This is a very old version of an article that has since been extensively rewritten. The latest version is The Guerrilla Guide to Interviewing, Version 3.0.
This version is here for historical reasons only.
Hiring the right people is extremely crucial to Fog CreekSoftware. In our field, there are three types of people. At one end ofthe scale, there are the unwashed masses,lacking even the most basic skills for this job. They are easy toferret out and eliminate, often just by reviewing a resume and askingtwo or three quick questions. At the other extreme, are the brilliant superstarswho write lisp compilers for fun, in a weekend, in Assembler for thePalm Pilot. And in the middle, you have a large number of "maybes" whoseem like they might just be able to contribute something. The trick istelling the difference between the superstars and the maybes, because atFog Creek Software we only hire the superstars. Here are sometechniques for doing that.
First of all, the #1 cardinal criteria for getting hired at Fog Creek:
Smart, and
Gets Things Done.
That's it. That's all we're looking for. Memorize that. Recite it toyourself before you go to bed every night. Our goal is to hire peoplewith aptitude, not a particular skill set. Any skill set thatpeople can bring to the job will be technologically obsolete in a coupleof years, anyway, so it's better to hire people that are going to beable to learn any new technology rather than people who happen to knowSQL programming right this minute.
Smart is hard to define, but as we look at some possible interview questions we'll see how you can ferret it out. Gets Things Done is crucial. People who are Smart but don't Get Things Doneoften have PhDs and work in big companies where nobody listens to thembecause they are completely impractical. They would rather mull oversomething academic about a problem rather than ship on time. These kindof people can be identified because they love to point out thetheoretical similarity between two widely divergent concepts. Forexample, they will say "Spreadsheets are really just a special case ofprogramming language" and then go off for a week and write a thrilling,brilliant white paper about the theoretical computational linguisticattributes of a spreadsheet as a programming language. Smart, but notuseful.
Now, people who Get Things Done but are not Smart willdo stupid things, seemingly without thinking about them, and somebodyelse will have to come clean up their mess later. This makes them liabilitiesto the company because not only don't they contribute, but they soak upgood people's time. They are the kind of people who copy big chunks ofcode around rather than writing a subroutine, because it gets the jobdone, just not in the smartest way.
The most important rule about interviewing:
Make A Decision
At the conclusion of the interview, you have to be ready to make asharp decision about the candidate. There are only two possible outcomesto this decision: Hire or No Hire. Turn to your computerand send immediate feedback to the recruiter. The subject line should bethe name of the candidate. The first line of the email should be Hire or No Hire. Then you should spend about 2 paragraphs backing up your decision.
There is no other possible answer. Never say, "Hire, but notin my group." This is rude and implies that the candidate is not smartenough to work with you, but maybe he's smart enough for those losersover in that other group. If you find yourself tempted to say "Hire, butnot in my group," simply translate that mechanically to "No Hire" andyou'll be OK. Even if you have a candidate that would be brilliant atdoing 1 particular thing, but wouldn't be very good in another group,that's a No Hire. Things change so often and so rapidly that weneed people that can succeed anywhere. If for some reason you find anidiot savant that is really, really, really good at SQL but completelyincapable of ever learning any other topic, No Hire. They don't have a future at Fog Creek.
Never say "Maybe, I can't tell." If you can't tell, that means No Hire. It's really easier than you'd think. Can't tell? Just say no! Similarly, if you are on the fence, that means No Hire. Never say, "Well, Hire, I guess, but I'm a little bit concerned about..." That's a No Hire as well.
An important thing to remember about interviewing is this: it is muchbetter to reject a good candidate than to accept a bad candidate. A badcandidate will cost a lot of money and effort and waste other people'stime fixing all their bugs. If you have any doubts whatsoever, No Hire.
While you are conducting the interview, don't worry that if youreject a lot of people, Fog Creek won't be able to find anyone to hire.That's not your problem. It's the recruiter's problem, it's H.R.'sproblem, it's Joel's problem, but it's not your problem. Keepasking yourself which is worse - that we grow into a big, lousy softwarecompany with lots of coconuts, or that we stay small but high quality?Of course, it's important to seek out good candidates and everybodyshould see it as a part of their mission to find and recruit smartpeople who get things done. But once you're actually interviewingsomeone, pretend that Fog Creek has plenty of great candidates. Neverlower your standards no matter how hard it seems to find greatcandidates.
But how do you make this difficult decision? You just have to keep asking yourself during the interview: is this person smart? Does this person get things done? In order to be able to tell, you're going to have to ask the right questions.
