“Maybe” is one option too many

When I’m planning an event, and I use a web service like evite® to send invitations, that web service offers three choices:

  1. Yes, I’ll come
  2. No, I won’t
  3. Maybe—I’m not sure

“Maybe” is one option too many. As a best practice, we should dispense with it, just as we should replace five-star rating systems with four-star ones.

The problem with five-star rating systems

Let users choose from five stars, and they nearly always pick three. Three is the little bear’s porridge, neither too hot nor too cold. Three is neutral—a safe place to hide. Even in the virtual world, where nothing more consequential is being asked than an opinion, many people would rather equivocate than commit.

But present these same users with a four-star spread and you leave them no cover. Two stars out of four is not neutral. Neither is three stars out of four. Any star rating they choose will reflect an actual opinion. There is no place to hide. When there is no place to hide, courage arises out of necessity. Force people out of the brush, and they develop the backbone needed to state an opinion.

The trouble with “maybe”

As data, “maybe” is as useless as a three-star rating in a five-star system—and as hypnotically compelling to users. “Maybe” is a button that begs to be pushed.

Maybe is a magnet for neuroses. It salves guilt complexes and incites passive-aggressive avoidance behaviors.

“Maybe” sometimes means maybe, but it can also mean, “I’m not coming but I don’t want to hurt your feelings.” Or even, “I plan to come but I reserve the right to change my mind at the last minute if something better comes along.” Some people even use maybe to mean, “I won’t make dinner but I’ll come for dessert.”

When you invite twelve people to a restaurant dinner via a web service, at least four will say maybe. Do you reserve a table for twelve? When eight show up and range themselves at opposite ends of the table (“because other people might be joining us”) you have an awkward table filled with gaps. The empty seats haunt the meal, suggesting social failure.

But if you call the restaurant at the last minute to change the reservation to eight, two of the maybes will show up, like ants at a picnic. They’ll have nowhere to sit, and they’ll blame you. (“I told you I might come.”)

How can you know what “maybe” means? In the context of a web service, you can’t. All you can do is phone people and ask whether they’re leaning toward coming or not—in other words, try to move them from a five-star three to a four-star two or three. If they’re the passive-aggressive type, they will continue to evade the snare of commitment. “I’m probably coming,” they’ll say.

What is the solution? Use web services that offer a binary choice: “I’m coming” or “I’m not coming.” If you can’t find such a service, build one. If you run a web service that includes “maybe,” offer an optional two-choice (“no-maybe”) version.

When demand an outright yes or no, people generally supply it. They only equivocate when handed the means to do so. Form is content.

[tags]design, usability, invitations, fourstar, fivestar, rating systems[/tags]

81 thoughts on ““Maybe” is one option too many

  1. It’s a problem on facebook that runs rampant. It’s likely because older teens (like me) and others on facebook find it easier to always say ‘maybe’ rather than really commit.

  2. Couldn’t agree more…

    No wonder why wedding invites are YES | NO. Too much money wasted for all those MAYBE people…

  3. In my experience with questionnaires, primarily in usability testing, when presented with a four star system, three of four becomes the new three of five. When people are genuinely apathetic about something, they’ll pick 3/4 basically to be polite, wanting to spare the designer’s feelings, missing the point of the questionnaire as an evaluation tool, rather than a tool for validation and warm fuzzy feelings.

    Trimming out the indifferent option may force people to make a choice or it may just destroy the balance on your scale, making it a lot tougher to make a good analysis. And the fun part is that you can’t really tell which one has happened.

    I personally like a 5 star system. 3/5 isn’t really enlightening, but at least it’s pretty reliably accurate. Sometimes people genuinely have no opinion or don’t care about something, and if you take that option away from them, you’re running the risk of just muddling your data with their personality.

    I totally agree that “maybe” is useless in the scope of invitations, though.

  4. There is a lot of evidence that 5-star rating systems are inherently flawed. It’s not secret that the big sites that use them (iTunes, Amazon for instance) offer users a “was this review helpful? Yes/No” link after each star-rating.

    How crazy is it that these sites put a rating on other people’s ratings? Clearly people abuse the rating systems and then other people have to rate the ratings to determine how useful they are. Not so useful when you think about it.

  5. I agree. Well… I don’t know, I suppose “maybe” could have some value. But then again, I can see where it wouldn’t. I guess I tentatively agree, for the most part.

