Faster learning cycles: the gains are exponential, not additive

Summary : Want to improve RoR (Return on R&D)? Faster learning cycles are probably the biggest reason why the best development teams regularly out-innovate their competitors. Faster learning cycles are far more important than money or resources and this is especially true when talent is equal.

Want to improve RoR (Return on R&D)? Faster learning cycles are probably the biggest reason why the best development teams regularly out-innovate their competitors. Faster learning cycles are far more important than money or resources and this is especially true when talent is equal.

The reason is due to mental disengagement of knowledge workers. They are not books, where returning to a bookmark brings them right back to where they left off. The longer they are disengaged, waiting for results, the more catch-up they have to do to recover their last position. 

There are three important time zones measure in each learning cycle: Time to Question, Time to Answer, and Time to Decision (TtQ, TtA, and TtD). Decisions invariably lead to a new question and the process starts over. Management has a role to play in minimizing all three via both the team and resources.

The biggest resource role management can play is in Time to Answer. It’s an area where I’ve seen a lot of once great IDMs fall out in recent years to become fablite. Capital costs often get the blame because no one wants to take responsibility for poor RoR. But I watched them set themselves up for failure as they sought to save costs by either centralizing or outsourcing their analytical capabilities just as the era of new materials dawned. The result was an average TtA shift from half-to-a-full day out to one-to-two weeks. This put billions of development costs at risk to save less than a hundred million. Here’s how it played out:

A lot of development is serial, so you must have an answer to decide on the next step. If one company’s TtA is 1 day and another’s is two weeks at a minimum, the latter is falling behind by 13 days with each learning cycle in the worst case. Moreover, this doesn’t even include the time lost as researchers work to catch back up to where they were mentally. So at a minimum the slow TtA company was behind by a full quarter after only 7 days had passed; a full year after a month and a full node after two months.

Now it wasn’t quite as clean as the math indicates, but the slow TtA companies did systematically slip behind as node development progressed beyond 130nm. By 65nm, most had given up completely. 

By G Dan Hutcheson                                     Copyright © All rights reserved.

You may like this also:

Access to and use of this Website is subject to VLSI's Terms of Use (including Copyright Policy & Claims) and Privacy Policy. By accessing or using this Website you agree to VLSI's Terms of Use (including Copyright Policy & Claims) and Privacy Policy.

Copyright © 2021 VLSI Research Inc. All rights reserved.