As web analysts, internet marketing managers, or web consultants, we’re all looking to prove our worth daily. Hell, everyone who cares about their livelihood is. So one of the things we, as a group, are leaning towards is prognostication. And that’s a dangerous line of thinking.
This month, Tim Schultz wrote a great piece for Playboy Magazine called the Outlaw Economists. To sum up, he indicates there’s a group of rogue number crunchers out there that tend to think that all of the complex economical models we’ve been building in money labs across the country are simply wrong. They use too many variables, the wrong variables, and predict the impact of financial regulation and policy changes (federal and corporate) using… well… bad math.
It’s argued that simply studying history, as opposed to modeling, would have let us see the corruption that we’ve seen on Wall Street coming far sooner, rather than subscribing to the “risk free” direction as dreamt up by the modelers.
I tend to agree. I also think you should read his piece when you have a chance. I realize that means you’ll have to buy or borrow a copy of Playboy. You’re welcome.
So what does that have to do with us?
Economic Modeling and Web Analytics
To me, it means we can’t afford to fall into the same trap. With the “advent” of multi channel attribution, multivariate testing in both ad copy and landing pages, and countless other measurements we’re trying to make all driving business decisions, I think we need to every once in a while take a step back and see the big picture. I came up with a very basic pyramid chart on how I choose to invest my analytics time, and what I study.
Don’t Predict the Future – Determine it with Fact
By spending most of my time looking specifically at what has brought revenue to the organization in the past, I can then plan incremental growth proven on SOLID THEORIES. Leads of course is the second most important piece of the puzzle, and I devote a lot of time looking to determine what is bringing in the best quality, highest converting inquiries. But I don’t know what those are until I understand specifically where the revenue comes from.
Traffic, Testing, and ? are all variables depending on situation.
Just a note on ?. Every organization has some of these. Little wild cards we play with on the website or in the advertising. They can be just about anything, and most of them are covered under Testing. Sometimes, however, we just make changes without testing them, based on supplier changes, stock shortages, etc.
So how do I determine the future rather than predicting it? By not modeling. By instead:
- looking at historical performance
- NOT adding in multivariate testing or changes
- NOT worrying about multi channel attribution at this point
- Simply incrementally growing what has worked
There’s a huge difference in making sound business decisions based on historical performance and fact compared to building complex models that attempt to encompass every variable imaginable. One drives business, the other simply has too many variables.
Small businesses in particular need to be especially careful in this regard. There needs to be a balance between planning and executing, and building complex prediction models is best left for the Business Plan days, and not in your analytics sessions.
A Special Note on Multi Channel Attribution
I’m a self professed lover of this stuff. But I use it for what it is, which is a page in the story of how we got to where we are. I don’t think it’s always a proper indicator of where we’re going. For example, if we find we’re crediting PPC and the Affiliate Channel for a whole lot of sales together, won’t we want to take strides to decrease our cost of acquisition by closing more deals on the first visit, and eliminating an added layer of expense? I’d rather see us credit only one channel, but do it fairly by sealing the deal when it’s in front of us.
Was Nostradamus Right?
Depends on who you ask. And that’s no way to run a business. I’d like to see more of us get back to simple measurements that mean something, rather than doing math like a physicist to determine how to better sell a pair of sneakers.
While there’s room for all of that in business growth discussions, I think the analysts ought to get back to what we’re supposed to be good at: Telling decision makers why what happened yesterday happened.
What do you think? I look forward to your comments below.