Should one always mention numbers/percents for algorithm results in an interview/cover letter?

Question

If I am referring to my implementation / use of a machine learning algorithm for some task that is relevant to the job, should I always be ready to give information about the performance of the classifier, e.g. its accuracy, recall, F-measure, etc.? For a non-entry level job, I am guessing this would be rather important.

But for an entry level job, is it ok to report not so great results or should I omit the results completely and wait for the interviewer to ask me if it comes up?

I understand it may depend on the task, e.g. a 50% 'accuracy' on generating live product reviews that get voted most helpful by humans (who don't know they are machine generated) would probably turn heads, but on a simple binary classification task this number would be almost comical.

These websites suggest numbers should generally be included in a resume/cover letter to highlight achievement, but what about when they suggest not so great achievement?

(1) http://career-advice.monster.com/resumes-cover-letters/resume-writing-tips/numbers-to-highlight-accomplishments/article.aspx

(2) http://www.smartrecruiters.com/blog/the-right-numbers-in-a-resume/

(3) http://careerrocketeer.com/2011/11/4-ways-numbers-can-make-your-resume-effective.html

Answer 1

Should one always mention numbers/percents for algorithm results in an interview/cover letter?

No, definitely not. Numbers may be helpful, but you shouldn't put numbers blindly everywhere - only where they really help convey the scale of your achievements and help you stand out from the crowd in a positive way.

As you mention, the measurements of an algorithm may be hard to interpret without the right context, and too detailed for a CV / cover letter. So use only numbers which are easy to interpret and unambiguous, even for people without a strong technical background - remember that the first readers of your cover letter and CV will be most likely HR / recruiters, not developers or scientists. And you must pass through their filters in order to get to meet a technical person in an interview, where - if asked - you can then give more details about the algorithm(s) you developed.

The referred articles suggest you use numbers which are measured in, or are easily translated to, quantities making sense to prospective employers, like money, time etc. In other words, strive to quantify the impact you made on the business, rather than just quoting raw technical data. So if you can calculate or dig out concrete numbers like

• product reviews generated by the algorithm I developed received 5% more "helpful" votes by humans compared to the earlier implementation
• the binary classification algorithm I developed ran 25% faster / was 10% more accurate / used 15% less resources compared to our original target

then by all means quote them.