Leverage Analytics with Contribution Analysis Tips Tricks and Best Practices: Part 1

Twenty-two centuries ago, Archimedes famously said, “Give me a lever long enough and a fulcrum on which to place it and I shall move the world.” Since your boss routinely expects you to move both heaven and Earth, it seems you need an even longer analytics lever; well, either that or become stronger than Achilles. Sounds daunting, but in this two-part post you’ll learn all is not lost. You can transition from exclusively using the common short levers (e.g., simple dashboards and reporting) toward including super-long, rare levers that empower you to deliver smarter analysis.

In my previous post, on Adobe Analytics’ new Contribution Analysis capability, I explained how a non-quant analyst can use Contribution Analysis to complete in minutes a task that would otherwise take even a data scientist weeks to accomplish. Contribution Analysis is indeed that uber-lever you’ve been looking for without even knowing it. Contribution Analysis unearths the driving forces that explain anomalous changes in your trended data, opening the door to proactive decision-making. How’s that for leverage?

In this first installment, we’ll recap what Contribution Analysis is and what it does, after which we’ll dive into the tips, tricks, and best practices for the critical first phase of your Contribution Analysis—selecting the right variables to maximize clarity.

Contribution Analysis

Think of how the proverbial 90-pound weakling, armed with a long-enough lever, can move enormous boulders that even a demigod would have a hard time shifting. Since demigods are hard to find, expensive to retain, and outclassed by long-lever-wielding weaklings, Adobe decided to help its clients by creating a long lever for your analytics practice. Because you’re no weakling, I imagine the heavens are already quaking. Adobe Analytics’ strategy is to provide simplified, yet powerful, tools that empower analysts and marketers to perform analysis tasks far beyond their intrinsic level. In essence, Contribution Analysis within Adobe Analytics is a manifestation of our approach in delivering against our strategy and vision to bring data science to the masses.

Contribution Analysis intelligently identifies possible causes, or contributing factors, that help explain changes in trended data and other anomalies. These powerful algorithms automatically query tens of millions of datasets, and then apply machine learning to determine the strongest contributing factors.

Before you object that such powerful algorithms may be as difficult to harness as Zeus’s thunderbolt, let me assure you that we’ve designed them for the everyday business user rather than the quant or data scientist. This means even mere mortals have already wielded them effectively in beta testing (see Figure 1), with Earth-shaking results.

John Bates 1
http://blogs.adobe.com/digitalmarketing/wp-content/uploads/2015/03/John-Bates-1.png

Figure 1. Beta test proves Contribution Analysis puts Adobe’s data-scientist-in-a-box at your fingertips.

In the remainder of this post, you’ll learn tips, tricks, and best practices to ensure that you get the most value out of Contribution Analysis on Day 1. So, let’s get started.

Combining Contribution Analysis Features for Cosmic Results: Tips, Tricks & Best Practices

Just as combining different metals can create a stronger alloy for your uber-lever, combining Contribution Analysis tips, tricks, and best practices will help you create the ultimate Contribution Analysis output. We’ll start here with variable selection, and will round off with other crucial features in the second installment of this post.

Variable Selection

This is the most important ingredient for getting the most out of Contribution Analysis. By default*, when you click “Analyze” on an anomaly within the Anomaly Detection Report, Contribution Analysis analyzes EVERY . . .

Although all these data are extremely valuable, removing some dimensions from the analysis may reduce clutter and maximize clarity. If you look at everything, you’ll probably find all your time-parting variables as the top contributors. This makes sense, but doesn’t provide much insight, so remove them and rerun the analysis. After removing variables that obviously shouldn’t be in your analysis, you may spot a few more that you know aren’t relevant to your contribution analysis, so hide these too.

Before your first analysis, I suggest you open the right-rail and click the gear icon (see Figure 2) to customize and remove generally uninteresting or duplicated variables such as (see Figure 3):

  1. Pathing (search “entry” or “exit”)
  2. Dynamically set variables (remove traffic variable version)
  3. SAINT classification parent variable
  4. Time-based variables (time-parting)
  5. Visitor Profile > Technology variables (color, java, etc.)

KEEP EVERYTHING ELSE!

John Bates 2
http://blogs.adobe.com/digitalmarketing/wp-content/uploads/2015/03/John-Bates-2.png

Figure 2. Settings (see red arrow) let you get your contribution analysis in the right gear (which is a sort of circular lever too…).

John Bates 3
http://blogs.adobe.com/digitalmarketing/wp-content/uploads/2015/03/John-Bates-3.png

Figure 3. Customize Dimensions page lets you remove uninteresting or redundant variables to improve clarity. After all, it’s easier to move heaven and Earth when you can actually see what you’re doing . . .

Adobe Analytics automatically deselects the traffic variables (pathing versions) and many of the visitor profile and technology reports that we’ve identified as being generally less interesting or valuable. However, you may override these default settings at anytime. Once you click “Save” within the Contribution Analysis customize dimensions settings, your dimension selections will persist for all contribution analyses you queue up in the future (these selections will not be applied to any contribution analyses currently running or already in your Analysis Queue).

Key Points

Contribution Analysis is a groundbreaking new tool that places Adobe’s machine-learning-based “data-scientist-in-a-box” at your fingertips. Selecting the right variables and deselecting the less-useful ones provides the clarity you need to get your Contribution Analysis going.

In the second installment of this post, we’ll delve into the tips, tricks, and best practices relating to interactions & actions, applying segments, interpretation, and top segments. Then, we’ll move on to list your likely next steps.

*Contribution Analysis is primarily an Adobe Analytics Premium capability. However, there is a much lighter version of Contribution Analysis for Analytics Standard customers. Analytics Standard customers can only analyze up to three dimensions (by default these include Pages, Products, and Campaigns) and you will only see the list of the top five contributors in the Contribution Summary.