How Microsoft Degrades Their Users (In a Good Cause)

blackbearnh writes "We all know that slow Web pages drive users crazy, but where is the boundary between too slow and too simple? As Microsoft's Eric Schurman points out, the fastest-loading page of all is a blank one, but it's also the most useless. In an interview with O'Reilly Radar leading up to his appearance at the Velocity Conference, Schurman talks about his experiences working on some of Microsoft's highest-volume sites, including the home page and Live Search. In particular, he discusses how Microsoft will selectively degrade the performance of pages to small sets of users so that they can see how various amounts of delay at different times and places affect user behavior. 'In cases where we were giving what was a significantly degraded experience, the data moved to significance extremely quickly. We were able to tell when we delayed people's pages by more than half a second, and it was very obvious that this had a significant impact on users very quickly. We were able to turn off that experiment. The reasoning... was it helps us make a strong argument for how we can prioritize work on performance against work on other aspects of the site.' He also talks about what it's like to be one of the most often-targeted DDoS sites on the planet."

Comments

As opposed to ...

[fahrbot-bot]

In cases where we were giving what was a significantly degraded experience ...

... the normally degraded experience.
(Ba da BOOM! Don't forget to tip your waitress.)

Re:As opposed to ...

[zxjio]

Experimenting by delaying a pageload for 500ms is worthy of ethical considerations? Would you like to sue Microsoft for emotional damage? Too many people are afraid of doing anything these days.

Re:As opposed to ...

[Hal_Porter]

Gmail does "selective degradation" really well. E.g. if you load gmail over a slow VPN over wireless connection it says "This site is taking longer to load than normal, would you like to try the Basic Html version or wait longer". Also you can choose basic html (i.e. less ajax and css) as your default view.

Basic HTML is quite usable these days - it even does email address autocompletion on Opera. So it can use ajax but it presumably doesn't depend of it. In a way it's a bit like a well written application which can use new features if they are present but run without them on downlevel systems.

Re:As opposed to ...

[Anonymous Coward]

You're discussing the wrong topic. "graceful degradation" != "selective worsening".

In Gmail, the degradation is helping you cope with a poor connection. In Microsoft's experiment, the degradation is hurting you and is imposed by a coin flip.

Re:As opposed to ...

[Jurily]

Experimenting by delaying a pageload for 500ms is worthy of ethical considerations?

No, they should be shot on sight.

Re:As opposed to ...

[EvanED]

That would probably completely invalidate the results though, for two reasons. First, the sorts of people who would opt into that wouldn't at all be representative. (It would take an unusual person to even find the opt-in, let alone volunteer for a degraded experience knowingly.) Second, knowing about it would be way too likely to affect how the people behaved.

You could get halfway by saying "would you like to help us do research" or something like that, without saying in what way, which would reduce these problems, but not completely.

Re:As opposed to ...

[Hal_Porter]

The summary is about graceful degradation. TFA is OFFTOPIC here.

Re:As opposed to ...

[Jesus_666]

That's half a second! Let's do the numbers:

We assume that Live search gets ten billion hits a day. We also assume that Microsoft degraded 5% of all hits. Thus Microsoft has wasted 1000000000 * 0.5s * 0.02 = ten million seconds! Microsoft wastes more than 26 years worth of productive time per day. Now, assuming that the computer of the Live search users consume 800W on average, we find that Microsoft wastes a whopping 20.9 watt-millenia per day. Assuming that 80% of that is turned into waste heat it's obvious that this has a non-negligible impact on Earth.

Gentlemen, I think we have found the root cause for both the energy crisis and global warming (and because our bitching about the oil price annoys the arabic world, also islamic terrorism). Now all we need to do is keep Microsoft from doing these experiments and everything's dandy again.

Re:As opposed to ...

[tannsi]

I think this would only actually be a problem if anyone used LIVE search.

Re:As opposed to ...

[jargon82]

Exactly. It's their site, and they certainly allowed to do with it what they want :). They could do "market research" and ask people how slow it could be, but instead they are collecting real world technical data and gaining insight on to how the performance impacts real people. Hopefully they then use this to decide where to spend time on performance.

Re:As opposed to ...

[OakDragon]

Modded as Funny (100%), but you've just laid out the environmentalist rationale for controlling everything everyone does, everywhere.

Re:As opposed to ...

[Anonymous Coward]

You'd probably end up fathering Seth McFarlane.

Re:As opposed to ...

[PitaBred]

We assume that Live search gets ten billion hits a day

While we're at it, can we assume that I have ten million bucks?

Select groups of users

[mysidia]

selectively degrade the performance of pages to small sets of users

In other words, Firefox, Opera, XP, and Linux users. And the experiment will get turned off, once they switch back to IE8 on Vista.

Re:Select groups of users

[perryizgr8]

no, he did not forget. they are just too insignificant to talk about.

Agile and all that

[BadAnalogyGuy]

What is cool about the Web is that it is the most Agile of all release environments. Unlike shrinkwrap software, web software can be changed very easily and universally for all users. It brings a raw edge to the development of the software.

In this, there is also the possibility of becoming complacent and ill-tuned to the needs of your users. Taking Google as an example, they keep their services in a perpetual state of beta, always in testing, never reaching a final v1. This type of reliance on constant feedback from customers may work for a short while, but unless the product reaches a state of relative stability (in terms of both not crashing and also not changing) the users will typically find some other software to use.

So when Microsoft decides to impact a few customers with degraded QoS, they may be setting themselves up for a bigger fall later. By introducing the possibility that MS may actively sabotage your user experience in the name of experimentation and testing, they degrade their own reputation (as much as it can be degraded from its current levels) and needlessly increase FUD regarding their proffered services.

It may be for a good cause, but customers should not be the ones testing Microsoft's software. As a professional software house, they should provide good quality control before software hits the servers. It doesn't matter if this is the age of Agile or not.

Re:Agile and all that

[Anonymous Coward]

In this, there is also the possibility of becoming complacent and ill-tuned to the needs of your users. Taking Google as an example, they keep their services in a perpetual state of beta, always in testing, never reaching a final v1. This type of reliance on constant feedback from customers may work for a short while, but unless the product reaches a state of relative stability (in terms of both not crashing and also not changing) the users will typically find some other software to use.

You just disproved your own point.

Re:Agile and all that

[Volante3192]

Yep. Half a second delay is the end of the world. Age of Linux as SaaS...or whatever buzzwords are in play here.

I'm up for a good Microsoft rant any time, any place, but if a small batch of users have to take a performance hit to improve the experience in the end for all users, isn't that a positive thing? You can't really beta test this stuff. You can try running simulations, but nothing beats real world numbers.

Would we view this any different if Apple tried it? Google?

(Disclaimer: This is a logic exercise. In reality, I doubt there's actually much MS could and would do to their site to improve my experience using it.)

Re:Agile and all that

[perryizgr8]

i think google already does this. i read about a ui designer who left google. he said that google relied too much on experimental data for their colors and ui than his advice. for example, if they had to choose the color of the search button in youtube, dark blue or light blue. so for a day, each color would be tried. due to the sheer volume of clicks, they would be able to see patterns and then decide which color users are more likely to click on.
but the key difference here is that changing ui is nowhere as prolematic as reducing speed.

Re:Agile and all that

[Hal_Porter]

if a small batch of users have to take a performance hit to improve the experience in the end for all users, isn't that a positive thing?

Didn't Jesus say "the needs of the many outweigh the needs of the few"?

Re:Agile and all that

[Fallen Seraph]

In this, there is also the possibility of becoming complacent and ill-tuned to the needs of your users. Taking Google as an example, they keep their services in a perpetual state of beta, always in testing, never reaching a final v1. This type of reliance on constant feedback from customers may work for a short while, but unless the product reaches a state of relative stability (in terms of both not crashing and also not changing) the users will typically find some other software to use.

Yeah! I mean, take IE6 for example. That didn't change in a REALLY long time, and lots of people use it! That makes it good, right? [/sarcasm]

Your statement neglects quality. Yes, people want sites that're stable and don't crash, and yes, changing the design every week is bad and confusing, but improving on the design and function of a site is always a good thing, so long as you do so at intervals large enough for users to adjust to. The design of Gmail has only changed drastically two or three times in it's history design-wise, but they still consider it a Beta (depending on what you consider a drastic change, of course).

The issue is that Google, once simply a search engine, is now in the Web Services industry. The fact is, no matter what the label says, Gmail and many of their other apps are not in Beta, and haven't been in a long time. They're just hesitant to call it "v1" or something because that has a sense of finality, like customers shouldn't expect it to change very often. With the web, and Web Apps in particular, that's no longer really the case. They are often redesigned and redone to improve their performance, effectiveness, ease-of-use, and even aesthetics. You even point out yourself how agile the web is as an environment for releasing software. You neglect, though, that this keeps it interesting for the users as well, because they like the feeling that their product is continually being improved at no extra cost to them (unlike many shrink-wrapped software) (Note: When I say "extra cost," I mean in addition to any subscriptions they already have to the service, if any).

The "Beta" in Google's case is very much a marketing issue as much as it is a technical issue.

Re:Agile and all that

[rishistar]

I always thought that was Spock.

Re:Agile and all that

[x2A]

"The "Beta" in Google's case is very much a marketing issue as much as it is a technical issue"

And perhaps a commitment issue, like people who stay engaged forever but never actually get married...

One of the most often-targeted DDoS sites

[Anonymous Coward]

Thanks for the reminder, it's already been a couple of hours since my last flood ping! Now if you excuse me...

The woods are lovely, dark and deep,
But I have promises to keep,
And pings to send before I sleep,
And pings to send before you sleep.

Re:One of the most often-targeted DDoS sites

[Rosco P. Coltrane]

So you ping and then you sleep()? Not much of a flood ping really...

Re:One of the most often-targeted DDoS sites

[RichardJenkins]

So you ping and then you sleep()?

TODO: Insert geek related sex-joke here

Punch your customers in the face, selectively

[syousef]

If I were running a fast food restaurant one of the first things it would make sense to do is pick groups of customer to punch in the face instead of giving them their order. It's all for a good cause. We want to know just how much abuse they'll take before they go down the road to the competition. That will help us figure out how good our food is. Now did you want a fries with that burger? *PUNCH* How about a *PUNCH* drink?

See how absurd it sounds?

Re:Punch your customers in the face, selectively

[lxs]

Don't be so negative. They're simply migrating the Vista experience to the Cloud.

Re:Punch your customers in the face, selectively

[totally bogus dude]

It sounds absurd because what you're saying is absurd.

If you ran an experiment where some customers had their orders delayed by a few minutes more than was necessary and had some kind of metric to determine their enjoyment of their dining experience, it wouldn't be so absurd. Perhaps you provide free internet access in your store, and the extra delay results in a greater chance of people making use of it. And once they've started using it, there's a greater chance they'll decide to order a coffee after their meal and stick around for a bit longer.

Or maybe you find they're less likely to return to the store. That might be hard to track, but the point stands. There are some things which are interesting and which may or may not give unexpected results when tried in real life. If an experiment like this shows that a few minutes delay significantly upsets customers, then it becomes clear that spending extra money to have more staff on is probably actually worth the expense. On the other hand if you can show that most people don't notice, then it makes sense to risk having a shortage of staff at peak periods if you can save a bit of money.

You might even find unexpected results, for example maybe a lot of people after waiting a few minutes with nothing to look at but the menu end up ordering more than they initially would, so it's actually profitable to make people wait longer. Who knows? The only way to find out is to experiment.

Re:Punch your customers in the face, selectively

[nEoN nOoDlE]

Hey, who invited the logic guy to this Microsoft bashing thread?

Re:Punch your customers in the face, selectively

[stephanruby]

If you ran an experiment where some customers had their orders delayed by a few minutes more than was necessary and had some kind of metric to determine their enjoyment of their dining experience, it wouldn't be so absurd.

Sure, it wouldn't be so absurd, because we all know that a Microsoft Live results page is just like a nice burger, or a nice frothy Guiness getting poured ever slowly. The slower it takes, the better it usually is.

In fact, that should be Microsoft new marketing campaign: "At Microsoft Live, we make all our results from scratch and we don't pre-index anything. It does take a little bit longer, and we may not be the biggest search engine around, but that's just a sign we're focusing on delivering quality results -- not fast results."

Re:Punch your customers in the face, selectively

[mh1997]

If you ran an experiment where some customers had their orders delayed by a few minutes more than was necessary and had some kind of metric to determine their enjoyment of their dining experience, it wouldn't be so absurd.

When I was a teenager, I worked at McDonald's. One day, some corporate people came into the restaurant with stop watches and notebooks. They had people pulled from the cash registers, then had extra people put at the registers. It appeared that they were doing something along the lines of what you are saying and what Microsoft did.

Re:Alkamai?

[Phroggy]

There's no L in Akamai, but yes, Akamai mirrors the static content, such as images. However, dynamic content such as search results are still served by Microsoft.

I submit to you

[ghjm]

a web page more useless than a blank page.

http://havenworks.com/

Thank you, and good night.