether that outlier is statistically significant based on your historical data.

Strategic decisions: Shape long-term marketing direction

Strategic decisions set the long-term direction for the business. For example, budget planning and demand forecasting. Making decisions on cross-channel budget reallocations will require methods like attribution modeling or marketing mix modeling, which generally require large amounts of historical data. Given the complexity of working with these large data sets, you’ll need support from your BI team or data team to make these decisions. However, this is often worth the investment due to the high impact of these decisions. 

Factors impacting marketing decisions

Marketing often moves faster than the data infrastructure that supports it. By classifying your decisions, you can select the best tool for the job.

Marketing often moves faster than the data infrastructure that supports it

Evan Kaeding, Lead Solutions Engineer, Supermetrics

When evaluating marketing decisions, you should consider:

• Frequency of decision-making: this can range from hourly to yearly. For example, budget pacing is something you want to do daily, but you generally don’t adjust your demand forecast on a daily basis. 

• Data volume required: some decisions require an amount of data that fits nicely into a spreadsheet or data visualization tool. While other decisions demand a larger amount of data that requires a data warehouse or data lake.

For example, if you need to know how your single-channel campaign performed yesterday, you can easily log in to the platform and get that number right away. You don’t need any tools at all for that. On the other side of the spectrum, if you’re trying to identify messages that have resonated with different audiences across platforms over a long time range, that’s going to require a larger amount of data.

Tools for effective data-driven marketing decisions

What I’ve seen from working with thousands of analytics-matured companies is that to make effective data-driven decisions, you need to have tools that can accelerate speed to insight for low-data decisions—optimization and tactical decisions—and big data decisions —operational and strategic decisions.

At Supermetrics, we refer to these as:

• On-demand data: data that is quickly available to marketers whenever it’s required
• Centralized data: data that is available through centralized storage and follows a company’s data governance guidelines

The way data is stored and distributed has a significant impact on your decision-making.

On-demand data for optimization and tactical decisions

With both optimization and tactical decisions, you want to do these relatively flexibly without necessarily going through the struggle of loading this data into centralized data storage, pulling it through the modeling layer, building governance, and creating dashboards to visualize it.

This is where you need tools that are optimized for on-demand data. Data is fed straight from the source into a spreadsheet or a data visualization tool where you can build reports that help answer tactical and ad-hoc questions.

Giving the marketing team direct access to the data they need will:

• Allow rapid tactical experimentation on existing and new advertising channels.
• Give ownership of the marketing data back to the marketing teams that are generating it.
• Reduce the marketing team’s dependence on a central BI/Data team for time-critical insights.
• Allow timely, ad-hoc analysis in a way that’s hard to design for in centralized systems.

Centralized data for operational and strategic decisions

On the other hand, when it comes to operational and strategic decisions like trend analysis or budget forecasting, you’ll need more historical data and support from BI and data teams.

With a centralized data infrastructure, data is typically stored in a centralized repository, like a data warehouse or data lake, and managed by a data team.

The benefits of this infrastructure include:

• Having a single source of truth by consolidating data from multiple sources.
• Ensuring full ownership and control of historical data rather than relying on external platforms for data storage.
• Granting your data team access to analytics to run more complex and advanced data models such as marketing attribution or marketing mix modeling.
• Implement data governance to maintain data quality.

Get the best of both worlds

Most companies often start with a marketing data architecture that is optimized for either on-demand or centralized data access. The trend we’re seeing is that companies with high data fluency who are far along in their data journey reap significant benefits from leveraging both models simultaneously. 

By doing so, marketing teams aren’t waiting on a data team with a million things packed into their weekly sprints. And data teams aren’t drowning in ad-hoc requests submitted by their marketing teams and their rapidly changing requirements.

By combining the best of both worlds, marketing leaders can give the marketing team the flexibility they crave and the data team the power they need to run complex analyses.