Three years ago, Odeo was a struggling startup on a path to nowhere. Odeo's core offering -- a set of tools for users to create, record, and share podcasts -- was facing serious competition from Apple and other heavyweights. The management team made a radical decision to "reboot" the company, and Twitter was born.

As I read the Twitter story, narrated eloquently by Dom Sagolla, I can't help but look back over the many startups that I've been associated with over the past 12 years.  In my various roles as a founder, an investor, a board member, and an advisor to startups in Silicon Valley, I'm constantly fascinated by the mechanics of reinvention. Which approaches to reinvention succeed and which ones fail?

Startups flounder for countless reasons. Perhaps the market opportunity is not as big as imagined, or perhaps there is a mismatch between the technology and the market. Maybe the world changed in some significant way, invalidating the key assumptions on which the startup was based. For example, an established company such as Google or Microsoft might enter the market. Or perhaps the deepest recession in recent history dried up demand for the original product or service. In these cases, the founders and management team have to ask themselves the question: should we push ahead, assuming superior execution will win the day against long odds? Or should we change what we're doing?

Companies that decide to reinvent need to acknowledge the bad news first: most startups fail, even the reincarnated ones.  Those are just the odds. The good news is that certain approaches to reinvention work better than others, and companies can increase their chance of success by carefully calculating their reboot strategy.

Core components that help a reboot strategy

Every technology startup has four core components: team, technology/product, market, and business model. Rebooting involves changing at least one of these components, while leaving the other factors unchanged. Let us look at each component in turn:

  1. Team. Reinvention usually leads to changes in the team. To qualify as a reboot rather than an entirely new company, however, there must be at least part of the team -- and usually at least one of the founding members -- who continues to remain with the company through the transition. In my experience, one model that usually does not work is when VC investors replace the entire founding team with new management. I've never seen a startup with none of its founders remaining succeed.
  2. Market. Many startups try the most tempting option: to keep the same technology/product and look for a new market.  After all, the investment in product development has already been made.  Unfortunately, while this approach seems the most logical, it is also the least likely to succeed. Why? The hardest part of a startup is understanding the requirements of the market, not building the product. After the dot-com bust in 2000, many consumer internet startups tried to reinvent themselves as enterprise technology providers (remember Chemdex?). The startup junkyard is littered with the carcasses of dot-coms that took this route and failed.
  3. Business Model. A very attractive strategy is to keep the same product and market, but change the business model. In my experience, this is the most likely option to succeed. For example, enterprise software companies can reinvent themselves by open-sourcing their software and providing consulting services, or a premium version. A software vendor can reboot as a software as a service (SaaS) provider on the Web. Consumer websites can move to a subscription model from an advertising model, or vice versa
  4. Product. Another smart reinvention approach is to addressing the same market (or a closely related one), but change the product or the business model. This option works best when the market need is real, but the product does not adequately address the opportunity. I've found that the key to success is to throw away the old product completely and start from scratch, using the hard-won learnings about the market acquired from the first iteration. In some cases, it makes sense to move the old product to "maintenance mode" and reassign the bulk of the team to developing the new product. 

I've applied this particular model of reinvention to both companies where I have been a founder -- Junglee in 1997 and Kosmix ten years later, in 2007.

We started Junglee in 1996 to create virtual databases that integrated data from multiple websites. Although we had some initial success, we quickly realized that the architecture of our first product limited our ability to deal with rapidly-changing information, a key success factor in certain markets. We completely rebuilt the product from scratch in 1997, and created the world's first comparison shopping service.  This service was enormously popular and led to Junglee's acquisition by in 1998.

We introduced Kosmix as a vertical search engine, initially in the health sector.  Our idea was to find a better way to help users understand open-ended queries such as "diabetes," which have no single right answer; that is, explore topics rather than find the needle in the haystack.

We'd planned to take a vertical-by-vertical strategy, launching sites named RightHealth, RightAutos and RightTrips. Very soon, however, we realized that the vertical approach carries severe limitations, because it's hard for consumers to remember to go to different sites for different topics of interest. We decided to rewrite the product from scratch, and in December 2008 we relaunched as a horizontal site.  Kosmix lets you explore any topic and gives you a 360 degree view of anything than interests you -- including videos, images, community discussions, expert commentary, news, and much more. This transition from vertical to horizontal was much harder than it sounds; it required us to rewrite our technology from scratch. But we did it because of our passionate belief that the problem is real and the market opportunity is vast.

While most startup reboots involve rethinking only one or two of the four core components, in some rare cases it makes sense to go the whole hog. Sometimes it pays to be bold: go after an entirely new market opportunity, create a new product, find a new business model, and make large-scale team changes. This approach is fraught with risk; but there have been a couple of spectacular successes. One clear example is Twitter. Another is Twitter's cousin SMS GupShup, a similar service in India. SMS GupShup was born as Webaroo, a company that wanted to create offline copies of large parts of the web so you could browse while offline. A couple of engineers there launched the SMS GupShup service as a lark and it took off; once the management team saw the traction of GupShup, they re-oriented the company around the new idea.

Some startups are born great: the right team starts with the right idea at the right time, and the rest is history. Some have greatness thrust upon them: the right conjunction of market forces propels an unlikely startup to dizzying heights. Other startups, not so lucky as those in the first two categories, need to earn their greatness. And sometimes that requires a reboot.

Anand Rajaraman is co-founder of Kosmix with consumer properties, and  He also sits on the board of several technology companies and is a consulting faculty member at the Computer Science Department of Stanford University. His latest thoughts and discussions can be found at