  6. Great stuff Jeffery, with maybe you are giving the user an option not to have to make a decision, and boy do we love not having to decide stuff!

  7. I think four stars is still too many. I want to know if someone would recommend it and if so, why.

    Whether it is movies or hotels, it has to be “perfect” to get 5 stars. If it only gets 4, what was wrong? Did the internet not work? That might be a deal breaker for me, regardless of how many icons it gets.

  8. I find the 5-star system especially annoying since, in most cases, the decision you’re reviewing is a binary one — should I buy X? The worst offender with this practice is Pitchfork, whose album reviews scale from 0-10 with a decimal point, making a 101-point scale. What is the difference between an album that gets a 6.8 and one that gets a 6.9? Is the latter really ever-so-slightly better? And isn’t the info I want ultimately “should I buy this album or not?” I prefer Ebert/Siskel/Ropert/whoever with the thumbs: leave subtlety for the exposition and tell me whether you thought it was worth the time and money.

  9. Heh, when I was younger (like in the 8-12 years old range) my nickname was “Mr. Maybe”. Seriously. I was a fence-sitter from the start. I like to think I may have outgrown it, if not become completely conscious of my “maybe” tendencies. So “maybe” options and the third star are tempting for a person like me and, agreeing with you completely, need to be banished. I have learned that actually making a decision is not the end of the world and is certainly prettier than the view from up on that fence.

  10. I disagree on not allowing a neutral rating. Suppose I am neutral on the subject (say, a restaurant). If I am forced to give a thumbs-up or thumbs-down, I probably will not vote, since I do not want to influence the outcome one way or the other. However, in the results, there’s a big difference between “2 negatives and five positives” and “2 negatives, 5 positives, and 340 neutrals.” In the former case, it would appear that we simply do not have a large sample. In the latter, on the other hand, it is clear we have a large sample, just not many people had a positive or negative opinion.

    I would advocate, on the other hand, a seven-point system:

    0 – very very extremely negative; hate.
    1 – negative
    2 – only slightly negative
    3 – neutral
    4 – slightly positive
    5 – positive
    6 – the best

    This gives people enough room to express not only their opinion, but also the strength of their opinion.

  11. Maybe you’re right, maybe you’re wrong. ;)

    Would one solution to have a deadline on answering? Obviously, you’ll have to order the tables at some point, and if they haven’t confirmed that they’re coming, you obviously shouldn’t order seats for them.

    Seems like an obvious solution to me. Same thing goes for going to movies, for instance. As for parties, perhaps maybe is an acceptable answer … “Maybe I’m coming, I have to find a babysitter.”

  12. My friends and I rate things as being either one star (suck) or two stars (awwwwwwweesome). It works well in practice, because why rate something that isn’t extreme?

  13. Hehe – just what I thought a week ago when I pulled all maybe options from a very simple app we’re using to find out when all bandmembers are able to rehearse… we’d always make phone/SMS rounds for double-checking, and that’s like Xeroxing every piece of paper you’re about to feed the shredder! :-)

  14. The company I work in just tossed out a 10 question, 5 point scale, call survey with a one question, yes or no survey. We ask, “Would you recommend our service to a family member?” That’s it. Binary is where the truth is. Love or hate, yes or no, dead or alive, save me a seat or don’t. Maybe doesn’t cut it – in life, or business.

  15. Odd numbers never worked out in playground sports either. The one is the middle, the ‘maybe’ is the guy left out.

    Maybe if we abolished the practice of fence sitting early on people would actually provide honest input, rather than no input. Unless you are sincerely tracking indifference I agree that ‘maybe’ has no place when seeking true feelings and impressions.

  16. Maybe is still useful in some situations, for example surveys where a person geniunely doesn’t care one way or another about an issue and it would be wrong to record a positive or negative response.

    But for something like attendance, yes/no makes a lot of sense. Plenty of review systems could be replaced with love it/hate it too since each user’s review is often a bit more like a single vote.

  17. One of my old favorites. Definitely maybe. Great song.

    After two decades of self employment I learned that anything but a yes is a no.

  18. Having recently tackled this dilemma on my own site, your post strikes home. I used to agree completely – yes or no, only. Now, I’m not so sure. I’ve spelled out my reasoning here.

