Why Leveraging Your First-party Data is Hard (and what to do about it)

I recently had a piece "Brand Data Becoming Pivotal To Digital Advertising" published on MediaPost. The TL;DR: with digital advertising data able to travel a two-way street, ad tech ceases to be an end unto itself and must take on a symbiotic role with brand data.

James Smith, President of J.R. Smith Group, submitted a great question (multiple great questions, actually):

"Enjoyed your thoughts. I wonder if you could address a few related issues?  First, some of the most valued data is lodged behind walled gardens and the gardeners are loathe to release it.

Second, data analytic skill sets across brands vary dramatically. Knowing the data, IT, and stats is one thing, knowing the business vertical, for those analysts is another (in short supply). Also read that as tech stacks and AI can't do everything effectively. 

Third is the issue of data overload. Many brands still have trouble calculating reliable ROI numbers for some digital messaging platforms--and they are awash in data. Thanks."

Considering that these are big questions, I figured I'd write these up here on LinkedIn. My responses to James' three questions:

How do you get the data you need from Walled Gardens?

First, I think it's safe to assume that the door is closing on access to highly granular, user-level data within Walled Gardens. The ethical, regulatory, and legal environment is leading the owners of large addressable user graphs (primarily Google, Facebook, and Amazon) to implement solutions like "data clean rooms" or to build prescriptive targeting and reporting restrictions right into their platforms.

The good news is that this resets the playing field for all advertisers. If you're having trouble getting your "data wish list" fulfilled by the Walled Gardens, so are your competitors. But what your competitors may be doing, instead of fixating on turning back the clock, is adapting to this new environment and finding new ways of measuring performance, building audiences, and generally getting answers.

Here are a few things brands can do:

1. Use independent (i.e., third party or "3P") ad serving and measurement providers (or even an independent bidder). These solutions lack their own user graphs, which makes person-based measurement challenging, but 3P tech does tend to provide more open access to the underlying data (log files, etc). However...I see "independent ad tech" eventually running into many of the same regulatory, ethical, and legal headaches that the "Triopoly" is dealing with (not to mention technological challenges introduced by Safari ITP, ad blockers, media fragmentation, etc). And if you're using 3P tech, there are then the problems of reconciling the data back to what the Walled Gardens are telling you.
2. Use the native features of the Walled Gardens! Yes, Walled Gardens protect their data (And wouldn't you? It's worth untold hundreds of billions of dollars to them.), but it's not as if the Walls don't have windows, doors, and "librarians" who can help you find the answers you need. Google's Ads Data Hub (ADH) is a specific solution to this. ADH basically gives marketers log-level insights, but ADH won't ever return data more granular than 50 users, making it impossible for marketers to gain access to private user data. It's not a bad compromise. Many of the measurement use cases solved by user-level granularity are probably served equally well at 50-user granularity; that's still pretty granular.
3. Flip the script. Instead of relying on digital performance metrics (such as CTR, a proxy metric that is inexorably becoming obsolete), structure your campaigns around demonstrable business outcomes such as Revenue, Order Value, Lifetime Customer Value, Foot Traffic, and Customer Loyalty. These aren't easy to tie back to digital campaigns, but valuable things aren't ever easy. The question an advertiser needs to ask themselves is "Why are my business outcomes so difficult to attribute back to my digital advertising?" Look at the whole business; the brand, the entire customer experience (CX), the entire customer funnel, the whole fulfillment chain, and post-sale activity. Is your business sending multiple marketing channels through the same order/fulfillment/CX process and thereby shooting the marketing team in the foot when in comes time to figure out which customers originated from where? Is customer rewards (e.g., loyalty cards) data not being used to measure outcomes? How much measurement is captured post-sale? (on that last one, I bet most marketing teams have little to no insight, despite the fact that customer retention is utterly critical to business health) Are there basic tracking elements (e.g., site analytics, call center analytics, store check-in analytics) not in place?
4. Take a second look at incrementality and lift-based measurement. Digital marketing perhaps got a little side-tracked over the last 10 years thinking that literally every last thing can and should be measured at a person (user) level. Some people in ad tech would call me a killjoy, but I don't think marketing and advertising will ever completely solve that problem; it's a mirage, so don't put all your eggs in that basket. There are many questions to which we'll only ever get inferred, indirect answers. Making sound decisions based on inferences (as opposed to verifiable facts) is how most of the world's most important decisions are made, but this brings many data-driven digital marketers out of their comfort zone. Nevertheless, the most competitive, successful advertisers will invest tremendous effort into finding the right questions to ask in the first place. In addition, incremental measurement requires careful setup of controlled experiments (the Scientific Method definitely applies). You can't just do what you've been doing and expect it to translate over to a lift-testing methodology.

