Saturday, July 23, 2011

Scrum Hates Technical Debt!

This is my phrase.  Ken Schwaber talks about Flaccid Scrum.  (Not my favorite metaphor.)  Jeff Sutherland talks about ScrumButt and the Nokia Test.  (I like this.)  Uncle Bob Martin talks about "the land that Scrum forgot."

People are doing Scrum in an unprofessional way, and then are not as happy with Scrum as I would like, nor than they deserve to be.  And often want to blame Scrum.  To me, this is not right in any way.  The most important thing: we want people to have better lives, and it ain't happening enough just yet.  Not for me.

So, allowing technical debt to grow is one key symptom of what I often call 'unprofessional' Scrum.

Now, we need a definition of technical debt, because not all of us know what it is.  Here is mine.  "All the things that we did or did not do, that are "in" our current product, that make it hard for us to change it quickly."  Examples: Lack of automated testing, reduced knowledge of the existing code (via many paths), duplications in the code, code complexity, code unreadability, lack of refactoring (at many levels), all the stuff we said we should upgrade 'real soon now', etc, etc, etc.  (I have assumed a software product, but the same concepts apply to any product.)

Here is Ward Cunningham's definition (and some other thoughts too).

My call to action on this (unprofessionalism and technical debt) is....

We engineers have to stop 'going along' when the business guys say 'you have to skip that stuff and just get more features out the door'.  We have to explain to them (and, to be fair to them, they don't know the facts) how we are only hurting ourselves (the firm, say) and the customers by 'going too fast'.  We have to explain it many times.

We business folks have to listen to the team, and learn how to understand technical debt.  And allow them to build quality.  As that book said a while ago, in essence, quality is free.  This is a difficult subject.  It is hard to understand, and some technical people give us too much BS.  But none of those excuses or problems give us a get out of jail free card on this important area of managing new product innovation.  We gotta do it.  Professionally.

Now, ok, I have to back peddle just a bit.
I agree that occasionally, close to rarely, there are situations where we should not 'do it right' and for the immediate release, we should just 'get it out the door'.  And THEN immediately go back and fix the technical debt.

And I agree that some legacy technical debt does not need to be fixed.  (For example, if we are never going to change that area of that system. Probably no need to fix it.)

And I agree that Technical Debt is ultimately a business problem. The business people must ultimately decide when and how much to fix.  But, they need the geeks to explain the details of the problem to them (typically they don't understand at all or very little).

Two fairly obvious things to say:
1. The bad news does not get better with age.  In other words, allowing technical debt to grow is not only unprofessional, and lying, but it is just plain stupid.  About 98% of the time (ie, so often that it is not worth asking 'isn't this an exception case?').

2. Scrum did not make you go fast.  Ok, ok, yes they are called sprints.  It sounds like we are in a rush. But pretending like the 100 yard dash is really the 90 yard dash is just unprofessional.  If the story is not done, done with a strong definition of done, then you're just lying about your velocity.

Three more fairly obvious things to say.
1. This is a hard problem.  It is almost impossible to keep technical debt from creeping up.  You will have to fight hard to keep things better.

2. It is worth the fight.  You can make your life, the lives of your teammates, the lives of your customers, better.  It a fight, a struggle, but actually fun to do as a team; and you will feel better for it.

3. Tools.  Yes, there are lots of tools, techniques, specific approaches, and patterns to follow to stop increasing technical debt and to reduce the technical debt.  But it starts with the human beings, the folks you work with, having the will to say: "We're fixing the problem with technical debt!  We have to!"


Vin D'Amico said...

I fully agree that too much technical debt is bad. At the same time, adding lots of extra effort to avoid any technical debt is equally bad.

As you point out, an area of code that is stable and unlikely to be revisited, probably doesn't need to have technical debt paid off. Another area of code that is heavily used and added to, should have it's debt kept to a minimum, if any.

In either case, technical debt should be documented and added to the backlog. There's no point in hoping it will just go away.

Joe Little said...

Hi Vin,

Good points.

In real life, I almost never see a team that does too much about technical debt. IMO. Only 90% of the time is it obviously not enough yet.

But I agree it is possible, and probably there are a few such teams out there.

Now, I completely agree that whatever one does, it must be done with some efficiency and some effectiveness. Not just pushing money or time at the problem thoughtlessly.

Thx, Joe

Unknown said...

Whilst I also agree technical debt is more often than not bad for business I think we as engineers striving to be the best we can be often forget that technical debt is a business decision (in fact I consider ALL of my decisions business ones)... It comes with costs and it comes with benefits.

I hate debt as much as the next developer. It's frustrating, annoying and feels so dirty. But sometimes it unfortunately just makes good economic sense!

Of course the tough part is in estimating the benefit of doing it right (aka the cost of not doing it right). What seems like x today often turns out to be 10x down the track. Add to that the fact that so much of that cost is hidden and I agree it's a brave man that makes the call "this technical debt is good for business".

Joe Little said...

Hi Tim,

I liked your comments very much. And agreed with them before you said them. In fact, your first point is to me so important that I revised my post.

As I already said, there are s few times to accept some technical debt. But I find very teams are rigorous in thinking about technical debt, must less acting on it. So, I tend not to go there first.