What should the outputs of an analytics program be?
I’ll argue the outputs should include: profit, evidence (historical), clarified-concrete-measurable goals, decision support (scenario analysis), forecasts, customer intelligence, and competitive intelligence: all resulting in an expanding base of knowledge and competitive advantage.
I didn’t mention mediums in that description.
The medium is both the message and the problem preventing most programs from becoming a source of competitive advantage for their organizations.
We’ve got some pretty big problems with the mediums.
There’s a huge amount of work that’s going into moving data from multiple systems into a single system: a process called aggregation.
What’s the most common aggregation point?
A spreadsheet. It’s excel.
It’s a not a very good solution. Usually it isn’t semi-automated or effectively QA’d.
I’ve pushed excel beyond it’s natural limit. Even when it’s made very pretty and functional it is at best a stop-gap solution.
Excel should be to an analytics practitioner as Visio is to an information architect.
It can be a useful tool to express the model, view, and controller to a tech: but it shouldn’t be the platform. It shouldn’t be the principle medium through which a practitioner communicates with a huge audience.
I’m expanding on that word: communicates.
Have you ever watched somebody open up an excel spreadsheet? Have you ever watched them consume the data on the page?
Just watch them.
Chances are you see a heavy sigh and a whole lot of squinting. You’ll also note that the time spent with the page will vary from just a few seconds to a few minutes (at most).
I’ve watched others take the spreadsheet and run sums and functions on the data. They’re effectively torturing it themselves to make the spreadsheet talk. They’re trying to learn something from the data.
Spreadsheets don’t teach on their own.
I’ve done it myself. Once I’m satisfied with understanding what is going on, I immediately jump to finding out why it is going on and how I can improve it.
In figuring out that why, my first instinct is to go to the go-to people and start asking questions. Frequently an hour worth of talking can save a week’s worth of digging. Sometimes I need more data – but I know how to phrase a query.
I can report that phrasing a query can be very hard to do and it’s seldom done really well.
The memegenerator below demonstrates what happens next.
Humans have a hard time guessing how long it takes to put something together. If it only took somebody a minute to read something – then it must have only taken the analyst ten minutes to put together. (right?)
Excel sheets within any organization proliferate (how secure are all those sheets, anyway?). The result is the perception that they’re cheap. And if they’re cheap, the demand for MOAR comes far and furious.
But that’s the wrong medium. They might scream for more dashboards – but a dashboard can’t possibly answer a complex query or tackle a complex problem.
A dashboard doesn’t tell anybody why something is happening. It tells them what is going on.
It offers a very small incremental competitive advantage. Sure, they might know MOAR WHAT is going on, but better decisions are made using causality, not knowledge of past state alone.
Analytics becomes a source of spreadsheets instead of a source of competitive advantage.
If Excel is not the right Medium – what is?
I’m arguing that the medium ought to match the objective.
If you are replying to a complex query, a presentation – or dare I even suggest it – an animation/video through visualization ought to be the right medium (GGOBI is free). A complex question typically results in a simple answer that need a long explanation to have face validity.
For instance – if the CFO were to ask the web analyst “what are the traits of our most valuable customers?” – the answer might be simple: “People who buy often and say good things about our products”. The story explaining why is just that: a story.
You don’t use a spreadsheet for that.
If you are asked for ongoing data so that a manager knows what is happening with their section of the website – then it ought to be a tight dashboard based on clear business goals: tied as closely to how that manager is bonused as possible. That dashboard ought to be in a web based format that is designed for dashboarding. Ideally, the dashboard would have a function that enables self-exploration. (Enter the world of vizualization and democratic access).
There’s another reason too:
Inquiries for WHY something is going on naturally lead to demands for MOAR metrics to be added. This means that what is born as a small dashboard of 10 metrics grows into a 111 metric disgrace of a report.
The medium of an excel spreadsheet is simply incapable of keeping up with the wave of human curiosity that is aroused from seeing the surface of the data.
If you believe that the principle output of an analytics program is just data: then the MOAR cycle of metrics is not a problem for you. That’s just fine and we have nothing to talk about.
If you believe that the principle output of an analytics program is competitive advantage for a firm: the mediums we use as practitioners must shift. If you agree with me – then we have a lot to talk about.