Interview with Mat Hauk of Datalicious.com about Big Data Marketing
A focus at Datalicious is data-driven marketing. This expression seems to imply that marketing decisions are often made with weak supporting data or no data at all. Is this still the case today?
Very much so. Marketers and brands are more aware of the data available to them, and in most cases collect valuable customer data (and have valuable customer data in old offline warehouses); but it’s still very common for these insights pulled from the data to not be used to influence future decision making. Data-driven marketing is specifically that: allowing the data to dictate the direction of your customer interactions. Too often, marketers are caught in the rut of habit, gut feel and intuition.
The complexion of a marketing team has changed. There are more statisticians and analysts that are involved in the planning process. There are also better technologies available to perform what used to be complex and daunting tasks, which allow marketers to make better decisions. But I’d say there’s still a ways to go.
What kinds of data do you think marketing decision-makers often overlook?
There’s probably no single source of overlooked data. The trouble I see is that there is simply so much data and it’s so disparate that most marketers choose to specialize. They elect to acknowledge their web data and ignore their mobile data because that’s hard. Or they know there’s a tremendous wealth of historical perspective in a data warehouse, but it’s too hard to integrate it or the internal politics make it difficult to gain access to it quickly enough to react to a market demand. Breaking those silos and giving yourself a worldview of your customer knowledge is key to data- driven marketing.
To what extent are today’s major marketing platforms designed to support collecting data for what Datalicious calls “data-driven marketing?”
The major marketing platforms aren’t designed for it at all. Most of the major platforms that exist today are built on 20-year old concepts by fast-moving startups. Those startups were all bought and aggregated into what’s being called “stack” solutions which were never well integrated. Ultimately, what we’ve ended up with is half a dozen providers who are reselling technology stacks that were acquired, not built; and then the customer has a lot of silos with the same branding that they still can’t leverage. Smart data-driven marketing needs to come from truly democratized data that has context to the rest of the knowledge of the business, not just context to itself. Next-generation solutions like Datalicious OptimaHub do exactly that. By giving yourself access to view all of your customer data in any context, you can begin to uncover deeper insights, more relevant trends and active segments that you never knew existed.
What needs to be added to major platforms to facilitate data-driven marketing?
There’s a great lack of integration between major platforms. I don’t see building connectors between disparate technologies as a solution. These are all platforms designed to serve specific analytics uses, but they were never made to work together. Yet we still see traditional BI companies buying up these vendors and putting together a hacked-together solution. I’ve yet to see a vendor crack a unified platform that takes care of collection, analysis and actioning. We built the OptimaHub to address this.
Is the problem a lack of data, hard-to-use analytics tools, or the way that marketing campaigns are designed and budgeted?
Part of it is the lack of true integration between analytics tools. This comes from the aforementioned: companies purchasing disparate technologies and forcing them to play nice together, which isn’t really a solution.
Another area would be a lack of analyst involvement in campaign planning. When it’s just marketers left to design data collection processes, it’s usually without the foresight of looking at what insights are actually required, as well as what data is required to generate these insights.
What is “multi-touch attribution” and why should Big Data marketing types study it more intensively?
Multi-touch attribution is where every touchpoint in the purchase path gets assigned a credit. There’s a few different methodologies that can be used (e.g.. semi-partial correlation or logistic regression), but the goal is the same: to gain an accurate view of media channel performance. Custom multi-touch attribution modeling gets marketers to the truth of media performance. Anything else will generally leave marketers with inaccurate data when planning future media spend.
This is in contrast to, say, last-click attribution, which gives all conversion credit to the last campaign touch point in the purchase path and ignores the value of all other digital advertising that contributed to the conversion. Unchallenged for many years, it has led to an outdated and biased view of the digital landscape.
What is your take on video ads, slideshows (e.g., SlideShare) and infographics? Where do these fit in the universe of display advertising?
It depends on what use the marketer requires them for, as each of these can be used as awareness or engagement pieces. Used in the right way, they can be very effective as either awareness or engagement tools. At the end of the day, the question isn’t whether they fit in, but where do they fit in and with whom? Having the right data-driven strategy will make it immediately apparent that the segment you’re addressing will respond to these tactics. Unfortunately, some marketers are just guessing.
Isn’t there a negative connotation associated with display advertising due to overuse or lame deployments of “banner ads” in the early days of web advertising?
Yes and no. They get mocked across the industry for reasons you’ve mentioned, although the data says they’re highly effective in generating awareness and even more effective in closing sales when used as part of a retargeting campaign. This was confirmed in an attribution study we ran recently with Facebook.
The algorithms and technology behind ad servers are getting really interesting, so we’ll learn more about their true value in the years to come. There’s a negative connotation, but the data says this is unfounded.
Are there differential impacts for display ads embedded in HTML email as compared to, say, sidebar display ads on a blog post?
Absolutely. You’re having different conversations with different audiences. Contextual information is important.
What is your opinion of “house” display ads? What can be done to make them more effective? Should they be rotated on a schedule, and if so, what determines the rotation schedule?
All things should be determined by what your data tells you. How people responded in the past, what segments are evolving and how your predictive models tell you is probabilistic for the future. House display ads are a tactic and they have their place. The trick is in identifying it.
Will Big Data for marketing improve display ad quality?
I think so. Things like viewability are really just getting us closer to properly tracking display performance. It’s giving us more data to assess their true value. This has a two-fold effect:
- Publishers will understand the true value of their placements and inventory, which will result in changes in where ads are served.
- Advertisers will get even more data on what drives awareness or engagement from readers.
What are the tradeoffs to be made when targeting ads to specific populations? Some bloggers have mentioned that ads can become “creepy” when they too closely monitor internet or other behavior.
Evidence suggests that consumers will accept the privacy violation if the value-add is equal to the level of violation. We’ve seen this in numerous campaigns: remember the airline that stalked social profiles of their customers to come up with the perfect Christmas present? Huge privacy violation, but obviously, the customers thought it was worth it because the airline offered great value in doing so.
We’ll see more and more of this discussion as always-on voice command applications become more prevalent. If Amazon Echo records all of your conversations and makes it easier to order through the store, I don’t think Amazon customers will have a problem with it.
Do you use data outside of display ads for analytics? E.g., holidays, weather events, launch dates for products or storefronts?
Personally, I believe flatly ignoring any data is antithetic to smart data driven marketing.
Connect Big Data marketing with Big Data sales data using Syncsort ETL and Hadoop with Syncsort DMX-h.