What is Google Next?
Next is the show that brings together Google’s different products in cloud computing, machine learning, big data, and advertising. At 25,000 attendees, it’s the largest event that Google has put on and represents their attempt to gain ground in the enterprise market against Azure and Amazon Web Services (AWS). Given the focus of the products, attendees ranged from analytics and data science leaders at large brands, to release engineers at startups.
I attended as many of the sessions as possible pertaining to customer data, marketing analytics and business applications of machine learning and noticed a number of consistent themes across both CMO and CTO worlds. Data is at the intersection of both, demonstrating yet another example where business and tech leaders need to partner closely.
Part One: Business themes and challenges across marketers in attendance
The well-traveled cliché of “the right (product/message/offer), at the right place at the right time” comes up often at events like this, though it’s still a far-off goal for many. The amount of data for marketers to sift through is growing, and consumer journeys are becoming more fragmented, meaning that the goal line is always moving. However, many customers were talking about success they’ve had building on the Google Cloud Platform (GCP).
Challenge 1: Measuring & attributing performance across channels
While the advertising platforms in use today all provide valuable insights and reporting for their channel, marketers lack a solution for viewing performance across all channels in aggregate for a given campaign. The problem is compounded when each platform uses a different KPI to measure effectiveness. The most basic approach -- copy and pasting values into a deck -- is a simple point solution, but doesn’t scale when you need historical views and have dozens, if not hundreds, of campaigns being run across multiple markets. PowerPoint and Excel can only take you so far, which takes us to the next challenge.
Challenge 2: Collecting and organizing data for analysis and visualization
Because Google already has a large amount of data for most advertisers in the form of paid search, video, and display performance, they’ve made it easy to load this data into their data warehouse product (BigQuery) to analyze of huge amounts (terabytes) of data in a matter of seconds. When combined with data from other channels via APIs, and matched against a lookup table, this provides a powerful way to automate the collection and analysis of omni-channel campaign data.
Lesson Learned by Many: To be successful here, you need to have a flexible architecture to grow into. Start with one data source and build from there, rather than trying to bring in every channel at once. This data solution will be the foundation for better digital attribution of fragmented journeys, but that shouldn't be the initial goal.
Challenge 3: Getting from measurement to activation
For marketers that do have control of their data, the next challenge was about being able to get the right insights and be able to activate against them in a reasonable timeframe. A common theme had to do with how long it takes to to perform advanced analytics and create insights, thus contributing to how long it takes to actually act on data.
The Solve: Several online retailers discussed how they built their own dashboard that brought together fast-moving data, along with data from their BigQuery database and provided immediate insights. In one example, near real-time search performance and site analytics data went through an algorithm to suggest different keyword strategies on the same screen. Using the AdWords API, those insights could be acted on immediately.
Challenge 4: Personalization
Customers who had solved challenges around data management had some of the most compelling examples of how they were driving personalization. One online retailer with 1.6B in revenue talked about how they don't have a single home page; they actually have millions of different home pages for each of their visitors. Making use of both known and anonymous data, they relentlessly test different algorithms for what converts best and segment smaller audiences for additional experiments.
Part Two: Common challenges for technology leaders in attendance
Tech Challenge 1: Data lakes or data warehouses?
The Data Warehouse is the most mature of the two and has a well-defined schema built on a relational structure. It will also typically have retention limits on the actual records, as they have been designed against specific reports that are used to run a business.
A Data Lake, on the other hand, isn't designed for a specific report or purpose. Its role is to collect as much data as possible, in an unstructured and limitless format, so that a data scientist can find a use for it in the future. Until recent advances in cloud computing were developed, this previously wasn't possible, because costs for storage and analysis of unlimited sets of data was completely unrealistic.
An emerging concept, the Customer Data Platform (CDP), is a use case for a Data Lake. The CDP is intended to connect, match and align all known data about a customer in one place with records that previously have never been matched. This is the "360-degree customer" vision that's been discussed for years, but has yet to truly be realized. This is an interesting space to watch – and to get started, the data lake is essential.
Tech Challenge 2: Marketing technology implementation fatigue
From ERP to CMS to CRM to E-Commerce, enterprises have invested lots of time and capital in implementing packaged software. While largely successful in the end, the journey is usually pretty bumpy. Many organizations take a pause after a large software project to stabilize and plan the next phase of their journey. With advances in Platform-as-a-Service (PaaS) offerings, it becomes more compelling to look at how business requirements can be met with off-the-shelf services instead of off-the-shelf software. Which leads to the next theme...
Tech Challenge 3: A reversal of "Buy before Build"
The longstanding strategy for most mature IT organizations was to buy software from a vendor before trying to build it themselves. No one would really consider building their own CMS when there are dozens of vendors with 20+ years of R&D experience. However, the opposing view is that all those years of R&D have led to large, complex products with more features than an organization may need. So why pay for features that may never be used? For some specific digital marketing analytics use cases, especially those pertaining to data management, the library of APIs available provides a compelling alternative. This is where DIY in this case makes sense.
Additionally, building bespoke solutions incrementally off of cloud APIs is also better aligned to agile software development. Getting a release out every 2-3 weeks with additional functionality is more appealing than waiting months, or years, for an implementation of traditional packaged software. This is a dominant model in the startup world where the capital to invest in licensing, training and implementation for a Big Bang release doesn't exist. The utility model of pricing for APIs is far more appealing.
Tech Challenge 4: Ownership
In 2013, the CIO of General Motors, Randy Mott, famously embarked on an insourcing program to reverse a trend at GM wherein 90% of IT was outsourced. The goal was to own the innovation and all aspects of GM's digital future. This trend of insourcing, and creating blended team models, is happening in marketing as well -- CMOs are looking to bring parts of their media programs in house. The goal in both cases is to own the data as well as the algorithm.
Technology and Marketing leaders both want to know who will own the data and where it will be kept. Solutions that sit in the middle of a brand and customer and take control of the data are non-starters. As an example, BMW just announced that they will be deploying Alexa in certain vehicles. What's most interesting is that they will proxy the driver's request through a BMW server (introducing an obvious delay), before going to Amazon, so that BMW also owns the data. It's an example of a brand compromising on a tightly controlled user experience to ensure they don't lose control of valuable customer data.
We believe there's a strong case to be made for both the Marketing Cloud approach (e.g. Adobe, Salesforce and Acquia) as well as the Cloud Platform approach (e.g. Google Cloud, Amazon AWS and Microsoft Azure), and would be happy to discuss more.