03 September 2014

To find smash-hits, ignore statistical significance !

It's not the end of the world if you're wrong !

The biggest mistake in market research is to worry about being wrong.

Ironically, research can learn this lesson from one if its great detractors, Steve Jobs. He who famously dismissed market research saying "People don’t know what they want until you show it to them".


He was right about customers' lack of self-insight, but wrong about the value of research.


Businesses want blockbusters, to crack the big time, go viral, be trending (upwards). They want and count hits as Steve Jobs knew and proved.

Steve Jobs is famous for having launched multiple commercial smash hits like Macintosh, iPod, iTunes, iPhone, iPad.
However, he also rather less famously launched multiple commercial busts like NeXT, Apple Lisa, Apple III, PowerMac g4 cube.
Jobs' skill was not in identifying what was true, it was his willingness to try.

Jobs was right to dismiss research - at least as it is currently conducted. Most research aims at minimizing mistakes - but in so doing, misses the opportunities.


The parent of research is the scientific method that reflects a fairly common human preoccupation with finding truth, with being right - or more exactly, not being wrong!
But it's time for business research to cut the umbilical cord. Let market research serve the needs of clients, not science. How do we do that?
Well, the misdirection and misuse of research in business is reflected in the ubiquity of inferential statistics. (Stay with me here - if you're scared of statistics, you're going to love the final message!)
Research in business (as elsewhere) is a process for reducing uncertainty. However, if there's one thing of which we can be sure, it is that uncertainty will always remain.

The key is deciding which errors we will accept. There are two possible errors that can be made in research:
  • One is to identify an idea that turns out to be false. That is, a "false alarm" (= false positive or Type I error).
  • The other error is to deny an idea that turns out to be true. That is, a "miss" (false negative or Type II error).
Academic research focuses on controlling false alarms. Scientists are searching for truth and don't want to say something wrong. This is underscored by the convoluted notion of 'statistical significance' based on p < .05.
However, business research is a search for opportunities, a new advantage, a new product, a new communication, a new tool that can advance sales, market share, profit and maybe all three.
Imagine a company runs a little internal study showing that offering a free doughnut and coffee will bring in 4% more customers. Not pursuing the idea because the result is not statistically significant (p=.11) would be very foolish.
The solution is to dispense with statistical significance!
  • Calculate and report effect sizes and confidence limits. Focus on the potential 4% gain!
  • Relax the alpha level. Set the cut-off for significance to p=.1 or even higher! 
Research should emulate the method of Steve Jobs, shift its focus towards the pursuit of worthwhile tries and away from worthless truths.

It's not the end of the world if you're wrong - as proved by the Ford Edsel, New Coke, and the Apple Newton. But it can be a long, slow, stagnating death if you miss opportunities.

--------------------------
For the evidence that experimentation is more likely to result in market growth, see Shugan & Mitra 2014. In fairness, experimentation is linked to market growth, not individual growth. Experimentation helps the market, but may see a number of participants lose.

1 comment:

  1. Apple's hits & misses: https://www.theguardian.com/technology/gallery/2019/jun/28/jony-ive-8-hits-and-8-misses-from-20-years-at-apple

    ReplyDelete