12

Answers of this site are of very disparate quality. We have a lot of bad answers; some have already been improved or removed, but others are lingering.

We're in the process of determining what to do with poor answers. We need to get an idea of the scope and nature of the problem.

This site currently has 392 non-deleted answers, which is a bit much to read in one setting. To focus on the answers that are most likely to be problematic, I've sorted the answers by length. Of course length doesn't make quality, but it is correlated: an answer that is too short surely has too little information.

Here's a graphical representation of the answer length distribution, courtesy of Undo. 131, 143, 147, 166, 183, 200, 207, 209, 219, 221, 228, 228, 231, 234, 234, 239, 241, 243, 249, 254, 255, 256, 258, 262, 274, 277, 278, 278, 279, 294, 308, 319, 323, 323, 328, 331, 331, 332, 332, 332, 333, 342, 342, 348, 348, 352, 353, 353, 353, 357, 359, 359, 361, 364, 366, 366, 368, 368, 368, 371, 378, 378, 380, 385, 385, 388, 397, 400, 400, 403, 405, 412, 414, 414, 418, 419, 421, 422, 423, 426, 428, 431, 435, 436, 440, 446, 452, 453, 458, 460, 462, 470, 470, 471, 475, 476, 478, 478, 480, 482, 483, 487, 492, 493, 495, 497, 497, 506, 509, 509, 513, 518, 518, 520, 525, 527, 528, 529, 530, 538, 539, 543, 548, 554, 562, 565, 565, 566, 569, 570, 570, 572, 577, 578, 579, 580, 590, 591, 599, 602, 602, 603, 604, 611, 618, 620, 623, 625, 633, 642, 644, 649, 651, 654, 655, 661, 661, 661, 664, 664, 667, 670, 673, 681, 681, 682, 683, 684, 686, 691, 692, 695, 699, 700, 705, 708, 719, 720, 723, 724, 724, 726, 729, 729, 730, 733, 735, 735, 737, 737, 740, 748, 749, 750, 752, 756, 758, 760, 764, 770, 774, 776, 778, 781, 785, 785, 791, 791, 793, 793, 793, 795, 795, 797, 797, 806, 820, 821, 825, 826, 829, 832, 837, 837, 838, 839, 839, 839, 839, 841, 841, 844, 850, 853, 855, 855, 864, 867, 868, 868, 868, 899, 903, 926, 926, 931, 932, 934, 938, 940, 942, 951, 963, 978, 986, 987, 989, 992, 994, 1002, 1004, 1005, 1007, 1010, 1012, 1018, 1029, 1029, 1029, 1038, 1058, 1059, 1067, 1067, 1071, 1083, 1087, 1095, 1096, 1099, 1099, 1102, 1105, 1108, 1119, 1124, 1125, 1127, 1132, 1133, 1133, 1144, 1145, 1155, 1163, 1165, 1171, 1182, 1194, 1206, 1207, 1209, 1222, 1232, 1238, 1245, 1249, 1251, 1262, 1266, 1267, 1274, 1300, 1309, 1319, 1320, 1333, 1344, 1365, 1376, 1391, 1397, 1398, 1401, 1432, 1434, 1456, 1460, 1469, 1476, 1517, 1520, 1530, 1558, 1559, 1572, 1622, 1641, 1645, 1672, 1680, 1689, 1705, 1712, 1715, 1719, 1737, 1739, 1741, 1743, 1746, 1758, 1760, 1774, 1776, 1823, 1834, 1835, 1851, 1860, 1864, 1884, 1893, 1899, 1935, 1969, 2003, 2056, 2097, 2154, 2212, 2238, 2265, 2290, 2344, 2396, 2404, 2443, 2524, 2593, 2642, 2675, 2795, 3018, 3056, 3109, 3236, 3573, 3824, 5558, 6032, 6697

I downloaded the data over the API with stackexchange-api-download-all then used shortest_answers to generate the list below. If you want to analyze the data, you're welcome to use adapt the scripts. They're writtin in Python.

Here is the beginning of the list of answers sorted by increasing length (69 answers of length up to 400). The number in parentheses is the length of the rendered HTML. The number in brackets is the current score, followed by one * per comment.

Please browse through this list and evaluate the answers. Are these the kinds of answers we want? Is the answer useful at all? If you think the answer should be improved, post a comment to explain how. If you think the answer is really bad, flag is as “very low quality” — but do leave a comment. Here's a comment I've started to leave, with minor tweaks as appropriate:

Welcome to Software Recommendations. We want answers that explain how the recommended product matches the requirements and fits the purpose of the question. What you've written here is little more than ad copy for the product. Could you expand it to demonstrate how this product fits? See this meta post for tips on writing an answer.

Please post your impressions here. How much crap do we need to get rid of to make this site viable?