  19. I don’t agree on the 7 point system, I think people will just end up choosing the middle one still.

    It might be better to ask the question of “If you HAD to choose ONE or the OTHER, which would it be?” and just give two choices. After that, you could have two scales dependent on their answer, one for the negative and one for the positive. Will take longer to fill out, but should be more reliably accurate. Of course, not necessary for invite, but might be necessary for other survey-style questions :)

  20. Hey do the math people – 3/5 = .6, which doesn’t indicate you have no opinion or a neutral view of something. Rather, it says you liked it, but only mildly. Netflix has it right, translating 3 stars into “Liked it.” On the other hand 2/4 = .5 : exactly half, meaning you didn’t like it any more than you didn’t like it – truly noncommittal.

  21. It’s interesting that you claim “Let users choose from five stars, and they nearly always pick three” because the only studies I’ve seen show that is, in fact, not the case (and, of course, just because I haven’t seen the studies doesn’t mean they don’t exist).

    Take this study dealing with Amazon, for example: http://sloan.ucr.edu/category/working-papers/product-reviews/
    … which shows that >75% of ratings on Amazon are 4-stars or above.

    That’s not to say I disagree with your point (with which I generally do agree). Just that I’m intrigued that you say most choose 3 when I have found that not to be the case.

  22. Perhaps a better option than maybe would be, “remind me later.” I have a hard time thinking to tomorrow, let alone knowing if I’m available two or three weeks in the future. In an invite example, your “ask me later” wouldn’t show up as an RSVP, but it would be off of your desk until you had figured it out.

  23. I definitely disagree about 5-star systems.

    I think there needs to be a neutral territory for those that don’t care, don’t know, or have no opinion. Perhaps it would be better to actually label the options rather than using numbers, but I’m amazed how many times sites give me a Yes/No question where I’d prefer to answer “Abstain!” I end up choosing randomly and polluting their data.

    The best system is probably “Like It, Don’t Like it, or Don’t Care”. Or a 5-star system with 3 clearly labeled “don’t care”.

    Words are better than numbers anyway. People don’t think “I 35% like this album”, but they can decide whether they find it “unlistenable” or “tolerable” or “amazing” pretty well.

  24. If you want to know if someone would recommend a movie to a friend or not, you should not ask them for a rating (regardless of how many stars). What you should do is build a tool that actually lets them recommend it to their friends. Then you just have to sit and watch what they actually do. It’s accurate by definition.

    Implicit is better than explicit.


  25. Nicely put!

    In most cases, people already have the option of not answering the question. You don’t have to respond to the invitation on evite, or rate a book on Amazon. It definitely makes sense to remove the ambivalent option.

    Do you have any references for how useless “maybe” is? :)

  26. For a rating system, i suggest the range of two thumbs down (-2) to two thumbs up (+2). The default would be no thumbs (0), so that “maybe” and non participation are effectively the same (in the traditional system, 0 stars is non-participation, and 3 is maybe). Giving something that you dislike a rating -2 is more satisfying than giving it “one star”. Aggregation can be done via summation instead of averaging to produce a sort of karma, like + 2,154 or – 19,001. Incidentally, this is how digg (-1 = bury, +1 = digg) and reddit (down arrow = -1, up arrow = +1) work.

  27. “I think there needs to be a neutral territory for those that don’t care”

    There already is one – it’s called ‘not filling in the form’.

  28. On invites “maybe” may not be very useful, unless you only want to get an indication on how many people will show up.

    I don’t realy see the problem of a 5 star rating system. I mean I see your point, but as it’s mentioned before the rating will probably be not very accurate if people are forced to choose an option they don’t realy want to. If you’re only interested in the positive and negative ratings it’s quite easy to filter them.

  29. While I see your point in things like evites where a definitive yes/no is definitely desired. In a 4 star system 2 being 50% really is being neutral in terms of how averages are really taken. So, while you might mean to say that you disagree, you might actually be voting for “i’m neutral,” depending on the algorithm.

  30. Unrelated, but people should stop putting trackbacks in the comments. They are not comments and it is disruptive to read something like Zeldman on Maybe as a comment which adds absolutely nothing.

    Separate trackbacks from comments.

    Sorry to be off topic, but I am annoyed by it on many blogs and have finally started to comment about it.

  31. First off, I hadn’t give rating systems much thought. I have gotten annoyed when there wasn’t a “middle of the road” option, but then proceeded to select a polarizing choice anyways. I think what you’ve said about forcing people either all the way to a black & white decision or gently shoving them over the cliff graduated true/false is a great idea. Thanks!

