Bring one real problem to the session
The people who get the most out of an hour with us arrive with something specific they are stuck on. Not a topic. A problem, with its detail...
Read moreAnyone can state an accuracy figure. The work is in showing how it was measured, on what, and by whom.
The ventures the group holds have one habit in common: they sell a check, not a claim. It is the thread running through these sessions too, whatever the subject on the day.
A claim is a number in a deck. A check is the method behind the number: what was measured, on which data, against what baseline, and what happens to the figure when the data changes. A vendor who cannot describe the method is not hiding a trade secret. They are usually reporting a number somebody else produced.
In a session this turns into a short list of questions you can ask anybody, including us. Where did the test set come from. Who labelled it. What does the figure fall to on documents you supplied. What is the failure mode when it is wrong, and who notices first.
None of those questions require you to be technical. They require the person answering to have done the work.
22
Sep
The people who get the most out of an hour with us arrive with something specific they are stuck on. Not a topic. A problem, with its detail...
Read more
16
Sep
Most of what goes wrong at the start of a seminar is fixable beforehand, and takes about ten minutes.
Read moreThis seminar has finished. The next one is announced here.