Hi April, Great post. For start-ups especially, we've always told clients 80% of the work in a launch is promoting the problem--not the solution. Not only that, but you can't properly position or try to steer the category people put you in without grabbing the problem definition by the horns up front. Adding social media to the equation makes it truer than ever, and as you noted, provides a much better means for valuable feedback and convos. --Steve Parker

Hi Amber, thanks! I've adjusted my view of measuring "everything" to exclude: 1) anything where the ROI is already known accurately enough for making decisions; 2) anything where the cost of measurement would exceed the benefit of the activity being measured. (This is how I summarize Katie's comment.) Part of what I was trying to say was that measuring (vs. not) should be the default. You two have reminded me to add, "as long as its practical." Your point about trailing vs. leading indicators is noteworthy--especially in these times of budget stress and risk aversion. Everyone wants their analytics to be "predictive," but getting there usually requires a series of trailing metrics. But even those have to be extrapolated and more or less taken on faith. "Proof" is hard to come by.

Point taken, David. When we can both demonstrate the product and entertain, like BlendTec did, we'll increase the chances of success. But I was looking for something more broadly instructive: some products really cannot be demonstrated at all, while others can be--but are excruciatingly boring. Then a new question pops up: can entertainment unrelated to the product ("undemo") still deliver viral marketing benefits? And how many viral campaigns, like the ads of old, might generate great creative and entertainment but fail to move the sales needle? Lots of viral video jockeys are settling in for a cozy Mad Men-like return to the days when "clever" and amusing was enough--not tying it all back to sales results. Looking forward to continuing this convo.

I now see I wasn't clear. I didn't say experimentation has no merit--and that's not at all what I meant. I was trying to comment on what makes experimentation useful. It's necessary--BUT. What I object to is the idea that it's all right if you don't/can't learn anything from one experiment to then apply on another. I have heard "experts" say outright that since "you never know what's going to work," you should just keep trying different things. My point is, if you never know what's going to work, you should get another line of work--you're not a professional. And repeating unique experiments that all fail until the cows come home is no path to success. The reproducible part of experimentation is the most important because it's that part that is instructive and that you can apply to the next project/experiment. Everyone or every organization has a limit on how many failed experiments they can afford or tolerate. That's what I meant by reproducible. It's irresponsible to recommend to apprentice, DIY or part-time marketers, or any marketer really without much creative experience, that they keep throwing those long bombs into the end zone because some day someone might catch one. It just so rare. It's like telling the general to take his cannons down to the seashore and start firing volleys randomly into the ocean because an enemy target might wander by.