If there is a scientific basis to the Three-Click Rule, we couldn't find it in our data.

Nor can it be found by flipping a coin... Analyzing the wrong data proves nothing, other than the person analyzing it doesn't know any better.

Posted by: Ron Zeno on April 17, 2003 12:21 PM

---

 

There must be some word for a rhetorical technique wherein a proponent of an argument looks at successive issues in isolation, and sees that none of them accounts sufficiently for a result, and concludes that none of them, therefore, are an issue; as opposed to looking at the totality of an issue, and seeing how they contribute to an end result by their being together (in modeling terms, an interaction effect).

This same outfit has told us that page download time doesn't matter. But something tells me that three clicks would be a much more significant guideline for a website where the pages take forever to load. As cilcks become more of a 'hassle', due to slower downloads, difficult language, obscure paths, etc., I bet Jared Spool would see more of a difference. And I'm not saying that the other factors would outweigh the number of clicks, but that too-many clicks in the rpesence of other factors would combine for a problem. (Those other problems might not seem to be an issue is the path is short.)

Perhaps that analysis exists, but it hasn't been done in this superficial treatment, where everythng focuses on one variable.

Posted by: Frank on April 17, 2003 02:13 PM

---

 

I think three-clicks is a useful suggestion for beginners because it encourages site structure and organization. If you've ever watched a CEO type design a website, everything is buried under one or two main links. First I go to Mens, then I go to dress shirts, then I go to accessories, then I go to jewelry, then I choose cufflinks, and Oh! we can throw a search box in so they can find it right away if they want. Brr.

I agree, though, that guidelines can be dangerous because it is too easy to see them as rules that only the uninitiated ignore. Basically, I think it is up to the designer to have the confidence to ignore guidelines if he thinks it will work for his particular audience.

I hardly ever see new designers apprenticed by more experienced designers, but they usually seek guidance from some source. Guidelines can help if they are treated as a suggestion rather than dogma.

Posted by: Lydia on April 17, 2003 03:51 PM

---

 

Frank,

I see your point. We definitely thought about this when we were conducting our analysis. In systems with many variables such as the web we would expect to see some correlation (however little) between whatever it is we're looking at (in this case both success and satisfaction) and some factor that may or may not influence it (length of clickstream). The idea that you're alluding to (as I understand your comment) is that many little correlations would ultimately combine to form an overarching account of what is really going on.

However, what our data shows is that there is effectively no correlation at all between clickstream length and either success or satisfaction.

But let's assume for a minute that there is a very small correlation between clickstream length and success or satisfaction. Does that justify the creation of a rule that designers should blindly follow? Our intent with this article, and with much of our research, is to direct designers to those things that have significant effects on the user's experience. We try to dissuade them from relying on red herrings like the Three-Click Rule. That said, the Three-Click Rule may be beneficial in the sense that it may make designers focus more on users.

What the creator of this rule did is work backward and take user complaints about clicking(q) and create a design rule that MIGHT explain user frustration(p). This is illogical: p => q != q => p

Posted by: Josh Porter on April 17, 2003 04:13 PM

---

 

Frank, could you be looking for the logical fallacy of composition?

Posted by: MadMan on April 17, 2003 06:14 PM

---

 

Josh, I think that for clarity I need to revert to modeling rules here. And I apologize to everyone here if this is not what you're used to (I empathize with arcane expressions with formulas). I also hate to shock regular readers of this board who will be surprised to see this side of me!

Imagine for a moment that the negative impact of number of clicks (in terms of task completion and/or satisfaction) only manifests when there is some other break-down in the web site (such as slow pages or other hassles).

A VERY SIMPLE expression of this might be:

Neg effect = B1*A1 + B2*A2 + B3*(A1A2) + y intercept

A1 = a threshold for number of clicks is crossed or not, and equals 1 or 0 (or make it linear if you want, where it's the number of clicks)

A2 = other usability issues with each successive click exist, and the value is 1 (they're present) or 0 (they're not)

A1A2 = the presence of the combination of the two

Bi = impact of each of the variables above, as measured through ANOVA or least squares regression, etc

According to modeling rules, if the coefficient B3 (for the interactive term A1A2 -- that is, there may be too many clicks at the same time as there are other usability issues with each successive click) is statistically significant (that is, the presence of the combination has a negative effect above and beyond what you see from the effects of either 'main effect' variable on its own), you can't remove either A1 or A2 from your model, no matter how small the estimate of their impact is.

That is, if B3 has a significant impact, you have to include the other components in your expression of the relationship.

Now for the English, in terms of the discussion at hand:

If an interaction effect exists, you don't say that number of clicks has no effect, you have to say, "well, it depends." Number of clicks might not be an issue if the user feels momentum. But if each page is a puzzle, number of clicks is a problem. (There may also be points where the user feels momentum, but 45+ clicks is a hassle!)

The other important analytic point is this: if the research didn't encompass sites/stimuli where each successive page presents a problem (as well as sites/stimuli where it's smooth sailing), then there is no way to know about the interaction effect. And in that case, the conclusion that number of clicks is irrelevant, is tenuous.

Now. I can understand that three clicks is not a rule, and in my meetings at my old Fortune 50 company I hated to hear people talk about it and leave off the "other things being equal" caveat. We are trying to get people through their tasks with minimal hassle. But I have to admit that I liked the winnowing design of the old bananarepublic.com where you went through gender, garment, etc. More clicks, but it built confidence along the way, rather than confronting you with a long list of buttons/links to take short cuts.

