81% of marketers are looking to increase budgets for data-driven marketing, while 83% of marketers believe it’s important to be able to make data-guided decisions.

That’s nearly every marketer.

Over half expect to see revenue growth as a result of data-driven marketing investments (only 7% expect a decrease), and 39% plan to increase spend on data-driven marketing initiatives.

Despite the popularity of data-driven tactics and tech, there is a lot of confusion about the nature of marketing data, and the possible implications it can have for marketing decisions.

Here are five ideas I’ve identified as fundamental to effective data-driven marketing.

1. There are two types of marketing ‘data’: Contact information and performance metrics.

It is strange how rarely this distinction is highlighted, despite how frequently marketers interact with both types of data.

For example, “database marketing” refers exclusively to leveraging email lists to engage with customers.

“Data management platform” (DMP) refers exclusively to leveraging a mix of IP addresses, emails and other contact information to deliver targeted advertising to customers across the web.

Part of the reason I’m writing this article is because of how confused I was by so many of these terms.

Databases can store many things, so why only point to use cases that involve emails? And DMP – such a broad, sweeping term for such a narrow use case.

Understanding the difference between the two methods is critical to knowing how and when to use them.

Many modern marketers are focused so heavily on contact-information-based use cases that they neglect the importance of measuring overall performance and tying it to revenue.

2. Data-driven marketing based on contact information involves tracking individuals in order to get them to buy.

It’s sort of like helicopter parenting – helicopter advertising.

Tracking individuals across digital media is becoming increasingly popular among marketers, in an attempt to make their marketing distinct among the hundreds of ads we each see every day.

“Marketing automation” is one of the earliest examples of this type of data-driven marketing, though people rarely classify it as such.

Marketers use a MAP (Marketing Automation Platform – a technology class created by Marketo, much like Siebel Systems pioneered CRM) to track individuals through the funnel, usually in organic (email, website) touchpoints, but also in paid social and display ads.

Then the MAP distributes targeted content based on where a person is in the funnel, particularly what last action they took (such as downloading content).

Another example is “Attribution” – a term which on the surface just refers to assigning various media, creative and audience segments with percentages of contribution to success.

Attribution usually involves tracking individual customers across the web, based on interactions with digital media properties, both paid and owned.

More advanced platforms can track whether display ads come into view on the visible portion of a person’s screen.

Attribution sometimes leverages DMPs and TMS (Tag Management Software), along with proprietary analytics, to attribute conversions and sales to specific ads.

Attribution is primarily focused on individual identification, but also relies on aggregate performance metrics to recommend adjustments in advertising tactics.

Data-driven methods that rely on contact information benefit from high levels of detail, but suffer from scalability (you can’t track everyone).

Personalization, customer experience optimization and other initiatives fall into this category.

3. Data-driven marketing based on performance metrics involves analyzing investments in and returns from marketing initiatives, in order to better drive results.

Performance marketing, quantitative marketing, media mix modeling – these practices involved rolled up streams of customer actions – i.e. performance metrics.

  • “Performance marketing” is just a way of saying “marketing where you actually look at what works and what doesn’t in order to drive outcomes.”
  • “Quantitative marketing” has historically referred to enterprise-grade initiatives that use statistics to optimize marketing outcomes based on investment (i.e. ad spend) and key performance indicators.
  • “Media mix modeling” – or media mix allocation – is sometimes classed as a type of quantitative marketing. It involves analyzing which channels (TV, radio, display, etc.) have the greatest impact on conversions through probabilistic statistics (data science).

Data-driven methods that rely on performance metrics benefit from the fact that every system produces some measure of reporting and thus are highly scalable, but sometimes more detail is needed.

A focus on performance metrics also has a natural bias towards objectives.

Customer data can be a black hole of possible activities, and many companies are stuck in the vortex of collecting, cleansing, weighing customer data.

But performance data always tells a story relevant to your objective.

4. There’s this thing called a “proxy” – it’s how you measure the intangibles.

Some things are difficult to measure.

For example, “Awareness” (which is an actual objective for many marketers) is really the sum of “how much time do people spend thinking about your brand/product/service?”

Obviously, we can’t strap brain sensors to everyone (yet). Most marketers use focus groups and surveys to approximate awareness.

Those may also be used to approximate “Brand Equity” – the amount people like and therefore demonstrate purchase intent towards a brand.

The problem is, surveys of all kinds are riddled with response biases – including the most subtle, such as familiarity bias and availability bias.

Why not supplement these traditional approaches to awareness measurement with more comprehensive, scalable (and faster, less expensive) tactics?

One increasingly popular method of measuring awareness via proxy is social listening.

This technology category is actually not focused only on social media, but the ‘social web’ – all those publications, blogs, articles, comments, social networks, forums etc… essentially most of the Internet.

“Best-in-class” estimates for the number of websites scanned by a social listening tool are usually around 100m or 200m websites.

For reference, Twitter counts as one website.

Digitally-transformed enterprises use these tools to monitor the success of paid and earned media initiatives, as well as establish competitive benchmarks for brand awareness, based on the assumption that some number of people who are aware of a brand/product/campaign will talk about it.

Other examples of proxies include NPS (Net Promoter Score) – a measure of the social equity a brand has with its customers.

5. You can’t quantify poetry

Marketing – getting people to invest time/attention/money in brands, products & services – will always live partially in the poetic.

Connecting with customers inherently has one foot in the abstract, ethereal, creative – and one foot in the scientific, mathematic, quantifiable.

After all, some of the best ideas are ones no one has thought of before.

This is why Jay Baer proclaims that “Data-Driven Marketing is a Bad Idea” – all he’s really saying is that you can’t forsake the creative for the quantified, but the title he actually used is more sensational.

We need a more holistic view of marketing data

What marketers need is to broaden the scope of the way data is viewed, valued and used within their organization.

Personalization and attribution have their place among the most academic of statistical approaches.

And macro-performance measurement must concede the fact that, sometimes, the devil is in the details.

Of utmost importance is the fact that no data is valuable unless it connects to critical objectives. For most marketers, this means awareness, brand equity and revenue.

For more on this topic, see: