Wednesday, January 02, 2008

New Year's Resolution 2008: Create Infrastructure Strategy

It’s January 2 and so, in case you haven’t already settled on your New Year’s resolutions, we’d like to suggest one: figure out your infrastructure strategy.
A good place to start is with the number of people you need to do the job you’re in business to do. Since you will always have failures, you need a minimum number of programs to achieve a minimum level of return. Once you’ve figured that out, staff to that number of programs.

The problem is determining when a program achieves its minimum level of success. To answer that question, we’d ask another: for what are you looking to get paid?

As it stands today, companies can get paid -- pretty well, too – for doing a variety of jobs: creating INDs, for example (like Plexxikon); or getting a compound through proof-of-concept (like Exelixis or Vertex); or taking a product from Phase II to approval (like New River). It is by no means always necessary to do all of these jobs -- and therefore no need to staff them.

Think about the drug business like professional sports: the same guys who play in the NBA aren’t ever likely to qualify for Wimbledon; and none of them are likely to end up playing for the New England Patriots or worrying Tiger Woods. The physical requirements are different from sport to sport. And where they’re not, the training and focus required to perform at a high level in any one sport usually precludes excelling simultaneously at another.

The real question is to figure out what game you’re playing – and which team you need to play it. Presumably, you’ll need different players for the IND game than if you play the Phase III game. And you’ll need different numbers of players for each game.

Most Big Pharmas, thanks to tradition, feel they need to play all the games and therefore staff themselves to compete in each. But in fact they have traditionally played only one game – the commercial game. The only way a Big Pharma wins is by launching a product successfully (remember: what you get paid for doing determines which game you’re playing).

In terms of infrastructure, therefore, Big Pharma is playing at a huge disadvantage. The math goes something like this: to get one discovery compound to Phase I, you need to start with about eleven programs – and by the time you’ve gotten your one successful compound into Phase I, you’ll have spent $23 million in cash, without adding any capital or opportunity costs. (See a more in-depth analysis here). Infrastructure: 50-75 people.

On the other hand, to be relatively sure that you’ll get one discovery program all the way to market, you probably need to start with more than 100 programs – or a discovery cash outlay of more than $200 million. Rough estimate: 500 – 750 people. That math works, incidentally, only if discovery infrastructure is scaleable – that is, if ten times the people can actually do ten times the work. Given discovery’s requirements for rapid feedback and a certain anti-bureaucratic creativity, it seems more likely that at some point, the larger the discovery organization, the less productive it is.

In any event, in the worst case, if you’re making your money at Phase I, you need just one-tenth the discovery infrastructure you need if you’re not getting paid until a product reaches the market.

Same logic with development. If you’re getting well paid by a licensee or acquirer for moving a compound from Phase I to Phase III – not from Phase I to the market – you need fewer compounds to succeed because you don’t have any FDA or launch risk in your business. Fewer compounds, fewer employees. Nor do you need the same kind of infrastructure our discovery-focused player required. You need a different kind of infrastructure for finding new compounds to develop.

For this logic to work, you need to get paid, on a relative basis, about as well for doing a more focused job as for doing the traditional soup-to-nuts work of the traditional Big Pharma. And in fact you can. Phase II compounds now generate upfront licensing fees of $70 million and up – with royalties in the high teens or higher (and there is an increasingly competitive marketplace of companies willing to buy out those royalties, in case you want your returns right away). Domain Associates has done well for itself in-licensing a Phase I compound or two, wrapping a company around it, and hiring no more than a dozen people to manage the compounds’ development, largely through a network of CROs – then selling off the result at huge profits to J&J (Peninsula), or Forest Laboratories (Cerexa), or Merck (NovaCardia).

You can argue how repeatable those models are and therefore how much additional infrastructure you might ultimately need. Celtic Therapeutics – the new follow-on private equity fund building on PE predecessor Celtic Pharma (see here and here, for more) – figures that Domain’s math of onesies and twosies won’t work consistently. Given standard clinical failure rates, Celtic is thus amassing a larger portfolio of projects for which they’ll need a larger number of managers. But because Celtic is focusing only on later-stage development, it will still only need a relative handful of workers (20 projects = about 65 people, says Celtic managing director Stephen Evans-Freke).

We admit that we are oversimplifying the infrastructure debate to make a point. There will be companies who do multiple jobs and will require multiple infrastructures. A Big Pharma might be able to get paid for Phase I to Phase II primary-care development (e.g., Bristol-Myers Squibb’s deals with AstraZeneca and Pfizer) and simultaneously get paid for launching new specialty medicines… or vice-versa. There will be Big Pharmas who can create INDs and get paid – in cash or kind – for distributing them to development partners (Lilly is doing something like this with its Nicholas Piramal relationship).

But the key will be figuring out which jobs you can consistently get paid for. And then to stop doing the jobs – and thus hanging on to the related infrastructures –you’re not getting paid for.

No comments: