In a few entries on my blog, I have referred to Velocity and only briefly explained what it is. I think it's about time I explain properly for those not familiar with it.

Velocity is terminology from the Scrum agile methodology and is basically the same concept as Earned Value in more traditional project management methods.

This is how it works...

1. Select a regular time period over which to measure Velocity. If you're using fixed Sprints or iterations, use that time period. Otherwise you can use weeks, fortnights or months. It doesn't really matter which as long as you're consistent.
   
   
2. Add up the estimates for all the tasks/deliverables/features in your chosen time period. It doesn't matter whether the estimates are in days, hours or even in relative points.
   
   
3. Only include the estimates for any items that are 100% complete and signed off within the time period. Anything still in progress counts as zero, as there is no value in incomplete work.
   
   
4. At the end of the chosen time period, the figure you have is your Velocity (or Earned Value).
   

You can then use your Velocity as a basis for your future commitments. As a result, it is self-correcting.

For example, let's say you estimate in hours and track your Velocity in 2 week Sprints. You know, therefore, there are 70 hours available in a Sprint, but find you tend to deliver a Velocity of 50 hours (because of under-estimating and other interruptions throughout the day). Tracking this trend will allow you to commit to your 'norm' of 50 hours per Sprint in future, because you know that's what you usually manage to achieve.

As a consequence of this approach, you don't need to be any good at estimating, and don't need to get any better at it. As long as you're consistently bad, you will still get better at delivering on your commitments.

In my experience, it's just as common for people to over-estimate by being too cautious. Velocity is also self-correcting this way around...

If, for example, you find you tend to reach a Velocity of 90 hours in your 70 hour Sprint, you are probably not a super-hero code warrier that eats problems for breakfast. You are probably just being too cautious in your estimating. In this case, in future commit to 90 hours. This might sound counter-intuitive, but go on. I dare you :-)

Although Velocity and Earned Value are project management techniques, why not know your own Velocity. If you do, you'll find you can gauge much more accurately how much work you can really commit to. Even if you're lousy at estimating!

And guess what? Everyone loves someone who delivers on their promises.

read more

One technique used by agile development teams is the idea of estimating product features using points. This has a few distinct advantages over estimating in physical units of time.

1. Estimating is very quick because it's an intuitive estimate of a feature's size.

2. An estimate in points indicates a feature's size relative to another, and does not give the illusion of being precise.

3. Over time, and using consistent iterations (e.g. 2-week sprints), you get a strong feel for how many points a team can deliver in an iteration. Even if a team is bad at estimating, as long as they're consistently bad, this makes a team's commitments self-correcting.

Some agile teams use Fibonacci numbers to do this. Fibonacci numbers are a sequence where each number is the sum of the previous two, i.e:

1, 2, 3, 5, 8, 13, 21 ...

Using Fibonacci adds an element of science to the concept of estimating using points, adding the laws of distribution as a dimension.

If you're interested in understanding more about the scientific background of Fibonacci numbers - which weren't invented for agile development by the way! - you might be interested to listen to this Radio 4 podcast on Fibonacci...

See also:
How to estimate your product backlog

read more

In step 1 - my first article in this series - I described 'how to get your backlog in order'.

If you've completed step 1, congratulations! Because it's the biggest step. And the foundation for all else that follows. Whether or not you implement Scrum.

If you haven't completed step 1, you must not go any further until you have.

So here's Step #2: How to estimate your Product Backlog...

High Level Estimates

You need to provide some high-level initial estimates, in order to get an idea of the size of your product backlog items.

This is helpful because it helps to inform the decision about priorities. And whether or not the features are likely to be worthwhile. And from a management point of view, gives a perspective of how big the team ought to be, commercials permitting.

But as yet, you don't know much about the items on the backlog. You don't know exactly what the features are meant to do. You don't know what tasks are needed to complete them. And you don't really know how you will implement them.

So you have to do a very high level, top down, indicative estimate. In fact it's a guestimate. Not an estimate at all really.

How many times have you heard someone say, 'don't worry, I won't hold you to it; I just need a rough idea'? And of course they do hold you to it. Of course they do!

