1. Find phenomena that you want to understand (more)
2. See what information there is about this phenomena, or similar phenomena
3. Collect data of the phenomena using whatever methodology seems appropriate
4. Construct a hypothesis as to why this phenomena occurs
5. Test your hypothesis through experiment and data analysis (start again if it doesn’t)
6. Draw conclusions

Here are some examples. First, you can use customer data to drive experiments to discover the experiments to which people are most likely to respond. Tesco does this their ClubCard program to experiment extensively (it’s described at the end of this case). They have a couple hundred thousand members, and by offering discount for using the ClubCard, Tesco gathers a huge amount of data on what people are buying. This includes information on brands, amounts, combinations of items, and so on. The experiments come in two forms. They test out new products with special offers to ClubCard members, and they test out incentives to offer the club members. Each year they run thousands of these kinds of experiments. Each one teaches them something. Sometimes it’s what works, others it what doesn’t.

Amazon does similar experiments with their homepage – they offer many different versions to people and discover what works and what doesn’t. Similarly, Google experiments with hundreds of tweaks to their search algorithms each year. So data-driven experiments are one way to improve innovation.

A second way to use experiments is through finding a small-scale method to test out products and services. I’ve talked previously about how Unilever Hindustan tested out a radical new business model in India. They started a program that transformed the way that the entire company does business in that market by setting up just seventeen women in the new business model. The point of the experiment was to figure out if they could help these women set up their own successful businesses by helping them build their skills, to figure out if there was sufficient demand for these products in the poorer regions of the country, and to sort out how to best build the distribution network.

Once they discovered what worked and what didn’t, the slowly expanded the program to include 45,000 women serving over 100,000 villages throughout India. If the initial trial had failed, the cost would have been minimal. And because it wasn’t a big flashy launch of a full new product or service, no one would have even known about it. The second way to innovate through experimenting is to find a small trial.

The final way is even cheaper and faster – and that is to figure out a way to try an experiment with an idea. That’s what we’re doing here on the blog. There are over 300 experiments here. Some of the ideas have worked, and some haven’t. The ones that worked have been scaled up into public talks, sections of lectures, article ideas for magazines and journals, and, if everything goes according to plan, a new executive education program that we’re aiming to roll out later this year. That’s a lot from just testing out ideas.

So the point with experimenting to drive innovation is to do it as cheaply and quickly as possible. If you can trial ideas, do so. If not, figure out a way to test out new products and services on a small scale. And if you have access to lots of data, then you can use this to drive multiple small-scale experiments.

Try things, learn, and scale up. That’s using experiments to improve innovation.

(photo from flickr/steve.hubbard under a Creative Commons License)