"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 Andrew Shebbeare, founding partner and global chief strategist at Essence.
I love digital media. I particularly enjoy the geekery, the data, the possibilities. The only trouble is that when faced with this wonderful box of marketing Lego, I sometimes lose track of what I was building.
In an industry as complex as ours it is probably inevitable that we sometimes miss the woods for the trees while we try to connect irreconcilable data sources or argue the nitty-gritty features of DMPs. Here are a few of the pitfalls I’ve encountered. I know I have been guilty of all at one time or another.
1. Not Having A Strategy
In the era of predictive analytics and auto-optimizers, it is easy to forget to ask what we should do and why. The temptation to flick the switch on your programmatic acquisition machine is strong. But don’t confuse a few smart tactics for a strategy.
A media strategy is an articulation of how media will help you achieve a business goal. Any tactics that don’t align need to be set aside. You might have a cool lookalike model bringing in cheap signups, but is that going to propel your business to success? Would you be better off creating brand love that will convert millions of loyal, impassioned users?
2. Thinking You Can Avoid Taking Risks
You can’t, at least not without making huge sacrifices. With the amount of data generated by digital campaigns, it is easy to fall into the trap of thinking that your historical performance will tell you what to do next. Unfortunately, if you want to generate really awesome results, you’re going to need to be much braver. You should use data to shape ideas and understand audiences and how they relate to your brand and products. But you can’t expect data to have ideas for you.
Brands regularly back themselves into a marketing corner by being risk-averse and forgetting how to be creative. They get caught in a “whittling trap” -- refining down to the perfect creative, media plan or landing page in small increments when a radical overhaul is what’s needed. Meanwhile, a new entrant is only too happy to disrupt through innovation, agility and a healthy dollop of gut feel.
I recently met with an advertiser stuck in this loop, unable to grow their business but refusing to accept that their narrow attribution methodology was holding them back. Their flawed reporting had become the unshakeable foundation of their marketing beliefs; to question it was anathema. This cycle is hard to break.
3. Believing You Can Measure Everything
You can’t do this either. Obviously, the data that the digital medium generates is pretty amazing. The large-scale integration of data across multiple screens, place-based actions and offline transactions is very exciting, but we should not mistake big data for perfect measurement. There are some fundamental statistical realities that can’t be overcome.
Until we can read people’s minds in real time, we can’t measure brand effects in real time. Until we have the ability to predict behavior for every individual and measure changes influenced by marketing, we can’t measure direct response in real time either.
Thankfully, there are helpful proxies for behavioral and attitudinal change. You use them all the time. They help us make educated inferences about what works, but you should remember them for what they are. Attribution models are still models. Brand studies are based on samples. Both have statistical error. Even black and white metrics like last-click ROI are really full of noise and shades of grey when you scratch below the surface. You should use all these tools, but keep them in perspective. They can help you optimize, but may not reveal the truth.
A common mistake is to rule something out because it is “unmeasurable.” What people usually mean by this is they don’t have a familiar proxy that is compatible with other proxies and will tell them how much to invest. Once you accept that nothing is totally measurable, having a few different tools in play may not seem so bad. Next time you rule out a marketing opportunity, don’t rule it out because you can’t measure it. Rule it out because it doesn’t feel right.
4. Brushing Bias Under The Carpet
Marketers love testing, but not all really understand it. The truth is that much of what is considered good practice can be flawed.
Media people, for example, talk a lot about “optimal frequency” -- an important input into any planning model. It’s harder to measure than many realize. If you see your lowest CPA after five impressions, is that because five is your optimal frequency, or because people who convert tend to see five impressions? The only way to really understand optimal frequency is to hold out separate audiences at each marginal impression -- a complex and expensive undertaking.
I will enter the confessional myself. A few weeks ago a client pointed out that because my control ad in a recent campaign required a lower version of Flash than the exposed ad, the incidence of backup images served was lower in the control -- introducing bias in the audiences exposed to Flash. Thankfully it was a small bias and one we could isolate, but a bias nonetheless.
Bias is everywhere: in frequency, recency, sequence, time of day, site, audience and screen size. The discipline of experiment design needs to enter the mainstream of marketing. These are hard lessons because they cast doubt on some of the basic “best practices” we have come to accept. It is tempting to say something is good enough, but when you are dealing with small changes in behavior or attitude, it may not be.
5. Losing Sight Of The Human Connection
In the programmatic era, it’s easy to forget that audiences are made up of real people. Sitting behind virtual levers and dials, you might feel like some sort of virtuoso manipulating millions of faceless zombies. Take a reality check and give people some credit; it will pay off.
Ads that have something to say work better. What is interesting to one may not be interesting to all. Instead of boiling everything down to averages, put yourself in users’ shoes, consider their context and how you can add value to that situation, not clutter.
Don’t assume you always know what’s best. You might laugh at your own jokes; don’t expect others to. You might get goosebumps from your new promo spot; that doesn’t mean YouTubers will. Judging from inside the bubble is tough. Your brand affinity, product knowledge and personal circumstances are skewed. If taking an objective view is hard, think about pretesting your work before going to market. You could save millions of dollars and months of time.
6. Forgetting The Sum Of The Parts
I once helped launch a new retail brand from scratch. We started with a blank slate: no traffic, customers or reputation. We activated paid media and started tracking sales. We found that on top of what we tracked, we got an extra 40% or so we couldn’t explain. Word of mouth, cross-device conversions, rejected cookies and repeat purchases were contributing a big chunk of revenue for which we couldn’t directly credit our ads. These network effects and wise investment helped us grow and reach profitability much faster than we would have done otherwise.
Since marketing diagnostic metrics are really just diagnostics, you need to understand their relationship to your business’s bottom line. I can’t promise you 40%, but it’s not zero. This isn’t just about attribution, it’s about what happens outside the model altogether.
This is the one I find hardest personally because I get so excited by technology and innovation. There’s just so much cool stuff to play with that knowing when to stop can be difficult.
In this game, whatever you are implementing right now you will be tearing up pretty soon afterward. No matter how flexible you try and make a system, you’ll be missing the one feature you didn’t think of. People talk about CMOs being bigger technology buyers than CTOs and CIOs, but they also have it harder, with a lot less clarity of the world for which they need to designing.
My advice: Embrace the hacks that get the job done quickly. Be pragmatic. Don’t get obsessed with future proofing. Don’t be afraid to change your mind. Create a lean infrastructure and avoid getting too wedded to anything. Make the best decisions that you can today, focus on what you know. Focus on getting “good” into market rather than waiting for “best.”
I hope this list is at least a little useful. Thankfully these seven sins aren’t actually deadly. If they were I wouldn’t be here warning you about them.