The art of estimating
Estimation of features is a critical part of the software development process, especially if you are using an Agile methodology. The ability to accurately estimate features is an often overlooked skill that every developer should have.
As a project engineer, I’m responsible for making sure that we set realistic project timelines and that we actually carry those out. Managing expectations is one of my main responsibilities. This is where Agile is so nice — as long as you do it right.
When I start on a new project, usually one of the first things that happens is meeting with the client or whoever is sponsoring the project. In these initial meetings, at some point I will probably be asked one of two things: 1) how long will it take to accomplish X? 2) how much can we get done before X (some date)?
Let’s think about what is happening here. Your project sponsor is asking you to define what is realistic! This is great! You have one chance to set realistic expectations, and that time is now. Of course, you have to be fair — if you try to sandbag and say that it will take 4 days to add a button on a page, you will probably get called out. But even worse is if you underestimate things, because you will probably be held to your estimates. Once you give the sponsor the project timeline, they will take it to their bosses and spread the word around. So if you don’t get your project done on time or if it’s over budget, then the sponsor is in hot water. Not good.
If you aren’t asked one these two questions (i.e. they say, “We need to have X done by Y.”), that can still be OK. But if what they’re asking for isn’t realistic, now is the time to do something about it. No one likes to be told that what they want to do can’t be done or isn’t realistic. But I guarantee that they will be happier knowing about that up front than when you miss your deadlines.
Things to consider when estimating
When we estimate on our projects, we break functionality into features. Our criteria for features is that they should (usually) be something that can be tested and usually takes anywhere from 1/2 day to 10 days. There are lots of ways to do Agile and people do Agile in different ways, but most people break things up into features in some way.
The biggest problem I see with estimates is that people frequently underestimate things, and that’s usually because they don’t think of everything that goes into developing a feature. When I estimate, here are all of the things that I take into account:
- Reviewing the specs
- Asking questions about the specs and chasing down answers
- Doing the technical design and reviewing it with others on the team
- Writing the code
- Writing the unit tests
- Manually testing the code
- Fixing however many bugs you expect to have
- Working with external servers, web services, and other things that might slow you down
- Extra time if there are lots of uncertainties
In some cases, there might be more things to take into account.
One line that we always say about estimates is that “all estimates are wrong.” This statement is true on a feature-by-feature basis, but not for the project as a whole. This is because your project sponsor is going to hold you to your estimates. Whether that’s always fair or not is debatable, but that’s the way it is. You may not have all of the answers when you estimate a feature, and that decreases the chance that your estimate will be correct. So we need to try and have as many questions answered as possible before doing the estimating.
Sometimes you’ll start work on a project and realize that you’ve underestimated everything. This is not a good thing, but it happens. When this happens, you should re-estimate the features again so that your estimates can be accurate.
You’re also going to have the not-so-fun situation where you have to let your project sponsor know that it’s going to take longer to finish than you originally thought. Obviously this is not going to be a fun conversation, but the time it takes to develop a feature is the time it takes to develop a feature. You can change that reality. So the sooner you can adjust the expectations, the better chance you have of refocusing the project and righting the ship.
Estimating is a great tool for tracking progress and managing expectations, and it’s a reason that I’m such a big fan of Agile. It’s also a skill that every developer needs to master.
Great post, Jon.
One way to help with the “being held to your estimates” portion of working with the project sponsors is to offer a range. A realistic range, but giving a range at the high level opens you up to get back with better details after the team does its estimates. I’m not saying we shouldn’t be held to them, but it’s easier to be successful with a 70-90 hour estimate than an 80 hour estimate.
Also, when agiling it up, when you encounter something that’s going to drastically effect your estimate, bring it up sooner than later. Like you said, your client may not like the news, but at least having the information allows them to plan for it. (Or they may like if the change brings you in well under your estimate.)