weeks - 12:50 pm on May 4, 2009 (gmt 0)

I worked for a large regional retailer for 25 years. It has always been this way. There is no way to predict patterns of sales on a day to day basis. We're really learned this with the web. It will drive you nuts.

It's even true of traffic flow. Roads will average x cars at y times, but on some days, for no reason, there will be 3X cars at y time, and then, one day, there will be X/2 on the road at y time.

There are firms that will take your sales data and link it to the weather, for example, where you are suppose to see how the weather impacts your customers' buying habits. The idea here is you can "forecast" sales like the weather. Ha. (Good weather, good sales. Bad weather, bad sales. Please send me a check for $1,500.)

This is why it vital to use trend data that comes from over a long period of time over a large sample. And, then, don't worry about day-to-day traffic, one way or another. Or even week-to-week.

Typically, a new enterprise will finally get one of those weird big-flow periods and instead of seeing it as unusual, they'll decide that, ah!, the world has finally seen my biz for the outstanding value that it is. Then, of course, sales will drop to normal and, then, a low-flow period will hit--for no reason anyone can see--and that will throw them into crisis, when in reality the long trend is good.

Savvy ad agencies & consultants (and even some webmasters) know to make some kind of change right away (green to blue, or blue to green, is always safe) when a biz in this kind of "crisis" comes in the door with checkbook in hand, because the sales flow will likely correct itself naturally but they want to get some kind of credit for the change, which is tough to do if they have done nothing.

"Wow, thank you. I didn't know blue was so powerful!"

"Yes, we professionals have found how these little things can make a big impact."