No crystal ball? No big deal!

Svenja Sommer, Professor of Operations Management and Information Technology - October 15th, 2008
crystal ball

Key ideas:

• Divide projects into parts and only use traditional planning for areas where all is clear.

• When risks cannot be pinpointed, approaches like trial & error learning or parallel trials will increase the chances of project success.

• Novel initiatives frequently require such approaches as well as atypical monitoring schemes. 

Svenja Sommer ©HEC Paris

In 2008, Svenja Sommer joined HEC Paris as professor of operations management and information technology. Svenja Sommer holds a PhD in management from INSEAD, and prior to joining (...)

See CV

How do you develop a good plan when you don’t know what tomorrow may bring? This is the challenge for managers in start-up companies or those involved in brand new projects. Human nature is such that people tend to keep making detailed plans even when they lack knowledge about certain aspects of their project—like how a new technology will translate on a large scale or what kind of support customers will need for a brand new product—but sticking to such plans often brings everyone down. Svenja Sommer and her colleagues have explored how start-ups use two principal alternatives to traditional risk planning to deal with unknowns more effectively, revealing where managers should use each in order to reach the desired outcome: success!


By definition, new projects contain at least a minimal amount of mystery. When you start developing a new technology, enter a new market, or decide to expand your customer pool, you have few precedents to rely on to design your project. And what further complicates the situation is that you do not always know what it is that you don’t know! Svenja Sommer suggests thinking in terms of “knowledge gaps” and warns managers to beware of detailed planning when they are significant.

• When there are large knowledge gaps, detailed planning is most likely a waste of time, and even worse, it tends to generate a counterproductive, false sense of security. For example, based on the small-scale success of a new technology, a detailed business plan was developed for a new plant. But the large scale was an entirely different story, and stubborn reference to “the plan” resulted in ignored problems, project delays, budget overruns, and ultimately a failed venture.

• Novel endeavours must be recognized to require on the- job learning, at least in certain areas of the project. There is no recipe, so you have to test options and integrate new information as it is uncovered rather than make hasty, assumption-based decisions.


When managers and projects teams have to identify the areas in which they are knowledgeable and those which present “knowledge gaps”.

• Break the project into parts and review each of them. Do you or do you not know all that you need to about pertinent technologies? Markets? Customer needs?

• For areas of the project that are clearly understood, use traditional planning methods, identifying potential risks and making contingency plans. In areas where you lack knowledge and cannot pinpoint specific risks, you cannot plan. These areas should be approached using one of two schemes: selectionism or trial & error learning.

• Selectionism or parallel trials means developing several options, seeing how each one works, and choosing the best of the lot.

• When an area of a project is interconnected with others, and changes in it provoke changes elsewhere, launching several versions of a plan will enable you to discover the most successful. Let’s say you need to choose an external development partner among an unfamiliar group. Give several potential collaborators the same assignment and see who is able to best meet your needs.

• This method must not be applied to an entire project; it would be far too expensive and time consuming. Rather, use this method for sub-projects. Exploring alternatives can make the difference between success and failure, but remember that parallel trials must be kept alive long enough until unknowns have been resolved. Technology-related exploration may need to be continued only up to the prototype stage, while unknown customer needs may mean pursuing parallel trials until customers actually come into contact with various versions of your product.

• This method might seem costly because money is spent on things that are ultimately not used. However it can drastically increase the odds of success.

• Trial & Error Learning is a means to fill specific knowledge gaps. Once you have obtained new information, you can adjust your existing plan, filling in the blanks.

• Unlike selectionism, trial & error learning focuses on reducing knowledge gaps but will not produce a complete solution. It will contribute specific information for improving your plan and even revising your target, allowing the venture to evolve toward success. Many breakthrough technologies, including Apple and HP’s PDAs, were successfully developed using this method.

• You must be flexible and repeatedly adjust your activities and targets. For example, to learn by trial &error, software programmers have customers try out a new program, collect their feedback, revise the program, and repeat the process as needed.

• This approach can be time-consuming, but in initiatives with many unknowns but few connections between problem areas, it can largely increase the chances of success.


It is important to determine the right management approach early on, as monitoring methods must be chosen accordingly. In traditionally organized projects, managers monitor project advances and success by tracking predefined milestones, but this is inappropriate for projects presenting many unknowns. Process is the watchword here, not product. Incentives and rewards should be based on observable effort and action rather than targets.

• With selectionism, keep track of performance in various trials. This will enable you to decide which trial(s) to stop and which to pursue.

• With trial & error learning, monitor learning cycles. Ask questions like: What have we learned? Are we resolving uncertainty? What do we base decisions on? Are we analyzing observations about the project? The answers will bring you forward.

Based on an interview with Svenja Sommer and her article “Managing Complexity and Unforeseeable Uncertainty in Startup Companies: An Empirical Study” (co-authored with Christoph Loch and Jing Dong, Organization Science , published online ahead of print, July 25, 2008) 


• At the start of a novel initiative, thoroughly investigate the various factors involved to clarify what can and cannot be planned. Get the hierarchy involved early on, make them aware of challenges, and ensure everyone knows that the goal is progress, not a specific target.

• Choose the right management scheme for each part of the project: planning where there are few unknowns; trial &error learning where unforeseen uncertainty (unknowns) is high; selectionism where both unknowns and complexity (interrelatedness) are considerable.

• Monitor effort and action. Recognize the ability to adjust to difficulties and reward performance, even when it is low, when outside circumstances are difficult.

• Ensure close communication with field managers. They are often in the best position to discover new opportunities. 


With the help of the Shanghai Venture Capital Association, Svenja Sommer and her colleagues surveyed and interviewed senior managers in 58 start-up companies in Shanghai in a variety of industries. They compared the effectiveness of traditional risk management, selectionism, and trial & error learning in business ventures dominated by unknowns. Their study reveals where each method is most appropriate.