14
  • Oooh, I can't wait for this site to show up in SEDE. We can do things like look for answers without links and everything :D
    – user46
    Commented Feb 9, 2014 at 15:39
  • @Undo We can already use the API. I downloaded all the posts on the site (it's still plenty small enough for that) and did the filtering at the Python REPL.
    – Gilles 'SO- stop being evil'
    Commented Feb 9, 2014 at 15:43
  • Maybe it is useful to also post a showcase of excellent answers? Examples of how it should be done?
    – Bernhard
    Commented Feb 9, 2014 at 16:50
  • 2
    Some questions are quite broad (an example) and so answers are very short. Is it a kind of question to improve (but maybe the asker doesn't need anymore requirement), a question to close (it's a bit light to be a question but it's not uninteresting) or there is nothing to do?
    – Fractaliste
    Commented Feb 9, 2014 at 16:51
  • @Fractaliste Indeed, I begin to suspect that many of these answers are adequate answers to mediocre questions. This question falls into the common trap of asking for alternatives instead of laying out requirements, I voted to close it. In such cases, I don't think there's anything to do about the answer.
    – Gilles 'SO- stop being evil'
    Commented Feb 9, 2014 at 16:59
  • @Bernhard That would be something to add in one of the meta threads about answer quality, maybe here.
    – Gilles 'SO- stop being evil'
    Commented Feb 9, 2014 at 17:02
  • Any chance that this post could be made editable by everyone (or everyone over 350 rep)?
    – Nick Dickinson-Wilde
    Commented Feb 9, 2014 at 17:38
  • @Nick-BriarMoonDesign You'd have to flag a community manager to make it community wiki (editable by anyone with ≥100 rep). Why? To remove answers that have been updated since?
    – Gilles 'SO- stop being evil'
    Commented Feb 9, 2014 at 17:41
  • 1
    @Gilles I started a list here. Hope other can contribute to that post too: meta.softwarerecs.stackexchange.com/questions/86/…
    – Bernhard
    Commented Feb 9, 2014 at 18:43
  • After looking through everything in the "≤250" range, quite a lot of this seems to be at a quality level that would require significant edits or deletions. While this isn't true in all cases (there certainly are a variety of succinct posts in this range), a lot of them should be deleted.
    – Flyk
    Commented Feb 9, 2014 at 20:04
  • @Flyk I concur. Ideally we'll give people time to edit (some have already responded), then we'll delete the dregs before entering public beta.
    – Gilles 'SO- stop being evil'
    Commented Feb 9, 2014 at 20:13
  • @Gilles: yep that was my thought.
    – Nick Dickinson-Wilde
    Commented Feb 9, 2014 at 20:54
  • I'd argue that length alone might not be a good indication of answer quality, other than the skinny ends of the bell curve. Someone could post a rambling un-readable mess (for a long crappy answer), or use screenshots but very little words for a good, though low-wordcount answer.
    – Journeyman Geek
    Commented Feb 10, 2014 at 0:31
  • @JourneymanGeek Of course! I wanted to select some of the answers that had a higher chance of being bad, so we could cull some of the worst. If answer alone was a good indication of quality, we'd simply impose a minimum length. Screenshots are worse about the length of the HTML, they don't replace a description of features.
    – Gilles 'SO- stop being evil'
    Commented Feb 10, 2014 at 0:41

1 Answer 1

Reset to default
6

Here are a few observations with examples. I read all answers of length ≤250 and sampled the range 251–400.

First, I note that some of these answers have already been updated in response to comments prompted by participants in this review effort. Let me extend my thanks to both the people who reviewed and commented, and the people who improved their answers: it's thanks to you that this site has a chance!

Some of the shortest answers say nothing more than “this application exists”. These are useless. My least favorite answer in the lot is

First hit on google search RMaps

This goes absolutely against the grain of this site: everybody can search. What we want here is answers based on experience that demonstrate the suitability of the application to the specified purpose.

This question has three answers, all posted within a 20-second time frame (fastest gun in the west), which recommend the same product — two of them also recommend another product. The shortest of these reads

I remember using VirtuaWin for this purpose.

That's marginally better than “found in a Google search”, but still short of what we want. Given the convergence on VirtuaWin, it's probably good, but why? What are its strengths, what are its limitations? If someone comes along and wants to make a better product for this purpose, what's the target to beat?

We haven't always been good at not upvoting mediocre answers, e.g. Web application that creates RSS feeds for any website (score +5).

A tentative answer where the answerer isn't even sure where the requirements are met is not a good answer. While comments should not be used as answers, this is not an answer: it is a suggestion for a potential answer. The product that is recommended needs to be investigated before this can become an actual answer. While this investigation is being done, a comment is appropriate.

A number of short answers are adequate answers to mediocre questions (examples: https://softwarerecs.stackexchange.com/questions/264/other-products-like-libreoffice-draw-for-mac/618#618, Flash player that doesn't try to sneak in extra software). The question is too vague or too broad (for example “alternative to X” without specifying which features of X are important); the answer does what it can. These cases should be resolved by closing the question.

Some questions do call for short answers, when they are looking for one precise feature. There is some correlation between the possibility of a short answer and the question being too broad, but it isn't always the case.

You must log in to answer this question.

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