> Hypothetically we *are* a "software company". Oh and hey, we do have
> planning meetings. Trouble is, the estimation process is roughly
> "everybody picks a number, and we see if the numbers look similar". That
> still doesn't seem very scientific. Then again, when you're asked to
> produce a new feature and you have no idea what libraries might be
> available to help you or what OS-specific hiccups will trip you up, what
> more can you do?

You iron out those issues in the "investigation" phase of the project, 
or whatever you want to call it. Here we call it Phase 3, which is the 
engineers scratching the surface, doing some rudimentary testing etc, 
and providing very specific deliverables at the end, which includes a 
detailed time schedule. Phase 4 is then where all the detail work gets 
done, of course you'll hit hiccups, but hopefully during phase 3 you 
minimised the risk of any big ones and identified the things most likely 
to trip you up.

Although we do products that consist of mechanical, electrical and 
software elements, I assume a similar development process is in use for 
software-only products. I can imagine phase 3 being selecting which 
libraries to use and doing some basic testing on them, identifying any 
high-risk areas and doing some testing on those, planning a framework 
for the product etc.

Post a reply to this message