CrowdWorx interview with Carol Gebert: Prediction Markets in the Pharmaceuticals Industry

Carol Gebert

Carol, you are one of the pioneers in using Prediction Markets in business. Tell us how you have applied Prediction Markets in a corporate environment?

We setup Prediction Markets at Eli Lilly to predict enrollment rates into a clinical trial that was taking too long to enroll its patients.  That put the entire drug program behind schedule, so Lilly wanted to know when the enrollment would be complete  The participants in the market were meant to be the clinical sites that were enrolling patients and performing the clinical trials.  However, ten years ago, few of these clinics provided easy access to websites.  We were warned that many of the clinics were low-tech, with little more than a fax machine.   Under the rules of the time, we could offer only token prizes such as coffee mugs.  Rather predictably, participation in the market was too low to allow liquidity, so the Prediction Market failed in its goals.

We were also involved in setting up a market aimed at the public, to predict market share in several drug markets.  That market was more successful in attracting traders, but we had to wait for one year to declare the outcome, and during that time, participation dropped off.  However, the predictions were pretty accurate.  So it was declared successful.

What were the biggest challenges in using Enterprise Prediction Markets in the pharmaceuticals industry?

One big challenge is the long timeline of development in the pharmaceuticals industry. Most traders in a Prediction Market would be reluctant to wait for more than a year for their bets to be settled. There are ways to overcome this, e.g. using the average price for giving interim payouts or splitting up a long-term project into short-term milestones and forecast the probability of meeting these goals.

Another challenge is the fact that in most companies, not just in the pharmaceuticals industry, forecasting means simply planning or goal setting. Somebody sets an optimistic goal and no one is interested in an actual forecast..  Perhaps Prediction Markets are better suited to industries where forecasts are truly actionable, rather than simple forewarnings.  When HP ran its pioneering Prediction Markets, it wanted to know how much manufacturing it should commit to, for each printer version, before the next model was released.  Thus, forecasts of sales were actionable and better forecasts resulted in cost savings.  But for some business applications, real-time forecasts do not result in more cost savings, compared to simple projections from historical numbers.  In these situations, forecasts are merely used to confirm that operations is being run according to plan.

If you could start out again today in pharmaceuticals, which areas would you try for Prediction Markets?

For a pharmaceutical corporation the most money to be saved is in the last phases of the clinical trials. Trials themselves cost many tens of millions of dollars to conduct, and each day of delay loses $1million in lost patent time. So potentially, the forecasting of outcomes is valuable.  However, these clinical trials are deliberately double-blinded to ensure purity of the data, so forecasting the outcome of such experiments undermines the process..

Therefore, I’d probably go into the earlier, discovery stage.. Here the savings are not as large but still worthwhile.  For example, a molecule showing promising activity during discovery may still run into solubility/dosing or metabolic problems and be shelved in preclinical or early clinical testing after spending perhaps $10million.  So predicting physical and pharmacologic properties of a molecular series when it is still in its medicinal chemistry cycles would be valuable, actionable and have some results known in under a year.  The traders in such a market should be the medicinal chemistry team, the metabolism, pharmacology and toxicology teams, and perhaps even the process development team.

I can see licensing professionals appreciating Prediction Markets, too.  The licensing departments of big pharmaceutical companies are constantly searching for good drug candidates partially developed by smaller companies.  They typically review dozens of these each year.  All licensors paint rosy cases for their particular molecule, so it is hard for the big pharma professional to pick which drug candidates really are worth the licensing fee and remaining development costs.  Licensors typically approach big pharma licensing offices just before a major milestone is due.  Maybe a Prediction Market designed around those promises would help the licensing team make purchasing decisions.

As for areas beyond my home turf, I think toys would be high on my list.  Toy makers get a lion’s share of their income seasonally, in December.  Their product successes are dependent on personal taste (i.e. fashion) to a degree that technology products are not.  And they need to commit to manufacturing many months in advance. That sounds like a clear need for good forecasting.

PMs could also make valuable tools for supply-chain cost forecasting where supplies tend to have irregular availability or volatile pricing.  For example, I think chocolate makers might like to better predict pricing for cocoa, sugar and cream.  Similarly, electronics manufacturers might like to have better forecasting of rare earth supplies.

You have been in the Prediction Markets industry already 10 years ago. In your opinion, what are the Top 3 developments for Prediction Markets since then?

One of the major obstacles in the past was the way Prediction Markets were accessed: Online via the Internet, which at that time meant sitting at a dedicated desk and going through several steps to even access a desired page.  And all pages needed to be visited separately, with conscious intent.  Nowadays, browsing habits have changed and people are getting used to visiting Intranet portals and consolidated social sites on a routine basis. Mobile phones and mobile apps have opened up another access channel which was not available 10 year ago.   Everywhere I look, I see people wanting to play with their mobile devices.  I think Prediction Markets would make a perfect mobile app for expert-level professionals.  For consumers, Prediction Markets embedded into Facebook would be the way I would go.

Ten years ago many of the things we wanted to forecast were not measured so we could not know the actual outcome, even if there was an objective event which could have been measured. Nowadays, much more data is collected and readily available.  So the outcomes vital to defining the stocks or bets in a Prediction Markets are available in a way that restricted us before.

Last but not least: Until not long ago Prediction Markets have been banned in the U.S. under internet gambling regulation. U. S. companies were very hesitant to apply Prediction Markets, even when there was just play money.  Around the time I was talking to Lilly, DARPA funded a PM pilot to predict terror threats, but quickly canceled the project when two politicians misconstrued the effort, and denounced it publicly.  In the US, if your words can be twisted and misconstrued, then they will be.  Pharmaceutical companies are very nervous of that public reflex.

But in 2006 Professor Robin Hanson, of George Mason University successfully championed an effort to exempt Prediction Markets from this regulation, on the rationale that it is a business tool. That was a big step forward for PMs.  Before this change, PMs were severely limited in the types of incentives offered.  The Hollywood Stock Exchange demonstrates that people will play for no material reward if there is entertainment value to participating, but I suspect this is true only for topics of special interest, like movies.  For most business topics, a significant reward is required to attract and maintain participation.  From first-hand experience, I can tell you that you cannot motivate professional participation with coffee mug prizes, alone.

Thank you Carol for sharing your insights and thoughts with us!