Don’t Forget the Art Behind Digital Analytics

marcrossenddtupdated“Data-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Marc Rossen, executive director of digital insights solutions at MarketShare.

When most of us think about digital analytics, we tend to think about the science. Get the science, the math and the modeling down, the thinking goes, and you’ll know exactly how to distribute your marketing spend and generate incredible ROI. All you need is the right algorithm and you’re set.

As someone who’s committed his career to the marketing sciences, I’ll be the first to say that the science is absolutely critical. But the funny thing about science is that, often, the more complex it becomes, the less cut and dry your answers will be. There’s an art to sophisticated science.

That’s unquestionably the case in marketing analytics. Analytics is a tool for delivering marketing results to individual businesses, which means that effective digital analytics come from fitting the science to unique business needs. That requires understanding the individual concerns, questions and operations that drive each marketing campaign, and shifting the models accordingly.

The best results don’t come from models that are created in a vacuum; they come from science that is heavily driven by the art.

As The Digital Program Changes, Analytics Change

For one prime example of the art of data, consider the following digital truth: Not all businesses are at the same stage of digital investment. Some businesses are fully invested in digital; others are just starting to test the waters online. That distinction can drive enormous differences in the kinds of questions analytics will need to ask.

Digitally committed brands will want to focus analytics on driving efficiencies across existing investments, such as optimizing in-market campaigns to generate incremental lift. Digital newcomers, on the other hand, will want to focus on proof points of what’s worked and where the strength of digital really lies. And because newcomers may face more senior-level digital skeptics, they also may want to put extra focus on strategic-level reporting to convey digital value to the C-suite.

Questions For Different Stages Along The Funnel

It shouldn’t come as a surprise that top-funnel branding should drive different questions than lower-funnel marketing. Obviously, those differences spill over into analytics as well.

Lower-funnel marketing is likely to focus on near-term conversions, often on an ecommerce site. That means lower-funnel analytics might concentrate heavily on online data – and will likely work within shorter digital attribution windows. By contrast, upper-funnel marketing often has no online conversion event, and can involve many months of branding efforts. That means upper-funnel analytics might be more truly cross-channel, and is likely to work on far longer time horizons of months, rather than weeks. In short, different marketing focus will lead to different questions.

The Broader Environment

No marketing exists independently. Big sponsorships, brand halo effects, the economy, gas prices, the weather and competitors’ advertising and pricing can strongly impact consumer reactions to marketing. That’s why the best models can’t just focus on the immediate impact of a single tactic or campaign. They need to take the full picture into account.

Different Executions, Different Data

Different types of executions call for different kinds of data as well. Effective programmatic buys require highly granular, up-to-date data. Manual buys can rely on more generalized data and don’t need the constant flow of insights. Email marketing, meanwhile, will likely need to incorporate offline CRM so marketers know they’re sending the right message to the right people. Different kinds of executions require different data inputs to succeed.

Align The Teams

Companies can’t build effective models without understanding which team has what data. And so, for the models to perform well, the teams need to be aligned.

Take creative optimization. To optimize creative effectively, brands need to understand the ROI on each message, which means comparing each unit’s performance, compared with that unit’s spend. If the creative or media agency only has high-level spend data, on a quarterly level, for example, then the agencies won’t be helpful at capturing ROI and it can’t be an effective resource for creative optimization. Even the best model in the world won’t change that.

The common theme here is that strong data and models aren’t enough. To get the most out of marketing analytics, brands need to understand what their business wants to achieve, the campaigns they want to execute and the questions they want to answer. Once they have that starting point, they can gather data and craft their models around those parameters.

The alternative is potentially working with a misguided model, a black box or both. It’s the analytics equivalent of buying the best car and using it to drive in exactly the wrong direction. That, to me, seems like bad art and bad analytics.

My advice? For the best analytic science, start with the art.

Follow Marc Rossen (@marcprossen), MarketShare (@MarketShareCo) and AdExchanger (@adexchanger) on Twitter.

 

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>