  32. To those of you saying that 3 stars out of 5 = .6, so a positive review, and that 2 out of 4 stars = .5, or a neutral review, that only works in a system where you can choose 0 stars. With a 4 star system without a 0, the only logical way to assign the stars worth is to give 1 star 0%, 2 stars 33.3%, 3 stars 66.6%, and 4 stars 100%.
    Same with 5 stars, 1 star being 0%, 2 stars being 25%, 3 stars being 50%, 4 stars being 75%, and 5 stars being 100%. So what he was saying in his article is correct, and I agree with him.
    Another interesting possibility, however, would be to offer a maybe early on, but when it gets close enough to an event, require all maybe’s to pick yes or no, or automatically assign them to no. That gives people a little leeway early on, but will still get you the information you need to act on by the time you need it.

  33. For innerTee.com (it’s undergoing an overhaul, right now) we decided on a yes-only system. If you like a t-shirt mix/design vote for it. It helps keep the community positive since we’re negative by nature (We’ve found this keeps people from being mean).

    That aside, in relation to CSS and web standards as a whole, I feel like we’ve been fighting the “Maybe browser” known as IE for quite some time (I’ve been fixing quirks today so it’s on my mind at the moment).

    Thanks for the post.

  34. Maybe the problem isn’t that maybe is one too many options, it’s that it’s one (or more) too few. In response to an invitation, as you quite correctly point out, “maybe” could mean a whole range of things. But what you seem to be neglecting is that in the world of human relationships, all those alternatives are possible, and if you only give people one option to say something other than “yes” or “no”, you lose that richness of discourse. If you’re going to scrap “maybe”, you need to give people some other way to express an answer more complicated than a single bit.

    After all, if the only colours you will acknowledge are black and white, who are you to tell the person attempting to describe turquoise that their vision is limited?

  35. I have to agree with Dave. I pick 3/5 only when The question is indifferent to me.

    For example, say one survey question asks me what I think of McDonald’s new hamburger (1 being terrible, 5 being awesome). If I don’t pick anything, the survey won’t let me continue. I have never tried it, so my only option is 3 (unless I just want to cancel the survey).

    I am on a feedback panel, and I get surveys like this all the time, asking me to rate this and that despite the fact that I don’t have any experience with this or that.

  36. Prime example. I just filled out a survey:

    “Below you will find a list of characteristics. For each, please indicate how important this characteristic is to you when visiting a rewards website”

    This assumes I visit rewards websites.

    There was no option for me to indicate I do not visit rewards websites. Needless to say, I voted “Neither important nor unimportant” for each one.

  37. I would guess that having four choices is optimal: yes, probably, maybe, no. Then the maybes would mean “probably not”, giving people a way to decline gracefully.

    This would also help planning the event: I’d bet that 0.9 y + 0.5 p + 0.1 m would yield a fair estimate of the event attendance. You could even have the organizers report back, and tweak the formula based on actual data.

  38. To be honest, I think websites get much more insightful comments when there is no rating system. People use the ratings as a gesture more than a means of explanation. It becomes a fight of ratings.

    Ratings are much more useful in a private environment. Think about it. When you use the 5-star rating to rate songs in your own itunes library, you reserve 5-stars for the few truly great songs you own. But, when you go into a public forum, you’re trying to convince other people with your vote.

    People just don’t vote the same way between private and public environments.

  39. I would have thought the obvious fix for a 3-value (“yes”,”no”,”Maybe”) system (where the mid value is uninformative) would be to move to a 4-choice system as you stated. For the “Are you coming?” type question the 4 posible choices would be
    Probably will
    Probably won’t


  40. If it was up to me I would suggest a 4-choice system:
    – Yes
    – Maybe (with a “+”)
    – Maybe (with a “-“)
    – No

    I agree, “maybe” is one option too many, but it is important to have a middle way for sometimes one isn’t sure whether he/she is coming… But we always know whether our “maybe” leans towards “yes” or “no”. That’s why I believe it would be convenient for a birthday person and fair to his/her guests.

  41. when offered the choice of “yes” or “no” i usually opt to not reply.
    thats my maybe!

