For years marketers have been promised that technologies like programmatic will unlock efficiencies, freeing them up to focus on more creative work. Google’s Michael Bailey explains how his team is making that promise a reality.
Back in 1930, the economist John Maynard Keynes made a bold prediction for the not-so-distant future: thanks to technological advancements and the resulting productivity gains, we’d all be working 15 hours a week. Almost 90 years later, the average American is working more than double those hours.
Does that sound familiar to any marketers? For years now, we’ve been told that marketing automation technologies like programmatic will unlock efficiencies, freeing us up to focus on more creative work. That vision has also yet to materialize.
So a short while back, my colleagues and I at the Google Media Lab — the team that manages the media strategy for Google’s advertising campaigns — laid out and started enacting a plan that we think will make that vision a reality.
1. Find data that predicts results
If there’s one thing that’s important to remember when we talk about an automated future, it’s this: machines are only as good as the data we feed into them.
Machines are only as good as the data we feed into them.
For brand marketers, that can be tough. After all, the goal of brand marketing is to boost people’s perception of your company or product. What real-time data points can possibly help you measure something so nebulous? That’s the exact question we’re constantly seeking to answer, as we deploy creative and media for our own campaigns.
When it comes to our creative, we’ve put in place a structured approach to learning. We take creative from campaigns, develop a bunch of hypotheses, and then produce variations of the ads to isolate key performance variables before testing them in the lab. The team of people who does this isn’t involved in day-to-day campaigns, so we’re completely focused on the task at hand. If we see consistent results we’ll define it as a creative best practice and disseminate the findings across the company.
We’ve taken a similar approach with our media campaigns. For example, a couple of years back, we mapped out all sorts of variables we felt increased the chances of an ad being effective — everything from video completion rate to the presence of audio to viewability. And because we measure our brand campaigns on an impression-by-impression basis, we were able to test how effective each variable was.
It turned out that many of the things we had been optimizing for — video completion, for example — were not predictive of brand lift. What we found was that when an ad was both audible and visible on completion, there was a statistically significant lift in brand awareness. In other words, a data point that predicted the results we were shooting for.
2. Optimize for those data points
But what good is all this data if you’re not optimizing for it? That’s why, rather than relying on out-of-the-box optimization solutions, we’re increasingly using marketing automation tools like Google Marketing Platform’s custom algorithm. Marketers can feed in their proprietary data and the tool then uses machine learning to optimize campaigns against the parameters they’ve set.
The results so far have been promising. For example, in recent Pixel and Google Assistant campaigns, we customized our optimization algorithm based on predictive signals data collected from over 300 previous campaigns and ran it directly against out-of-the-box optimization tools to see which one was more effective. We found that our custom algorithm drove a statistically significant higher lift in brand awareness.
3. Build templates that can scale
The beauty of digital marketing is how easy it is to be contextually relevant. Unlike TV, which requires a one-size-fits-all approach, with digital you can take one creative template — incorporating all the best practices you’ve already identified — and then modify it to match hundreds, if not thousands, of different use cases. These use cases span unexpected moments and contexts — where we dynamically update the ads to reflect location, weather, or even sports scores — or expected moments for which we have planned scenarios built.
The beauty of digital marketing is how easy it is to be contextually relevant.
For example, for the release of the Google Home Mini, we did what we call a moment-mapping exercise. We outlined all the big and small moments where we thought our product could be useful — everything from sporting events like the World Series to holidays like Thanksgiving to small daily moments like making a grocery list. We then tailored our template to each of these cases and used programmatic technology to serve up the right ad to the right person in the right moment. If someone was searching for a turkey recipe, say, we could serve them a contextually relevant ad. Thanks to this automated marketing strategy, we were able to balance dynamic creative and scale, delivering more than 1.5 billion impressions while driving a 6% lift in awareness and a 5% lift in consideration.1
4. Use the time you’ve freed up to think beyond ads
We’ve spoken a lot about the role of algorithms, data, machine learning, and artificial intelligence so far. Is there a part for humans to play in our automated future? Absolutely. In fact, by doing everything we’ve discussed, we’re freeing up people’s time so they can spend it on more interesting creative work — the real custom stuff that can’t be forced into a template.
For the release of the Pixel 2, we partnered with the Guardian on a branded content-meets influencer campaign that generated both buzz and real business results, including a 40% increase in purchase intent and a 76% increase in people describing the device as “prestigious.”
But this type of creative work is only possible if long-standing norms and processes, built for the old way of doing things, are dismantled. That means we’re changing how we staff our teams, adding more creative talent. In fact, in just the past year, our staffing of creative talent in the Google Media Lab and across agency supporting media is up 5X.
We’ve also changed our processes. For the Pixel 2 campaign, for example, we split what is normally one process into two streams, each one with its own timeline and budget. Stream one, which included all the automated template work, got 20% of our time and 90% of our spend. Stream two, which consisted of all the custom elements of the campaign, got 80% of our time and 10% of our spend. We’ve found that by setting up our teams and processes in a way that reflects what we want to achieve, we’re more likely to make it happen.
Preparing for a future of marketing automation
The idea that we might one day work 15 hours a week still seems far-fetched. But by following these four steps, we’re confident we can make the second promise — of a marketing industry that uses automation to unlock new efficiencies, freeing up people’s time to focus on the more creative work — a reality.
