At Numeric Analytics, we have key partnerships that are core to our business – (in alphabetical order) Adobe, Ensighten, Optimizely, Tableau, Tealium among others. Critical to our strategy is staying in touch with these partners to keep a handle on what is happening in the marketplace for Analytics and Optimization.
Adobe, Ensighten and Tealium have just finished their annual user conferences. In this post, we’ll share with you the big ideas and trends that we believe are going to shape Analytics and Optimization in the next couple of years.
Trend #1: Bottom Up vs. Top Down
The basic approach to analytics has been to first understand your KPIs (key performance indicators) and then gather the data necessary to report on them. You know what you need to know; you just need the data to tell you how you’re doing. There were largely three reasons for this:
- Cost of storage was high enough that it made sense to limit the data you collect.
- The technical limitations of relational databases make mass collection and analysis impractical.
- But most important, the inability for the human mind (unaided by technology) to see connections in massive volumes of data made it unnecessary to collect more than your KPIs require. (See trend #2.)
We’re seeing a shift toward “collect first and ask questions later,” which was a key theme on the final day of Ensighten Agility. Customers experimenting with NoSQL data stores are on the forefront of this trend.
This trend signals a shift away from thinking that we already know what is important to admitting that there are connections and correlations in the available data that, when combined with Trend #2 (Algorithmic Discovery), can be the basis for discovering new KPIs, secondary metrics and opportunities that we’ll miss if we simply focus on what we believe to be our KPIs and the factors that influence them.
This certainly doesn’t mean that KPIs are unimportant – they’re actually more important than ever. But they will play a different role moving forward, which leads us to Trend #2.
Trend #2: Algorithmic Discovery vs. KPI-driven Reporting
KPIs aren’t only represented in rearview-mirror reports. They should provide the context for dealing with the bottom-up approach to data collection. For example, if Average Order Value is a KPI for you, today you are likely pulling a report that assembles data from various sources and puts it in an Excel spreadsheet for distribution. You may meet with key stakeholders and wonder why from month to month it changes (or remains steady). You may decide to try some things to drive it higher.
By 2017, we’ll see more “algorithmic discovery” against this data. This means that there will be an increasing number of programs that will help you discover trends, correlations, and connections that are not easily identified by a lone data analyst looking at the data. Maxymizer (an optimization tool) proactively identifies segments based on the data that it collects. BeyondCore scans large, curated collections of data to suggest correlations that would take data analysts many hours to figure out.
This algorithmic discovery will be the starting point for further human-driven questions and analyses as well as your optimization efforts. For example, you may find that “customer sentiment” and “fuel prices” have a key effect on Average Order Value. This would lead to important new secondary metrics to monitor against your primary KPIs. It can also help your optimization team develop new approaches to experimentation and testing that would not have been possible without algorithmic discovery.
Trend #3: Omni-Channel vs. Multi-Channel
“Omni-Channel” was a recurring theme at the early year conferences for Adobe, Ensighten, and Tealium. Let’s be clear about what this means. In a multi-channel worldview, each channel stands alone and has its own funnel and KPIs. Digital marketers recognize that people traverse channels on the way to a purchase, but their ability to understand the “customer journey” (another ubiquitous phrase) across channels is limited. Why? Because each channel typically has its own analytics data silo. Analysts must mash up this data to get any understanding of this “journey”. That mashup takes time (and may not even be possible), and we typically see metrics like 4:1 or 5:1 with respect to how analysts split their time between working with data and actually delivering insights that drive real value.
To make matters worse, attribution is massively difficult in this world, thus making it hard to manage vendors who each take credit for a “lead” or a “conversion.” If you can’t understand how a customer traversed channels to get to a conversion, then it’s your word against your affiliates’ when it comes time to negotiate the bill.
“Omni-Channel” is our industry’s way of recognizing that customer journeys are messy – messy, that is, if your perspective is of multiple independent funnels each with its own siloed analytics solution. The major players are taking this on. Tealium’s Audience Stream, Ensighten’s Activate, and Adobe’s Audience Manager are all focused on solving the problem of tracking a customer journey by tying session information back to a unique individual identifier. Add in Data Management players like LiveRamp and Krux to bring additional data into the process, and your ability to understand (and optimize) the customer journey across channels starts to seem possible.
Trend #4: Segmentation-Orientation vs. Page-Orientation
When optimization tools first came on the market, they gave us the power to run tests on pages to optimize for particular behaviors we wanted to happen. For example, if we move the “Add to Cart” button, will we net more cart adds? Optimization tools (A/B and Mulitvariate Testing) were designed to identify clear “winners.” This meant aggregating the results and declaring that one variation among all of the variations tested was the best.
We could not, however (at least in the early days), slice our results by segment – for some people (e.g., late-night page scanners) it worked better in the upper right; for others (e.g., lunch-time detailed readers) lower on the page worked better. Declaring winning designs ignored these segment differences thus potentially leaving money on the table even as we increased conversions.
To recapture the money on the table, optimization tools started taking segmentation into account, which has led to our current focus on personalization – serving the most effective content to the visitors who occupy a given segment.
What we’re seeing today is the ability for tools to proactively identify segments (see Trend #2 Algorithmic Discovery). When you combine this with Omni-Channel capabilities (Trend #3), you can shift your focus from optimizing for pages to optimizing for segments.
For example, if your analytics capabilities can associate a unique individual with a bunch of independent sessions across channels – online, mobile, and offline – then you’re a short step away from being able to say, “We are underperforming in conversions for ‘News Junkies in the Northeast 25-35’” (or whatever your segments are). And if your optimization tools can integrate this segment information with triggers on multiple channels, then personalization now becomes an Omni-Channel possibility. You will have combined Trends #3 and #4 into a powerful analytics and optimization platform that undoubtedly will combine all four trends into a digital marketer’s dream.
Further Reading
If you’re interested in getting started with personalization using Adobe Marketing Cloud:
This Customer ID Is Critical for Omni-Channel Marketing
Adobe Marketing Cloud Visitor ID Implementation with DTM
If you’re interested in optimization:
Getting a Testing Program Off the Ground
Six Best Practices for Maximizing Conversion Rates
Optimization: The First Step