  42. Neutrality is important sometimes. It just doesn’t belong everywhere.
    Forcing someone to pick love or hate doesn’t give you a better idea of their feelings, it just gives you a less detailed or less accurate view of their feelings. Maybe they are neutral because your product or proposition doesn’t excite them, or because your survey is dull dull dull and they’ve gotten 2 other dull surveys already this week. You can’t fix the problem by being pushy.

    And for inviting 8 people to dinner – you don’t use evite. You ask them.

  43. Here is a question about neutrality. In a four star rating system, what would you check if you truly didn’t care one way or the other? Without a neutral option, are you forced to give a dishonest answer?

    Is it possible to be truly neutral? Where you truly could care less? That seems like an extremely fine point to balance on, and when forced to give an answer, there usually is something that will tip the scale.

    Forcing your users out of passivity should create a more active community.

    As for evite, I wish someone would build a better service. It seems most people reside in the “Have not Responded” category.

  44. Sure, ‘maybe’ should be dismissed. But isn’t 3out of 5 stars simply 3 out of 5, or 6 out of 10, or 60%? Isn’t 2 stars out of 4 really 50%. And doesn’t 50% seem an awful lot like a ‘maybe’ to you?

  45. I agree hugely on ratings. I’d prefer a 0-100 score myself…. But maybe is necessary for events. Commitment is hard with meetings or shows or conferences. As a user, I’ll want to express interest to show others (maybe they’ll convince me to go) or to keep track of the events I’m interested in (see upcoming.org).

  46. It’s probably too late to comment on this and still receive any answer, but I’ll go ahead anyways…

    For the purpose you described, you are right, you need to know what is going on. But in surveys I find myself rather not saying anything than lying. In my opinion, this says more about a person than a lie. A question where it is quite obvious that a “yes” would show bad on your personality, you rather not lie and go with the middle approach. If you process your data accordingly, you might get better results than leaving the “user” no choice, or no “win”-situation.

    So far I have not found anyone expressing this and no data processor taking this into account. I wonder why? I think a tendency to the middle says as much as the tendency to the extremes, because the middle is an extreme, too. Ask Switzerland ;P

  47. I think five-star rating need to be changed to seven-star. 0 means very very very bad and 6 means perfect. But now there are a lot of things that can take -1:) but things with 6 almost isn’t present:(

  48. I’ve always had that notion when creating survey questions et. al. Offering a middle-of-the-road choice can be a disaster if you’re really after solid and telling information.

    Take, for instance, the question “How are we doing?” With three responses, “bad,” “okay,” and “good,” the users who chose “okay” don’t really reveal a thing thus the thing becomes a waste of time. If, however, the survey offers two choices, “bad” or “good” the person taking the survey is literally forced to provide more telling and directional information by having to select good or bad. This type of data, even if it’s not completely accurate (assuming the don’t think “good” or “bad” are truly fitting), having to choose one way or the other tells us in what direction they lean.

    Granted this works better, and in more fairness to the person providing the answers, in a five/four question survey, but the point remains the same. It’s not about making the survey easy, it’s about getting useful data.

  49. Star systems fail to be useful usually because they lack any context of what the stars mean. One user may regard ‘very bad’ as being 1 star, with 3 stars as ‘not sure/middling’, and 5 stars as ‘very good’ while another will regard ‘very bad’ as not voting at all, where 1 star is ‘bearable’, 3 stars is ‘good’, 4 is ‘very good’, and 5 is ‘excellent’. Neither of these are compatible.

    So we need to regard accurating ratings as being multi-dimensional and encourage some way of capturing this. When you rate something using a star system, using what language are you actually rating it? If you give a CD 5 stars, you can legitimately be saying you’re entirely satisfied with the product quality, service and delivery, but that the music is actually dire. If instead you think it’s one of the most influential albums you’ve ever heard, but that it’s still pretty dire, you are giving it a rating for a different reason again.

    Different rating contexts require different gradings and some should cross into the negative side of whatever dimension you’re looking at (5 star gradings are usually either 1-5 or -2 to 2). Take two common evaluation questions:

    – How important to you is X?
    – How satisfied are you with X?

    Satisfaction seems to be OK with Very Unsatisfied, Unsatisfied, Neutral / Don’t Know, Satisfied, and Very Unsatisfied.

    Importance then somehow ends up being Very Unimportant, Unimportant, Neutral, Important, and Very Important.

    What is the different between Very Unimportant, Unimportant, and Neutral? If something has no importance, it is all three of these.