Just for fun, here is the worst interview question on Earth: "What'sthe difference between varchar and varchar2 in Oracle 8i?" This is aterrible question. There is no possible, imaginable correlation betweenpeople that know that particular piece of useless trivia and people thatFog Creek wants to hire. Who cares what the difference is? You can findout online in about 15 seconds!
Actually, there are some even worse questions. I'll get to that later.
So now we get to the fun part: interview questions. My list ofinterview questions comes from my first job at Microsoft. There areactually hundreds of famous Microsoft interview questions. Everybody hasa set of questions that they really like. You, too, will develop aparticular set of questions and a personal interviewing style whichhelps you make the Hire/No Hire decision. Here are some techniques that I have used that have been successful.
Before the interview, I read over the candidates resume and jot down an interview plan on a scrap of paper. That's just a list of questions that I want to ask. Here's a typical plan for interviewing a programmer:
Before the interview, I am very, very careful to avoid anything thatmight give me some preconceived notions about the candidate. If youthink that someone is smart before they even walk into the room, justbecause they have a Ph.D. from MIT, then nothing they can say in 1 houris going to overcome that initial prejudice. If you think they are abozo, nothing they can say will overcome that initial impression. Aninterview is like a very, very delicate scale -- it's very hard to judgesomeone based on a 1 hour interview and it may seem like a very closecall. But if you know a little bit about the candidate beforehand, it'slike a big weight on one side of the scale, and the interview isuseless. Once, right before an interview, a recruiter came into myoffice. "You're going to love this guy," she said. BOY did thismake me mad. What I should have said was, "well, if you're so sure I'mgoing to love him, why don't you just hire him instead of wasting mytime going through this interview." But I was young and naïve, so Iinterviewed him. When he said not-so-smart things, I thought to myself,"gee, must be the exception that proves the rule." I looked ateverything he said through rose-colored glasses. I wound up saying Hire even though he was a crappy candidate. You know what? Everybody else who interviewed him said No Hire.So: don't listen to recruiters; don't ask around about the personbefore you interview them; and never, ever talk to the otherinterviewers about the candidate until you've both made your decisionsindependently. It's the scientific method!
The Introduction phase of the interview is intended to put thecandidate at ease. I spend about 30 seconds telling the person who I amand how the interview will work. I always reassure the candidate thatwe are interested in how he goes about solving problems, not theactual answer. By the way, in doing the interview, you should make surethat you are not sitting across a desk from the candidate. This creates aformal barrier which will not place the candidate at ease. It is betterto move the desk against a wall, or to go around and sit on the otherside of the desk with the candidate; this does help put the candidate atease. This results in a better interview because it is less distortedby nervousness.
Part 2 is a question about some recent project that the candidateworked on. For interviewing college kids, ask them about their seniorthesis, if they had one, or about a course they took that involved along project that they really enjoyed. For example, sometimes I willask, "what class did you take last semester that you liked the most? Itdoesn't have to be computer-related." Actually I am usually pretty happyif they choose a non-computer related course. Sometimes you look attheir schedule, and it looks like they are taking the bare minimumnumber of Comp Sci courses, but every elective is something related toMusic. Then they will tell you that their favorite course was ObjectOriented Databases. Yeah, right. I'd be happier if they admitted thatthey just liked music more than computers, instead of sucking up.
When interviewing experienced candidates, you can talk about their previous job.
In this question, I'm looking for one thing: passion. When you find a project that the person worked on recently, these are all good signs:
OK, the third thing on that list is the impossible question.This is fun. The idea is to ask a question that they have no possibleway of answering, just to see how they handle it. "How many optometristsare there in Seattle?" "How many tons does the Washington Monumentweigh?" "How many gas stations are in Los Angeles?" "How many pianotuners are there in New York?"
For programming questions, I ask candidates to write a small function in C. Here are some typical problems I would ask:
You don't want to give them any problems that take more than about 5 lines of code; you won't have time for that.
Let's look at a couple of these in detail. #1: reverse a string inplace. Every candidate I've ever interviewed in my life has done thiswrong the first time. Without exception, they try to allocate anotherbuffer and reverse the string into that buffer. The trouble is, whoallocates the buffer? Who frees the buffer? In giving this question todozens of candidates I found out an interesting fact. Most people whothink that they know C really do not understand memory or pointers. Theyjust don't get it. It's amazing that these people are working asprogrammers, but they are. With this question, here are some ways tojudge the candidate:
For #3, you can see how well they learned the bitwise operators in C.... but this isa skill, not an aptitude, so you can help them with these. Theinteresting thing is to watch them write a subroutine that counts allthe bits in a byte, then ask them to make it much, much faster. Reallysmart candidates will create a lookup table (after all, it's only got256 entries) that they only have to create once. With good candidates,you can have a really interesting conversation about the differentspace/speed tradeoffs. Press them further: tell them you don't want tospend any time building the lookup table during initialization.Brilliant candidates might even suggest a caching scheme where bits arecounted the first time they are used, and then stored in a lookup tableso they don't have to be counted if they are used again. Really, reallybrilliant candidates will try to devise a way to compute the table usingsome kind of a shortcut taking advantage of the patterns that occur.