So, I guess that I endorse recommending against the rigid adaptation of a three-click rule, but I think more work needs to be done. And I also think that designers are capable of seeing the complexity of "well, it depends."

RE Argument of composition: I don't think this is quite what I meant, though I do see the interactive component. I was actually thinking more along the lines of the argument of the beard, and critical mass. No single hair makes a beard, but put 'em together and they do.

Apologies again to all those shocked readers who were surprised to see this side of me. I was a bit of a quant jock here and there in my professional career.

Posted by: Frank on April 17, 2003 09:02 PM

---

 

BTW, I apologize for using the word "outfit," it sounds needlessly pejorative. No offense meant.

Posted by: Frank on April 17, 2003 10:11 PM

---

 

Frank,

Let's see if I understand what you're saying. I think you're saying something like:

"More clicks is bad if each click gives you something that sucks." or "More clicks might be positive or neutral if each click gives you something that is great or doesn't suck."

Sounds logical.

So "all other things being equal," no one wants to click at all - a "NO clicks" situation is best - "three clicks" must've came out of a committee and involved some compromise with pointy-haired types.

This leads me to my life's work (and a free non-returnable gift to Webword readers): "Lyle's No Thought Rule" - users want you to give them what they want/need before they even think of it. (All other things being equal. While supplies last. Not available in all areas, some exclusions may apply. Contains explicit lyrics, may be objectionable to some people and cats.)

But seriously, this topic relates to the "scent of information" thing. Which, if web sites were movies, says that sitting through long movies sucks, unless it's an entertaining or informative movie, in which case, long movies may be just as enjoyable as short movies, and in some cases even more enjoyable. If you can cram the same enjoyment into a short movie as a long movie, that's great go for the short film. UNLESS you can charge more for the longer movie... BUT if you only have enough info/tainment for a 30 second commercial, then go into advertising instead of making an epic. But remember that advertising doesn't work on the web. Jakob said so years ago - it's even written on some stone tablet somewhere.

Josh makes a good point in his article when he says that the 3-click rule might be good in that it makes people think about users. While thinking about users is good - if that "thinking" results in a worse design than not doing that "thinking", then it's NOT good. Personally, most experience with the 3-click rule is that it's usually quoted or mis-quoted by someone on a team who's really just using it to win some design debate. It's quoted as irrefutable gospel to be complied with -- there's not been a lot of thinking or user involvement going on at the time either.

Let's face it, debunking "3-clicks" doesn't take much effort. Just ask for 3 or 4 good sites that actually follow that rule. Lots of the best sites don't.

Posted by: Lyle - Usability Guru on April 18, 2003 12:22 AM

---

 

To debunk a rule, first research the rule well enough that you understand it... (Hint: the three-click rule predates the internet AND it deals with time to complete a task).

Posted by: Ron Zeno on April 18, 2003 01:06 AM

---

 

Lyle, your re-characterization of what I wrote is fairly spot on. And I'm in favor of debunking the three-click rule if it makes sense. But if we conclude too quickly that number of clicks doesn't matter — something different from the three-click rule — we risk giving license to burying content and thus increasing task difficulty.

One of the problems I have with focusing on number of clicks, or page download time, or any other single factor, is that we lose site of overall task difficulty in doing so. My copy of The Humane Interface is on low-likelihood-of-return-loan, but I remember that there are many components to task ease, and that anlyses (is HOMES the the acronym? no, that's the Great Lakes...) can be done which include these. I won't list the other issues; but if they're not measured, we can't see how they interact to contribute to task difficulty.

I'm glad UIE did their analysis where the entire task was the issue, and not a single page, but I do believe these things add up.

And I also remember the caveat in another thread (was it on page download time?) where we were reminded that in a lab setting, where respondents know they're there for an hour, they don't have the at-home distractions and priorities which call to task abortion.

Posted by: Frank on April 18, 2003 06:36 PM

---

 

Does some this depend on the task? If its something I really want to find or do I may be willing to click more times to succeed?

Posted by: daniel szuc on April 19, 2003 03:59 AM

---

 

Not sure how soon I'll be back in a profession where this matters, but is the acronym which I was looking for GOMES?

Posted by: Frank on April 19, 2003 10:29 PM

---

 

Frank,

I think I said this before, but it's worth repeating: All other things being equal, more clicks is a Bad Thing.

Keep in mind that "task difficulty" and "perceived task difficulty" are two different things. Sometimes getting the "answer" faster may actually reduce user satisfaction (but in general, faster is better). I think I've seen research to this effect, but it's too late for me to go hunt up a reference. It could be that a 'fast answer' experience doesn't allow for serendipity or encourage other exploration...not sure what the underlying reason would be - likely some psychology thing like anticipation of enjoyment actually enhancing enjoyment (e.g. enjoying making the coffee, not just being able to drink it).

Meta topic:
Even though it's less than a week old, this discussion is likely to go cold soon since it's way below the fold on the Webword home page...sigh. See my comments on the Webword redesign post...

Posted by: Lyle - Usability Guru on April 22, 2003 02:35 AM

---

 

Lyle, no disagreement here, all other things being equal.

Posted by: Frank on April 22, 2003 08:52 AM

---