    A 2 dimensional rating system (actually 3, because you’re always looking at volume of each response) may be effective at times – use a good/bad Y axis to measure the quality of something, and an impact X axis to measure the significance of this measure of quality. Then, a user can say ‘I think this is really good but in the grand scheme of things it’s pretty insignificant’ (x=1 on 1 to 5, y=2 on -2 to 2) or they can say ‘I’m not really sure if I like what you’ve done with the design but at the same time it isn’t really affected my user experience so I wouldn’t worry about it’) (1,0).

    Alternatively conceive your 2nd dimension by asking the user prioritise the different qualities/characteristics they’re evaluating, as well as a single (useful) grading scale independent of the priority.

  50. Here’s an alternative, it lacks some of the features of evite, but at least adds the ability to control the answer set.

    We just let iPoll loose on the world, build your invite in an web phone or browser and then send it to your list via email. There is also a public sharing feature, but that’s probably not appropriate for an invite. I hope it fits in somewhere – thanks.


  51. When my wife and I moved to Pittsburgh, we went out to a thai restaurant to celebrate our move. When ordering my dish, the waitress asked me, “How spicy do you want your dish on a scale of 1 TO 10″. I was like, damn, that’s a lot of choices. Turns out all the other restaurants ask me the same thing. Indian, mexician, they all have given me 1-10 to choose from. I thought 1-5 was bad enough…

  52. the same thing happen to me and my boyfriend. at different restaurants they do sometimes offer a scale for how hot you want your food

  53. To me, there are two different situations being discussed here, one: will you/won’t you be doing something, and two: giving your opinion. As much as I agree with the need to have only yes or no on an invitation, the commentary that followed on the uselessness of a 5-star rating system was misplaced. After all, “Will you be coming to dinner?” is a completely different type of question than “Do you like broccoli?”, as in the end, there are only two options on the dinner: eventually you will either go or you won’t go. However, in matters of opinion I find there is rarely an absolute yes or no, and how do you capture any of that variety if all you give people is yes or no.

    Perhaps one could look at it in that if you are getting the majority of maybe’s for responses, what is truly not working was the question rather than the rating system. “Do you like broccoli steamed until just short of crisp?” is a much more complete question, and much more likely to garner you a yes or no response. Tailor your question with an eye to how much information you want in response, as it were.

    I suspect many maybe’s could be “It depends” in disguise, since even a 5-star rating system only allows for degrees of yes or no, and many times the answer varies with the situation. I don’t like broccoli at all if you boil it the way my mother does, but it’s my favorite thing in a stir-fry. I liked the functionality of the feature as described, but as it didn’t actually work in my browser, how am I supposed to answer yes or no? Do I answer no and have you thing I didn’t care for the feature itself, or do I answer yes and have you totally miss the fact it didn’t work?

  54. This five-star rating system does’t appear to be that crazy if you compare it to the school rating system. In many european countries there are marks from 1 to 6 or a to f which is a 6 level rating system. So when rating you just give out a mark. Three is in the middle so its mediocre. I think 4 level rating system would be too less.

  55. Another person’s “five star” rating is much more useful to me if I think their tastes are similar to mine.

    Someone mentioned that most reviews at Amazon are positive. I think it makes sense: who would bother reviewing a bluegrass album unless they’re a bluegrass fan who owns this album? In which case, they probably like it. They may imagine that non-bluegrass fans will see their rating and be convinced to try it; therefore, it’s a review with an agenda.

    If I want to discover new music, I find it more useful to find something I know that I like and look at the “people who bought X also bought Y” information. It’s also cool to look at the “favorite albums” lists of people who share my opinion of this particular album.

  56. Pingback: Profiblog
  57. I prefer the 5 star rating system because three stars is a perfect rating for a pleasant, but forgettable film. The type of movie that you don’t think was a waste of your time or money to see, but you couldn’t be bothered to bring it up and recommend it to friends.

  58. Interesting post! Great point about 5-star systems.. but I have 2 important counterargs:

    1) The histograms (of rating distributions) that I’ve seen don’t support your hypothesis that many users rate items 3 stars. Why would apathetic/callow users bother rating at all?

    2) A sophisticated system can normalize the ratings data by, eg, discounting ratings from apathetic users (who tend to rate everything the same way), or even disregarding 3-star ratings altogether.

Comments are closed.