Why Smart Firms Should Experiment

My AFR op-ed today is on the use of experiments by businesses as a means of improving productivity. Full text over the fold.

Joshua Gans and I have been contemplating running a conference on this topic in 2011, so if there are Australian academics out there who are doing this kind of work, please drop me a line.


Have a Go, It Can Bear Fruit, Australian Financial Review, 5 January 2010

In a British fruit farm, the seeds of a quiet revolution in management are being sown. While the neighbouring farms go about business as usual, the owners of this establishment see each season as a chance to run experiments aimed at raising productivity and profits.

Halfway through the 2002 season, the pay scheme for all fruit pickers was suddenly switched from one that paid workers relative to their co-workers, to a scheme that paid based on absolute performance. Midway through the 2004 season, a managerial bonus scheme was introduced, allowing the best managers to increase their earnings by around 25 percent. In 2005, fruit-picking teams found their performance charted at the end of each day. Later that season, the best teams began winning bonus payments.

In each case, the changes were engineered by a trio of young economists: Oriana Bandiera, Iwan Barankay and Imran Rasul. Tracking the changes, they found that rewarding workers for absolute performance raises productivity by a whopping 50 percent, and that managerial bonus schemes boost output by 21 percent. Yet when it comes to motivating teams, it turns out that charting performance is a bad idea (the underperformers become discouraged), while rewarding performance is effective (the best try even harder).

To see how radical this approach is, imagine how the typical firm would have approached the same problem. Step one: hire a team of well-dressed management consultants. Step two: pay six-figure bill. Step three: hold nose and implement recommendations.

Business experiments turn this logic on its head. Rather than paying for expert wisdom, the experimental approach says ‘try it and see’. Where the traditional literature is passionate and confident (with titles like ‘Straight from the Gut’ and ‘Re-engineering the Corporation’), the experimental approach is modest and open-minded (with titles like ‘Team Incentives: Evidence from a Field Experiment’).

While the traditional approach listens to the consultant or the highest paid person in the room, the experimental approach listens to the data. Just as new drugs are evaluated using randomised trials, business experiments aim to bring both simplicity and scientific rigour to evaluating the impact of new strategies on the bottom line.

Perhaps the aspect of business that has made the best use of experiments is advertising. In formulating advertising campaigns, firms are increasingly recognising the limits to what can be learned from experts and focus groups. Rather than pick between two possible campaigns, the smarter approach may be to find test markets, and try both. This is most easily done in direct mail or online campaigns, where it is possible to get customer-level variation (indeed, Google’s AdWords software even allows small-time users to run such marketing experiments). But by exploiting variation across chain stores, or in regional markets, firms have found ways to test other marketing strategies too.

Business experiments have also helped to quantify the value of a strong track record. In a series of experiments on eBay, Paul Resnick and co-authors compare the prices of batches of vintage postcards sold by an established seller with a strong reputation versus a first-time seller. They find that first-time sellers receive prices that are 8 percent lower, illustrating that reputation on eBay carries considerable financial value.

Experiments can even help firms decide how to set prices. Economist Steven Levitt gives the example of his work with a direct-marketing travel service, which experimented with randomly changing prices on its mailings, and found that demand barely fell when it raised prices. Not surprisingly, the company now charges considerably more than it did in the past.

Randomisation, Levitt argues, ‘cuts through the complexity’. He points out that ‘All you have to do is flip a coin and put people into treatment and control groups. In the end, you just have to be able to add it up and say, “Am I selling more or am I making more profit in the group that got the treatment versus the group that got the control?”’.

From fruit farms to travel agencies, experiments represent a powerful challenge to business as usual, who are increasingly embracing the opportunity to get better feedback on everything from their human resources policies to their pricing strategies. Why settle for grizzled wisdom when you can try it and see?

Andrew Leigh is an economist in the Research School of Economics at the Australian National University.