When you watch somebody write code, here are some techniques that may be helpful:
Inevitably, you will see a bug in their function. So we come to question 5: Are you satisfied with that code?You may want to ask, "OK, so where's the bug?" The quintessential OpenEnded Question From Hell. All programmers make mistakes, there's nothingwrong with that, they just have to be able to find them. With thestring functions, they'll almost always forget to null-terminate the newstring. With almost any function, they are likely to have off-by-oneerrors. They will forget semicolons sometimes. Their function won't workcorrectly on 0 length strings, or it will GPF if malloc fails... Very,very rarely, you will find a candidate that doesn't have any bugs thefirst time. In this case, this question is even more fun. When you say,"There's a bug in that code," they will review their code carefully, andthen you get to see if they can be diplomatic yet firm in assertingthat the code is perfect... In general, it's always a good idea to askthe candidate if they are satisfied with their answer before moving on.Be Regis.
Part 6: the design question. Ask the candidate to design something.Jabe Blumenthal, the original designer of Excel, liked to ask candidatesto design a house. According to Jabe, he's had candidates who would goup to the whiteboard and immediately draw a square. A square! These wereimmediate No Hires. In design questions, what are you looking for?
Which brings us to #7, The Challenge. This is fun. Throughoutthe interview, you look for the candidate to say something that isabsolutely, positively, unarguably correct. Then you say, "wait aminute, wait a minute," and spend about 2 minutes playing devil'sadvocate. Argue with them when you are sure they are right.
Admittedly, in an interview situation, you are not equal parties.Thus there is a risk that the candidate will be afraid to argue with youbecause you are in a position of power over him. BUT, goodcandidates will tend to get fairly passionate about the argument, andthey may momentarily forget that they are in an interview, and they willget very involved in trying to convince you. These are the people wewant to hire.
Finally, you should ask the candidate if they have any questions.Some people like to look to see if the candidate will ask intelligentquestions, which is a standard technique in the interviewing books.Personally, I don't care what questions they ask; by this point I'vealready made my decision. The trouble is, candidates have to see about5-6 people in one day, and it's hard for them to ask 5-6 peopledifferent, brilliant questions, so if they don't have any questions,fine.
I always, always leave about 5 minutes a the end of the interview to sell Fog Creek. This is very important even if you are not going to hire them.If you've been lucky enough to find a really good candidate, you wantto do everything you can at this point to make sure that they want tocome to Fog Creek. Even if they are a bad candidate, you want to getthem excited about Fog Creek Software so that they go away with apositive impression of the company. Think of it this way: these peopleare not just potential hires; they are also customers. They are alsosalesmen for our recruiting effort: if they think that Fog Creek is agreat place to work, they will encourage their friends to apply.
Ah, I just remembered that I promised to give you some more examples of really bad questions to avoid.
First of all, avoid the illegal questions. Anything related to race,religion, gender, national origin, age, military service eligibility,veteran status, sexual orientation, or physical handicap is just illegal.If their resume says they were in the Army in 1990, don't ask them,even to make pleasant conversation, if they were in the Gulf war. It'sagainst the law. If their resume says that they attended the Technion inHaifa, don't ask them, even conversationally, if they are Israeli. It'sagainst the law. There's a pretty good discussion of what's illegal here. (But the rest of the interview questions at that site are pretty stupid).
Next, avoid any questions which might make it seem like we careabout, or are discriminating based on, things which we don't actuallycare about or discriminate based on. The best example of this I canthink of is asking someone if they have kids or if they are married.This might give the false impression that we think that people with kidsaren't going to devote enough time to their work or that they are goingto run off and take maternity leave.
Finally, avoid brain teaser questions like the one where you have toarrange 6 equal length matches to make exactly 4 identical perfecttriangles. If it's an "aha!" question, you don't get any informationabout "smart/get things done" by figuring out if they happen to make themental leap or not.
Interviewing is more of an art than a science, but if you remember the Smart/Gets Thing Doneprinciple you will be in good shape. When you get a chance, ask some ofyour co-workers what their favorite questions are and what kinds ofanswers they look for. In the Building 16 cafeteria in Redmond this is aperennial favorite topic of lunchtime conversation.