18

I see a lot of Excel questions, but also ones involving command lines, batch files, and other similar scripts, that entail some mental challenge. I'm not referring to ones where a novice needs some basics and the solution is straightforward.

Users will post what looks like a good answer, reflecting a significant time investment to figure out the solution and write it up. Often, the solution is obtuse, or long and complex, not something you can glance at and recognize that it will obviously work. The author may include explanations of how and why the components of the formulas should work.

The Problem

But people sometimes don't bother to actually test it, or fully test it against the range of conditions in the question. Even if they test it, they don't go the last mile to include a demonstration of the results, like a screenshot of the output, so there's no way to know the extent of any testing. In some cases, the sample output could also add value by showing exactly what the results will look like, which can make it easier to understand the formulas.

Without that, I can't tell if it's correct without investing the time to recreate the spreadsheet. Sometimes just dissecting the formulas can miss the same error as the author did on something untested that looks superficially like it should probably work. There have been countless times where somebody has posted a solution that looked great, then the OP or another reader commented that it didn't work, or the author returned later and changed it because they discovered that it didn't work.

The Ramifications

So I don't like to upvote the answer if I don't know that it actually works. If I happen to catch it later, and the OP has accepted it or commented that it worked, I can take that as evidence. But at least for me, a lot of contributors are leaving easy upvotes on the table by not "finishing" the answer.

I suppose I could leave a comment. But then I would look like the only dummy for whom the result wasn't obvious. And the author may have discarded their work by the time I comment, so going back to reconstruct it may seem like too much effort at that point.

My guess is that the author is focused on satisfying the OP, and expects that the OP will be doing the verification. But that misses the community at large. Good answers should address the audience beyond the OP; "proven" solutions.

Questions

It could be that my expectation is excessive relative to the majority of users, so my questions:

  • In terms of the potential scope of lost rep opportunity for contributors, do people besides me not upvote answers that they can't easily verify?
  • Do others agree that on non-obvious or non-trivial answers, it's important to include the results as proof?
  • If so, is there something we can do to educate and encourage authors to do that?
  • 1
    I don't like to upvote answers unless the OP has also indicated that the answer solves the problem, but I'll upvote occasionally when the answer includes "above and beyond" explanation/notes that teach me something, whether or not they solve the original problem. If I answer a question with code that's a bit arcane, I try to include add'l explanation but if a fix needs just a correction to a user's existing code, I don't usually do that; reputation points aren't a big motivator for me, so leaving a few on the table is a non-issue. – Steve Rindsberg Dec 25 '16 at 17:48
9

Users who ask questions that require complex scripts/formulae should have included some example input and the desired output. I find it highly likely that a user who went to the trouble of writing a gigantic solution tested it with at least the case(s) provided in the question.

Of course, there's no guarantee they did test it thoroughly. (Depending on how paranoid you are, you might want to test the answer yourself either way - a screenshot or text spew could be doctored.) Therefore, I really like answers that explain how the solution works. An Excel formula with a bunch of nested functions would optimally be accompanied by at least a note of which functions do what. That way, if you're familiar with most of the individual pieces, you could check that they're all used in a way that makes sense. Even better, people who don't know the pieces can learn and maybe apply them in other places.

This might not apply quite so much to Excel, but for scripts that generate files or objects, example output is helpful but not required. If it's provided, everyone can see right there which sections/properties are what they need.

Ultimately, you're free to vote as you like. I also tend to withhold my vote (up or down) until I can verify that the answer's information is correct. Voting behavior conveniently incentivizes the things you vote up - you're marking answers with demonstrations useful while not boosting answers you find lacking.

  • Just to be clear, I agree with everything here. My request for results wasn't meant to imply that those replace the explanations and other aspects you describe that make a good answer. I meant that it should supplement it. It sounds like we might assign different priorities to including results, but that's why there's both vanilla and chocolate ice cream. :-) – fixer1234 Dec 24 '16 at 1:26
2

Answers which present a valid approach to solve the OP's problem are good answers by my standards, even if they contain mistakes. While I certainly appreciate answers which are tested to perfection, I never expect code found on SE to be 100% plug&play, and I always test it before using.

How you set standards for upvotes / downvotes is very subjective, you might as well refrain from upvoting a 100% correct answer which is riddled with spelling mistakes and slang.

  • A few special cases: 1) Find a similar problem or duplicate question. The reader expects to have to adapt the solution. 2) A novel solution with mistakes. I might upvote the methodology if it is especially useful in general, but comment that it contains mistakes. Otherwise, the majority of these questions are posted by people with limited subject knowledge, or they wouldn't have the question. Same for future searchers. So posting an answer that doesn't work isn't much help to them; it doesn't solve the problem. An idea that needs to be developed into a working solution is more of a comment. – fixer1234 Dec 31 '16 at 20:12
1

In terms of the potential scope of lost rep opportunity for contributors, do people besides me not upvote answers that they can't easily verify?

YES.

Do others agree that on non-obvious or non-trivial answers, it's important to include the results as proof?

YES.

If so, is there something we can do to educate and encourage authors to do that?

Other than Commenting on each such not-easily-verifiable or proven response, can't see what. I'd set up a block/template Comment to that effect, short and sweet, and use it.

1

In terms of the potential scope of lost rep opportunity for contributors, do people besides me not upvote answers that they can't easily verify?

Yes, I agree. I upvote only answers which I can verify (AND which I consider to be useful, gives additional information to previous answers, ...)

Do others agree that on non-obvious or non-trivial answers, it's important to include the results as proof?

In general, yes. (of course there're also exceptions)

If so, is there something we can do to educate and encourage authors to do that?

I've experience mainly with Excel questions. There are two types of questions here:

  • questions asking for formula ():
    here, adding a screenshot which shows the formula and the correct result could be a quick win, this would already increase the quality of the answer to an acceptable minimum, however that still doesn't make it a good one (see this question too: Should I write detailed answers to trivial questions?)
    When I see answers with the formula only, or even worst: "try this...", I'm tempted to flag it as comment only
  • answers containing VBA: for some people on this site writing (short) macros is a quick win for answering, even if OP asked for a formula, or the questions is of a low quality.
    I find very few good quality questions asking for code (containing: what do you need? what you've tried?) and really hate when people comes here just asking for the code (Is there a macro, which can...?)
    As I generally don't agree with the approach, I'm more strict here: upvote only answers which are answers of a good question and also add some explanation to the code.
    For the original question regarding education: for me adding a proof here is a low priority, before they should understand: answer good questions only; consider coding level of the OP (if you write a code for a novice in VBA, then also include link for VBA tutorials to teach him how to use it).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .