Build Your Own Data-backed Personas

[tuna]

Notice: This is a discussion thread for comments about the SitePoint article, Build Your Own Data-backed Personas.
__________

One major issue with Personas, they are not real and are often based on non existent people. Try explaining to a client that you have just spend x hours building personas that are pretend people that don't exist based on no real user research data. That my friends is a hard sell.

If you base the personas on more than the statistical data at hand and user research methods such as interviews or focus groups are used, then I have found that you often end up with more or sometimes less personas than you expected as they are based on real people. Yes this costs a little more but its easier to sell to marketing and the bean counters.

That's not to say the quick and dirty (read agile) personas generation method isn't good. It's just it will only be a guideline at best.

[AlexDawson]

I don't agree with the concept of personas because they stereotype people (usually wrongly) into groups which assume that they are likely to behave in a certain way, personally I think the grouping of people into split's of target audiance is about as wooly as those TV psychics, the psychology of humans is much too complex to just be reduced into such vague and misleading stereotypes. If it was that easy to simply push people into a single group or target audiance, I would be a millionaire by now from using psychological triggers to get a certain type of person to mass buy stuff.

[gary.turner]

I don't agree with the concept of personas [sic] because they stereotype people …

I think you've got it backwards. The idea is to eliminate the stereotypical by giving them personalities. Consider the personae developed in the article:

• the front-end coder
• the workaholic business owner
• the forum junkie
• the troublemaking teenager
• the newbie hobbyist
• the enthusiastic blogger
• the Googling geek
• the visually-impaired student

Now take the first one. The workaholic has more to him than just that. The idea is to flesh out the simplistic description; give him a set of interests and concerns. Make him real, so that when you consider his needs, and how you can serve them in your site, you have a mental image of a real person, not some one dimensional stereotype.

Do this for each group of people you have reason to serve, and you're developing for real people, not some stereotype; the opposite of what you are concerned about.

cheers,

gary

[AlexDawson]

Yes but the problem with trying to turn a figurative ideal of an audiance member into a living entity is that you are more than likely to place him in a stereotypical role, even though that character will have unique quirks you simply cannot place people into a labelled category and treat them as a breed, making assumptions about a certain character of an individual is a stereotype. While the idea is in the right place it doesn't draw away from stereotypes of an "average" user, it simply builds a physical presence for those individuals which is inherently more dangerous, your essentially trying to build an identity on the back of what you assume that person may be interested or concerned about however because that figurative individual was constructed by imagination it is seeing the individual only in a one dimensional process from the viewpoint of the flawed subjective perception of the creator of that character rather than the voiced concerns of an individual. If you want to know what a teenage rebal thinks and what they expect, ask one... don't create some illusionary figure who is moulded upon the perceptions and flawed realisation of some portrayel which has been built up over some construct or observation, it's the equilivent of a man creating a persona of a pregnant woman and then somehow claiming they have insight into the womans needs when clearly they aren't experiencing that reality so those needs aren't likely to reflect the same view as the woman.

[gary.turner]

Alex, go back and read the purpose for the profiles, and the methodology used to create them. Your objections are answered in the introduction to the profiles.

cheers,

gary

[codestyle]

Print style sheet is badly messed up in Firefox 3.5

[AlexDawson]

Alex, go back and read the purpose for the profiles, and the methodology used to create them. Your objections are answered in the introduction to the profiles.

Gary, I did read the methodology behind them and I understand why they would want to do it, but it makes more sense to me to get real peoples feedback rather than trying to profile people in such a generalistic way, it just doesn't hold a candle to the meat of feedback which is the people who actually use the services.

[mattymcg]

it makes more sense to me to get real peoples feedback rather than trying to profile people in such a generalistic way, it just doesn't hold a candle to the meat of feedback which is the people who actually use the services.

Fair points Alex. However, to play devil's advocate, it would be folly to suggest that conducting focus groups or in-person user research didn't also have its potential flaws—one of them being that what people say they do isn't always what they actually do. I'm by no means suggesting that the method I used here is perfect, but it's certainly true that anonymous analytics information is in fact based on real usage data by real people who actually use the services, and are not subject to trying to make a good impression or intimidated by the video camera over their shoulder watching their every move.

Sure, there may not actually be a visitor to sitepoint.com out there who owns a Persian kitty named Ruffles. But even if there was, it wouldn't help you, because that kind of detail is purely there to help the design team form empathy with the persona. It's backstory, but it doesn't really matter in the grand scheme of things. What matters is how the majority of your visitors use the site, and the data is a pretty conclusive way of obtaining that.

the psychology of humans is much too complex to just be reduced into such vague and misleading stereotypes

Like I said in the article, personas are not the be-all and end-all of the design process. They're just one of the tools to aid the process.

[mattymcg]

Oh, and one other thing—great points, everyone. I find this stuff very interesting, and this is a valuable discussion. Thanks for being part of it!

[mattymcg]

Off Topic:

Print style sheet is badly messed up in Firefox 3.5

Um, it looks fine to me, codestyle. I'm running FF 3.5.2 on a Mac. Could you PM me and include your operating system and a screenshot? Thanks!

[mattymcg]

One more comment that came to mind:

That's not to say the quick and dirty (read agile) personas generation method isn't good. It's just it will only be a guideline at best.

My question is, what assurances do you have that user research-based personas accurately represent the bigger picture? If one of the users you interview turns out to be an anomaly, how would you know?

Not suggesting user research and focus groups aren't worth doing (we're actually going to run some of these as well, plus in a way we are doing some validation how accurate these data-backed personas are by recording everyone's answers to the quiz). But I think that user research also comes with its own pitfalls, regardless of which is an easier sell.

[felgall]

My question is, what assurances do you have that user research-based personas accurately represent the bigger picture? If one of the users you interview turns out to be an anomaly, how would you know?

The only solution to that is to include all of your potential visitors into the focus group at the start. If you don't then you are just substituting the particular concerns of the people selected for your focus group for the concerns of the personas you create and so target your web site specifically at people like those in the focus group rather than at the particular types of people that you want to target it at (which you define using personas). Unless you plan on using millions of people in your focus group then you have to slot people into categories.

[johnyboy]

i was once told by an art teacher that the best bird book is an illustrated one, not a photographic one. on the face of it this doesn't seem right. but i think it might be. if you take the best illustrated bird book and the best photographic bird book the illustrated bird book will be better (that is more useful to people who use bird books to identify which birds they've seen in the wild) because any specific instance of a bird will have particular idiosyncrasies which others don't. a photograph of a bird can only be of one instance, unless you have a series of multiple photos, then you can have however many you want to show but that's going to get silly, showing 50 red shanks, 50 gannits, etc. be much better to have just one (or two if the two sexes are different, or four if juveniles and different but not many male adults or whatever). the person who makes the best illustrated bird book will study multiple (and i mean hundreds) of instances of one bird type to draw just one or a small number of drawings. their drawings will sum up everything they've noticed. if all or most instances of a type of bird show a particular trait -- which will vary -- they try and sum up and create the typical version -- the typical red splodge on the breast or whatever. it's a bit like averaging -- bearing in mind the average is, counter intuitively, often quite a rare occurance. and that's how and why a fictitious version can be closer to reality.

what i'm saying is a good illustration of a bird is like a good persona; sums up the basic characteristics papering over the individual oddities which actually get in the way and obfuscate the underlying basics/core of what's being got at/illustrated. you're not being led by stereo types, you're being led by observations, and you're summing up those obvservations.

> If you want to know what a teenage rebal thinks and what they expect, ask one... don't create some illusionary figure who is moulded upon the perceptions and flawed realisation of some portrayel which has been built up over some construct or observation

i haven't read the article yet, but i've read a few persona books. there's obviously different ways of putting personas together. data and research driven personas are based on doing exacly what you're suggesting; talking to the people -- many times. and then summing up what you gather into a simple easy to use, not loads of bind boggling data, representations.


in my limited experience of making websites for small clients what i find is in the absence of who their site's for (their prospective customers, current customers, whoever) the site gets made for unconciously the makers of it. in the absence of who you're talking to there's a basic human trait i think which subconciously substitutes themselves for the audience -- people are talking to themselves. but if the makers are different, and they always are in these situations, very different, that's disastrous.

such a redicously simple example which personas would have helped with (i.e. forced the website makers to consider who it's for rather than simply making it for themselves) is given in the 'letting the words go' book: an umbrella website for all types of london public transport. one of the main links/buttons on the homepage is (or was, presumably it's changed now) for 'oyster cards'. apart from londoners and people who have visited london previously, an 'oyster card' is pretty meaningless. i bet the clients and makers all lived and worked in london. therefore they knew exactly what an oyster card is, therefore no need to actually bother to describe it on the button. no one thought, hey, people who've never been to london won't know what an oyster card is even though tourists who have never visited london before are surely highly likely users of that site?! i mean how bloody obvious can you get? -- but that's the point. which ever body of people were the clients (not too far from government in this case, maybe the mayer's office) plus the web company who were presumably fairly reputable collectively failed to realise a button on the home page for oyster card is useless to people who live outside london and people who live outside of london are exaclty the type of people who'll be using this site!!! i mean the word morons springs to mind. but that kind of blatent **** up is common. and that's the type of thing i think personas help avoid.

[xpcomputers]

Surely the clever thing here, is that having created some "fake personas" with which to analyse the sitepoint website, they then created a "fun quiz" to see which one you are!

Really what they are doing with that is gathering more info surrounding the personas which will either validate them as a whole, or show them crucial missed areas or wrong assumptions they've made. If any of their percentages are wildly out, then this quiz will highlight it for them.

So these personas might get smarter and more accurate in their 2nd incarnation!

They've also asked if anyone thinks there is a missing persona, to report it.

As I say, I think the site point visitors feedback is the cleverest bit of this strategy...

It almost doesn't matter how accurate the first attempt is, although if they are too inaccurate, then users might not respond with empathy and therefore not be bothered to take the quiz. If they are too accurate, then individuals will instantly see themselves, and therefore not need to take the quiz to find out which they are... you need to be just "fairly accurate" to get the best response in the refinement stages (in my opinion!). Of course generalised grouped personas will never be VERY accurate, so most of us should see facets of ourselves in several characters, as we will be represented by several for different traits of ours.

eg I was pitched in the quiz as being the Workaholic Business Owner, which describes some of me, but misses vast chunks of reality that is picked up by other characters in the set. I could see how the personas as a group might represent my viewpoints quite well.

It is important to not get caught up in the detail, but to see them as representative of the breadth of views & personalities. The detail is just there to give empathy to the people.

I'd be very interested to see how the quiz validates or challenges the personas!

Mike

[AlexDawson]

My question is, what assurances do you have that user research-based personas accurately represent the bigger picture? If one of the users you interview turns out to be an anomaly, how would you know?

You do the same as you would in any psychological study, you take into account the wide range of results given and determine which fall under the norm's (using various analytical methods) to weed out the results which are anomalies. The problem with using personas forged from generalisations of your users is that you are building profiles of individuals who encompass stereotypicsl rules of various users, granted normal research isn't perfect and you do have issues of compliency however at least those variables can be controlled, there is nothing more uncontrollable as the bias which is evoked from data-backed personas where you are using reductionism to try and filter everyone into a box.

[xpcomputers]

Alex,

Surely they don't need to be anywhere need perfect, they just help with checking that the site makes sense to the wide range of users that will use it. It is only a vague reality check, but potentially a very useful one.

I think, whilst your ideas might have merit if the budget was limitless and the results required where needed to be more accurate... for most cases, these personas can be used to check whether a website meets the needs of various types of people, and are MUCH better than nothing.

For example:

There is a user from the UK, a user from Germany, user from India, a user from Aus, and 4 from US. It is easy to be designing a site assuming everyone has the same cultural background as yourself. You might use slang colloquialisms because they give the human chatty style your are going for, but forget that users from a different country might not understand. Having a few different countries represented gives you that. You don't need a persona from EVERY country that has visitors to your site. As soon as you are considering, "does the guy from India get this?", "does the girl from Germany get this?", you are thinking of specific examples, to help you connect with a user from ANY other country (and different first language).

Or another example:
The article lists 4 buying styles. All that is really happening is that the site designers are thinking through each of the buying styles to make sure that all 4 are covered, so that they maximise sales. This makes perfect sense. Having 2 personas with each buying style helps you to picture this is a more creative and visual way, and reminds you that other factors are at play, so that you can visualise real people using the service. Of course the 4 buying styles are distributed randomly, but that doesn't matter as long as all 4 are covered. It doesn't even matter if most coders or bloggers are stereotypically a different buying style to the one assigned in the personas, as they are just there to flag all the different aspects for consideration. If all 8 personas are considered, then all 4 buying style are covered, and the same for each other aspect.

They serve their purpose perfectly without needing to perfect representations.

Of course if you design you site ONLY for these 8 people (and for no one else), as though they are perfect representations, you might get it wrong, but no one is suggesting that is the approach to be used. Personas are a way of helping you connect with raw stats, and consider the needs of the people represented by those stats. As long as you know the personas are flawed, and are statistical generalisations, and more than that a bit of a random mix too (like which buying style goes with which browser, goes with which reason for visiting), then you will gain HUGE benefit to have them rather than not.

It is not a method for everyone. If it isn't for you, then you can use a list of:

• different browsers to design for
• different OS users to design for
• different nationalities to design for
• different buying styles to design for
• different technical competencies to design for
• different reasons to be visiting the site for
• etc etc

It is a bit more staid, but if that is your thing and works for you, then go for it. Personas is just presenting this information in a fun and approachable manner. There is no longer a mythical 1.5% of users from India (that can be ignored), there is a "real" person with a real name, who has real needs, who is harder to ignore. The fact that this "real" person is made up doesn't alter the emotional connection you can establish with his needs. That is a very powerful force for tweaking design changes.

I must say I have never used personas before, and may or may not in the future, but I can see their worth and their limitations, and feel they have huge value as flawed as they are.

Mike

[AlexDawson]

Perhaps it's just because I studied psychology at college so I am used to analysis under observational or measurable conditions? I do understand why they may be helpful indicators and I can see how they could help those less aware of the people who use the sites gain some insight through expanding their empathy but I guess I would prefer to do the full scale study using a mixture of statistics, observations and the natural empathy skills I have attained from years of listening to wish lists.

[mattymcg]

@xpcomputers: Mike, you are spot on, both for the ways that we intend to use the personas and for the value that we aim to glean by validating them against all of the responses that we get to the survey. Glad you agree that it's worth doing.

@AlexDawson: Don't worry, we'll be including research based on conversations with real people, too. Definitely agree with your comments on how much credence to place on something that is obviously going to include a bias. Thanks for your comments, it's all useful stuff to hear.

Matt

[AlexDawson]

@AlexDawson: Don't worry, we'll be including research based on conversations with real people, too. Definitely agree with your comments on how much credence to place on something that is obviously going to include a bias. Thanks for your comments, it's all useful stuff to hear.

What would be interesting is rather than prebuilding personas using the current formula to offer a form on the site where people could enter a bit of personal information (that helps you categorise them - so it's researched on real people), mixed along with their browsing habits of the website (all the statistical information) fed through an algorithm which calculates based on what they baught, browsed, viewed, replied too, etc would build a bigger picture and then merge their results with other browsers to build up a list of what could almost be described as "e-genetic traits" to categorise them

Might not make much sense but I was thinking along the lines of a social experiment which places people into categories which best suits their dispositions in the same way Google Adwords looks at what you browse for to gauge suitable advertising or like those dating websites which match your personality with others.

[Stomme poes]

Anyone who's ever done anything anywhere with statistics, questionaires and such research knows that people LIE. They do it repeatedly. They do it every time they can get away with it, especially when it invokes their egos. I never trust questionaire-based research on things like placebos, weight-loss programs, smoking, or browser use (not that they're not useful). Why the Nielsen rating found such a freaking huge difference in what people were watching when they switched from questionaires (people were watching Masterpiece Theatre and other Good Things) to a box that measures what people actually have tuned on (suddenly everyone was watching porn and daytime tv!). Never trust them. They LIE.
Why double-blinds are the top of the pile. Nobody on either side knows enough to lie.

In the other thread I thought the spammers and SEO junkies were left out. But they might just be terribly vocal minnority who also don't buy SP books, so maybe it doesn't matter.
I also noticed everyone lived in a big city, when SitePointers likely don't (hope this gets asked in the "further research"). This surely has impact for things like how many users have dial-up vs something faster, which should affect things like the popups and scripts on the site etc.

I suppose it's time to take the quiz to see what the thing says I am.

edit:

10. Which of the following best describes the way you decide to buy something?
I research everything about the product, weigh up competing products, and then eventually choose the best one.

I see something that catches my eye and, in the heat of the moment, I'll usually just go for it.

I check out whether the company selling the product cares about its customers, and if I feel like I can trust them and the product is good, then I buy it.

I know what I want in advance, and if something comes along that looks like it fits the bill, I'll grab it there and then.

I picked number one but I'm not really that anal. I also didn't see the option for checking the voodoo google entrails first.

Huh, I ended up being the blind student. Never mind that I answered that I was salaried and not a student, or that I was kinda forced to choose methodical in my purchasing when that's not terribly true (it often depends on the book/subject, and sometimes I'll choose #3(O'Reilly books) or #4(need to get x Drupal book).

He uses the Web and email extensively to conduct his grocery shopping, pay bills, purchase music, and connect with friends.

Sounds like a terribly online person who uses all the social stuff. I selected none of the reads, feeds, blogs etc on the blogs and newsletters questions (I don't participate in any social networking stuff unless you count last.fm).

I wonder if he was the answer I got because I chose the methodical answer in the last question, don't earn money on a blog, and am not a business owner. Are the other questions superfluous?

[mattymcg]

Huh, I ended up being the blind student. Never mind that I answered that I was salaried and not a student

Hmm, that's obviously an indication that I need to tweak the decision making behind the quiz. I don't pretend that there is any particularly clever intelligence behind which box you get put in, but obviously the combination of answers you selected meant that the points allocation didn't quite work out.

I wonder if he was the answer I got because I chose the methodical answer in the last question, don't earn money on a blog, and am not a business owner. Are the other questions superfluous?

Nope, actually that's the only question that doesn't influence the outcome (if you're likened to a persona but have a different buying behaviour from them, then it points that out separately).

I'm pretty happy with the combination of answers that are required to deliver each of the other personas, but the student is different from the rest for a few reasons. Basically, this persona exists for different reasons from the rest—we included Scott because we want our design team to be reminded of visually impaired users when making design decisions (whereas the others are included because they represent a statistical majority). So while it's easy to assign an arbitrary weighting to some personas because of their revenue streams or the amount of time they spend on the forums, the student is kind of the "I couldn't put you in any particular box so I'll guess you're this one" category. I don't pretend it to be particularly scientific—at the end of the day it is really just a bit of fun.

I don't pretend it's any more accurate than any similar quiz that you might do in Cosmo or Vogue magazine. That said, we can obviously do something about how your answer to the employment status question affects this persona. I'll take another look at those point allocations tomorrow! (about to head to bed).

Thanks for all the feedback guys.

[mattymcg]

Oh, and just one more thing, as I couldn't resist!

I don't participate in any social networking stuff unless you count last.fm

I guess the SitePoint Forums don't count for you as being social or "connecting with people" ???

Posts: 2,330

[Stomme poes]

Mm, I meant things like Twitter, MySpace, Facebook, LinkdIn, Flickr, other bloggitty blog blog blogs...

I consider forums to be social but I have them mentally in a completely different category than social-networking sites. : )

The student profile is a heavy, extensive web user, and there are lots of people like that. I'm half-Luddute : )

[AlexDawson]

Anyone who's ever done anything anywhere with statistics, questionaires and such research knows that people LIE. They do it repeatedly. They do it every time they can get away with it, especially when it invokes their egos...

Sorry but I don't agree, of course there are people who will lie and there are those who want to conform (especially on matters of sensitive subjects) but it depends entirely how the study is carried out to how the person will answer, you are making out like every person on the earth is at best a lying fraudster who is paranoid of their own input, it simply isn't the case. While there is a potential for individuals to undermine the study, if you are careful and sensible in how the information is collected you can avoid such issues to the majority extent.

PS: Forums are social networking sites, whether you think of them as such or not

[Stomme poes]

you are making out like every person on the earth is at best a lying fraudster who is paranoid of their own input

The differences between questionaire vs empirical measurements in studies that have gone from one to the other lead me to believe that this is true, even when not done intentionally or with evil intent.
*edit I said I'm on here 1-10 hours per week. Does that sound right?