3 Challenges in Multichannel Analytics

Gone are the days of Web analytics as once we knew it. And even the newer term “digital analytics” hardly captures the constellation of sources we look to today to gain insight into our digital properties and campaigns.

Sometimes it seems there are almost as many “solutions” on the market today as there are data sources. Each one claims primacy in one area or another — usually in visualization, “actionability,” or the combination of online/offline data. In the past I’ve called this phenomenon “Convergence Analytics” but most know it as “multichannel.” No matter the solution, they all are prone to rather serious pitfalls. These typically take the following forms:

1. Data Access

We’ve heard of ETL (Extract, Transform, Load) capabilities, but this problem actually precedes it. Let’s suggest you’ve identified a dozen data sources, and that you want to pool them into a single repository for later visualization. We don’t yet have robots that can gather user IDs, passwords, and auto-request APIs from vendors; nor can any ETL tool ask questions of a data provider, arrange payment for access, or talk through the undocumented access points you’ll need in order to make live (or even scheduled) connections. Therefore, be prepared to send lots of emails asking for same. Who owns the data account? Is it personal or corporate? What is the UID and password? What arrangements already exist with the data provider(s)? How much will they charge for access? Is there even an API you can deploy? In many cases, there isn’t.

There’s a great deal of human input needed to get the multichannel engine started, and lots of tinkering needed before any visualization can even begin.

2. Custom vs. Packaged

This also is not a new thing, but is likely to be a challenge faced by anyone hoping to build a multichannel insight engine. It’s nice to think there’s a single solution that can supply all of your multichannel reporting needs, and plenty of vendors who claim they’re it; but too often you’ll find that what the vendor thinks is all-inclusive in fact leaves out exactly that one thing you’re looking for. In which case, it may make sense to build your own out of stuff laying around in the yard. And there’s plenty laying around in the yard.

At least one implementation I know of deploys a version of Microsoft SQL Server with Tableau on top of it for visualization. It’s capable of pulling in quite a number of data sources (see above) and then relaying it to a reporting layer – like Tableau. Some folks have also gone to Alteryx or Birst for different flavors of same.

The downside, as usual with building your own, is that there’s little help to be had in streamlining intake and output. You’ll need to cobble together automation scripts for both — probably more so for intake, because this is where the raw data is, and often it’s hardly suitable for export. You’ll need to map data, adjust columns, and continue making inclusions and exclusions until it feels right.

3. What’s Useful to Know?

This problem could be the subject of multiple columns, or perhaps a shelf of books that would include a great many pages devoted to cognitive philosophy.

Suffice it to say you can, theoretically, visualize almost anything. How many umbrellas got sold in Appleton, Wisconsin, on days where it rained more than 0.25 inches? There’s a graph for that. But do you care? Yes, but only if you sell umbrellas in Appleton.

The greatest threat to your clarity is an overabundance of choice. Today, visualization layers are almost magical in their ability to create ostensibly helpful graphics to help you understand data trends. But you have to focus. Building a dashboard of a dozen or more metrics may make you feel powerful, but can you possibly care about a dozen things at once? And that’s in addition to breathing, typing, and letting the cat back in.

Focus your efforts. Chances are, you already know what you care about. It’s just as likely that you feel like you’re oversimplifying by saying it. In theory, you might find more to care about if you could just see what there was to care about. But trust yourself. If you sell umbrellas in Appleton, then stay focused on data that helps you do so. Forget about galoshes in Galina. Or average number of clicks to oblivion.

We don’t have the space to discuss how one arrives at key metrics, but I can say that it’s often much more obvious than you think. What is your business? If you did not have a computer at all, what would you care about? This is exactly what you should be deploying your multichannel tool to help you understand better.

Related reading

site search hp
ga hp