These days, selling Business-to-Business can seem like risky business. CSO Insights' latest Sales Performance Report notes that in 2009 the percentage of reps making quota dropped to 51.8% from 58.8% a year earlier. The response? Most firms have increased quotas for 2010, seemingly hoping that an improved economy will be the tide that raises performance. In CSO Insights' view, it's a risky approach prone to fail unless other things occur.

The right interventions to improve sales productivity seem equally uncertain. New sales compensation models? Re-organizations of sales operations? New technologies? Sales training? One of the challenges to buyers is that these choices tend to be silo-ed silver bullets which, when shot, end up being advocated most by those who've shot them. There's no independent proof that the chosen tactic had more impact than any other tactic would have had.

So, what's one to do? In my view, there's value in thinking like a successful start-up. For start-ups, uncertainty is a given; the trick is drive down the risks arising from uncertainty. From Eric Ries's thoughtful post on balancing process and innovation, these best practices emerge:

• ACCELERATE LEARNING.
  The speed at which a startup can learn is its competitive advantage and the defining factor in its success. Any change that accelerates learning is a win, and everything else is waste
• MEASURE CUSTOMER-VALIDATING THINGS.
  Startups measure their firms' impacts based on customers' experiences. They do so in ways that create a common measurement language for all members of their team. This eliminates risks that different parts of a firm's team "learn" in their own private reality. It lets teams, when facing difficult choices, come together and make informed, fact-based, decisions.
• TRIGGER LEARNING VIA FEEDBACK.
  The faster feedback on customer experiences arrives, the better. The feedback loop between taking an action and seeing the results should be as short as possible. To provoke learning, simplicity matters. Reports that bury users in data won't affect behaviors. Simple ones will. It's not about how much data you can store. It's about how much learning you can provoke.