Friday, November 05, 2010

Reverse-Tolstoy, or Tolstoy reconsidered

Rather famously Tolstoy started Anna Korenina with the words:

“Happy families are all alike; every unhappy family is unhappy in its own way.”

I like many others have repeated this quote, and even applied it to business and, in particular the act of creating software. I think, although I dread to check now, that I even open one of the chapters of Changing Software Development with the quote.

I dread because I have come to the conclusion that it it the wrong way around - at least for companies, and specifically the bits of corporations which produce software but which sell something else (i.e. Corporate IT departments) the quote should be:

“Unhappy companies are all alike; every happy company is happy in its own way.”

Blame (credit?) Jason Yip and his “Problems I know you have” blog post that sowed the seeds. But after a series of corporate clients in the last couple of years I’d like to offer my own list.

Here is a quick list of the mistakes I see big corporate IT departments making in the development of software. Actually, there are too many points here, there will have to be a part 2 to this post...

Too much Work in progress
Companies ask so much of their software development groups and these groups don’t say No. As a result they have far more projects in the pipe than they can ever hope to deliver in the supposed timeframe. Earlier this year I met a manager who had about 10 or 12 people reporting to him, he was trying to schedule them to about 17 projects. Any attempts to do anything other than run all 17 at once were an anathema to him and his managers.

This leads to....

Resource foobar
For “resources” read people. There are only so many people in the corporate IT department and in an effort to do work people are assigned to work on multiple projects. So #1 their productivity drops because they are now task switching, #2 quality drops because they are thinking of so many different things, #3 predictability falls because you don’t know when they’ll be working on which project, #4 arguments follow over who should be working on what.

At one client I analysed the (highly inaccurate) time tracking system and found some people worked on 14 projects in one week. Although about 54% of employees only worked on one project a week those who worked on more than one averaged 2.7 projects per week each.

Part of the problem is...

Resource (people) pools
People aren’t allowed to stay with the same piece of code for very long. Which means those who manage them must argue over what project they are assigned to and when. And when they aren’t assigned to a piece of work, but they are the only person who can fix a problem, then all hell breaks loose which destroys the predictability of the project they should be on.

Successful sports teams (think Manchester United) don’t break the team up at the end of the season. Could you imagine the manager saying: “Well the season is over, I don’t need you for a few months so I’ll assemble a new team then, goodbye.”

Unsuccessful sports teams (think Everton, I know I do) usually sell the best players at the end of the season and need to rebuild the team at the start of the next.

Sometime they are pulled off work because of....

Year long portfolio planning
Which means the company decides on some arbitrary date what it will work on for the next year. Which supposes #1 the company has perfect foresight, #2 nothing will change, #3 all projects will run as planned.

Question: How can you be Agile if you decide at the start of the year what you will be doing at the end?
Question: Did your company have a year long plan on Friday 12 September 2008? Was the plan still valid on Monday 15 September 2008?

One side effect of this is: demand for “architects”, “designers” and such is very high at the start of the year, while at the end nobody wants an “architect” but everyone wants Testers. So you create your own resource crush.

Too many Chiefs, not enough....
Big companies don’t like coding, its dirty “engineer” work. So that is all sent offshore or hidden somewhere. Which means they need people to manage it: Development Managers, Project Managers, Programme Managers, Work Package Managers, Test Managers, Development Leads, Test Leads, and their friends the: Architects, SMEs, Business Analysts, etc. etc.

Rule of thumb #1: if the number of coders and testers working on a project is less than than half you have a problem.
Rule of thumb #2: if you find yourself sitting in a room with other (none coders or testers) discussing the work of coders and testers with none of them present then you have a problem

Of course, that assume you have a room....

Too many meetings; too few rooms (projectors, teleconferences, etc.)
When you can’t have a meeting because you can’t get a room its a sign that either you have too many meetings, or.... no, its just a sign that you have too many meetings.

When a company saves money on projectors, teleconferences kit, video conference kit, etc. it saves money on the bottom line but people waste far more value faffing around with making do with inadequate resources.

No quality practices at the code face
And when you have too many chiefs you are likely to find that all talk about “Agile” is about Iterations, Stand-up meetings, User Stories and so on.

Wake up and smell the coffee: if you can’t get quality code done Agile isn’t going to save the day. Invest in TDD, CI, Refectoring, etc. etc. While some big companies do this many don’t - many small ones don’t either but they are not the subject of this post.

There are assumptions at work here: quality is expensive, quality isn’t achievable (all software has bugs doesn’t it?), quality is something developers do, or should do, either way, its something people with dirty under their finger nails are involved with. (I talked about this at the Agile Business Conference last month, Quality – How much quality can we afford?)

Even raising the quality question can lead to accusations that the questioner doesn’t “see the bigger picture.”

(Sometime when you dig into things developers are already moving in this direction but they are doing so without making a song and dance about it.)

Hierarchy
This deserve a blog entry on its own. Inherent in Agile as a whole, but not spelt out explicitly, there is an assumption of a equality. That we are all in this together, we all have a common goal, and we aren’t going to let job titles, length of service, age or other stuff get in the way.

Unfortunately most of the corporate IT groups I’ve seen in the last few years have various people who do stand on hierarchy.

Functional silos
Why, o why, do companies organise themselves as: Test group, Development Group, Analysis Group, and so on?
No one group can deliver a product on its own.

Once you have silos you have people (managers) who need to justify and protect their silo. Solutions which mean people are more integrated worry them.

Lack of knowledge / Knowledgeable people
Another reoccurring bottleneck is the lack of people who actually understand the systems and code base. This is sometime made worse by earlier redundancy programmes.

Kelly’s Knowledge Law #1: It takes time for people to become expert on a software code base and system
Kelly’s Knowledge Law #2: Knowledge transfer programmes don’t fix this any time soon
Kelly’s Knowledge Law #3: You are where you are because of earlier decisions not to recruit more people, not to train more people and to fire people

Don’t believe me? Check out Brook’s Law

And to make things worse....

Knowledge doesn’t accrue in offshore partners
The likes of Wipro, TCS, InfoSys and so on rotate their people every few years. So just as Sid is getting to be proficient in your code base he is moved to another client.

That is, of course, assuming Sid stayed with your supplier long enough to get knowledge of your systems.
Assuming he did: remember he works for your supplier not you

Which leads to....

Hollowing out your knowledge pool
In the early days of outsourcing some really stupid companies gave their staff to the new supplier. When they tried to take IT back they found the people and knowledge stayed with the supplier.

So today most (clever) companies keep the key people - people with experience, “architects” and the like. But, these people are growing older, and sometimes they leave of their own accord. Which begs the question: Are you creating the next generation?

How does someone get to be an Architect? (i.e. a developer who might code)
Or an Subject Matter Expert? (e.g. a system analyst who doesn’t code)

Answer: they get to know this stuff because they once cut code.
These people became Architect and exports because you decided their knowledge was too valuable to have leave the company or do dirty things like coding.

So what are you going to do when they move on? Or retire?

Remember: the thing that makes them valuable is a) they work for you, b) they worked for you for so long they know how it works.
Outsource the doing today and you loose you knowledge tomorrow.

I see these issues repeated over an over again. Now I’ve told you it isn’t rocket science so why do companies do to themselves?

As Tolstoy might have said: “Unhappy companies are all alike; every happy company is happy in its own way.”

I’ll continue this rant in another blog posting soon.