|
Can people offer their experience on what is the best way to bill a data-mining project. Hourly is one option. OK. Other options involve (a) pricing for a spec'ed algorithm to be implemented, or (b) pricing for a "satisfactory" output data. But both (a) and (b) are tricky. For (a), usually the algorithm changes as you work, so it is infeasible to spec it ahead of time. For (b), with data mining jobs there is no good (and practical) way to measure result quality (in fact, often result quality is based on eye-balling). So (a) and (b) don't seem like good ways to go. Thoughts? |
|
Whatever you do, build in some mechanism for charging for increases in time or other resources which arrive at the client's discretion. Setting a fixed price per project seems like a good idea, until you have a client who wants to add "one more thing", etc. Also, get any travel expenses paid up front. I know consultants who've been forced to pay theirown way home when on-site with a client when thigns went sour. |
|
I'll say hourly is probably your best option, regardless of your algorithm, if you are good enough, changing the implementation of basic "out of the shelf" algorithms shouldn't be that difficult, (I'm thinking on easily available algorithms, SVM, HMM, where the only thing that change is how you handle the data) I agree, have your travel expenses paid ahead. And if you are charging by hour, just cherge what is fair. It depends on your expertise and your experience. |