How do you find people to do this stuff?

There isn't an easy answer to this question. My employer MightyHive (we're a services partner for digital advertisers, so "Renaissance Digital Ad Experts" are the core resource we offer to brands and agencies) finds this to be the true limiting factor in digital transformation today. It's looking more and more like the next-gen digital marketer is going to require a Unicorn-level skill set. There are not enough people to fulfill the vision.

What can you do?

1. Hire a partner. An agency, a consultancy, or ahem...my employer MightyHive (sorry for the plug, but this is what we do and why I write on these topics). It's not much different than when you need to remodel your bathroom. The rare homeowner can do this (and, perhaps just as importantly, do it well) themselves, but most can't, so they get help. The key here is to find a partner that's going to "teach you to fish." If the key to success for brands is leveraging their own data, they're going to need people internally to maintain and improve upon a system set up by an external partner.
2. Break the problem down until you've got job descriptions and training curricula that aren't works of fiction. Many advertisers might need a consultancy to help with this, since it could touch many areas of the business itself as well as the external tech the business employs. It's going to take a lot of time for businesses to undergo this transformation and find the right blend of partners, technology, automation, and staff (real staff, not Unicorns) that comprise a sustainable system.
3. Go outside the Marketing world. And I'm not implying that you hire data scientists and engineers (although that's part of it). I'm suggesting finding people who understand business operations: commercialization, pricing, order fulfillment, backend enterprise software (e.g., order tracking, accounting systems, and CRM), sales, customer service, and so forth. Generally speaking, career Marketing folks don't have a grasp on these areas of a business and "don't know what they don't know." Get them some help connecting the dots.

How do you handle data overload?

I've addressed aspects of the "data overload problem" above, but data overload is indeed a massive problem. Many marketing teams aren't exactly lacking in data, but they are struggling to connect dots, get a toehold, and generate insights they can use to make better, faster decisions. In short, I'd say: be picky. Not all data is created equal.

If all data is not created equal, how can an advertiser classify the inherent value (or cost) of investing resources across various data sets?

• Is the data set a facsimile of customers (facsimile = is it just cookies? is it "probabilistic" or "deterministic?"), or does it represent actual people? The Triopoly (Google, Facebook, Amazon) has actual users. An advertiser's first-party systems (e.g., CRM) contain actual people. There are a slew of business and technical reasons why actual people ultimately trump cookies, device IDs, and so forth (which is not to say that third-party or other data can't help or should be disregarded, but it is several steps removed from reality).
• How easy is it to get the data? Is there an API? What format is the data delivered in? How many developers (or developer hours) would it take to build a connection to the data? Can your partner (whether they be an agency or a tech partner) take the initiative deliver the data to you (or directly to one of your systems) in a way that's more convenient or automated?
• How timely is the data? And I'm not implying that real-time is always better. But think about how you're using the data and whether its timeliness supports your use-case. For example, do you need a massive Amazon Redshift deployment to process real-time data for a monthly report? Maybe not.
• How granular is the data and what level of granularity supports your use-case? This leads to a bigger question of what use-cases you are going to move forward with.
• Are you able to test the data? Is there some way to verify that it's accurate? This doesn't mean that the data itself actually needs to be 100% accurate, but if it's only 90% or 80% accurate, you need to know that in order to build tolerance into your models. And further, it helps inform what level of tolerance you can accept as an advertiser.

If you were to make a matrix to evaluate data sets/sources, there are plenty of other dimensions you could bring in, such as price, data security, etc. But in the end, a process of evaluation then enables a more diligent approach to tackling data sets iteratively rather than succumbing to a deluge of data sets you haven't bothered to understand.

I hope this helped

This was a long-winded response, but if there are one or two useful nuggets to take away, then I'm thrilled. Thanks for the great question(s), James!