Estimate Product Backlog in Points

The answer: Estimate your product backlog in points. Not in units of time.

Repeat: Estimate your product backlog in points, not in units of time.

No, I haven't gone mad. I know it sounds a bit whacky. But I'm going to ask you to trust me on this one; it does have its reasons, some of which will only become clear later in the series.

In the meantime, I ask you to accept that development teams are more readily able to give guestimates of 'size', without giving an estimate in time that they might be held to, and without having all the gory details.

So we're not asking the team 'how long will it take?'. We're asking 'how big is it?'

I also ask you to accept that Product Owners are more inclined to take this as a guestimate - as it's intended - and not as a premature commitment.

Now I realise that points could be seen as useless to a Product Owner in terms of making a business case for funding. Certainly until a team has a track record and we know roughly how many points they tend to deliver in an iteration. But I'll come to that later. Certainly, it is still helpful for prioritisation and to get across the relative size of a feature to a Product Owner.

Use a Points System

So what scale should you use for your points system?

Personally I like Fibonacci numbers.

For clever people, click on the link above to understand what Fibonacci numbers are all about. For simpler people like me :-) they're basically a sequence of numbers that some very old and very clever people (generally Italian, as per usual) have worked out have a slightly spooky, but very scientific significance. And this significance all relates to physics and the laws of distribution.

More about Fibonacci is really beyond the scope of this article, but they are a scientifically significant set of numbers, where each number is the sum of the previous two. They are:

1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610, 987 ...

For the sake of simplicity, when using these numbers for indicating size, I would suggest you use the range 1-21. Certainly for bug fixes and enhancements on products in the BAU (Business As Usual) cycle, this should give you sufficient a range. Maybe reserve 987 for that daft request you get sometimes to fly to the moon and back in an ice-cream carton :-)

The key here is about relativity.

A backlog item describes a feature. Maybe, for example, it's a report. You've done similar reports before, but it does have some complexity in the underlying data, so you decide to call this a 3.

Next on the backlog is another report. You size this one relative to the other one. Is it bigger or smaller. Clearly 21 is a lot bigger. 2 is a bit smaller. And so on.

To make sure the scale works for you, I suggest you start by picking what you think is the smallest thing on the backlog. Give this a 1. Then find the thing you think is the biggest thing on the backlog. Give this a 21.

Now you have your markers, size the backlog, working from the top, using the Fibonacci numbers.

When you get further down the backlog, you'll get to a point where the items are really rather fuzzy. And rather low priority. In fact you not sure you'll ever get to them in your lifetime. Please don't feel you have to size the entire backlog. Size enough of the items to see you through the foreseeable future. Remember it's already been put in priority order. So make sure you work from the top.

Estimate as a Team

Size your backlog as a team. There's a whole philosophy about the Wisdom of Crowds. Two minds are better than one, etc, etc. If there are big differences, use this as a discussion point to understand why. Did one person see issues and complications the other person didn't? Did one person see a nice simple approach that others didn't?

Consider playing Planning Poker. This is a fun technique to ensure that people don't influence each other. Each team member writes their estimate on a card, and everyone shows their answer at the same time. It helps to ensure less experienced members of the team are equally engaged and are not over-influenced by more experienced team members. It also helps less experienced estimators to learn from others. And it helps to avoid stronger, more vocal characters having too over-bearing an influence on the result.

From this exercise, negotiate the size of each backlog item as a team.

Review Priorities

Once you've sized up the backlog - or enough of it - ask the Product Owner to have another quick look at priorities. Maybe now they can see the relative size of the features they've asked for, they might change their view of priorities. 'Wow, if that's a 21, I'd rather have the other stuff first', or 'if that's only a 2, let's get it in the next release'. If any priorities are changed, simply move the item's position in the order of the backlog.

Stick with the Programme

Resist the urge to adapt this step. Ken Schwaber's book 'Agile Software Development with Scrum', which is highly recommended by the way, says this: If you are not yet an expert in a subject matter (e.g. Scrum) do not attempt to adapt it. Scrum is an adaptive process. But you are in no position to adapt it until you know it well, and have experienced it in operation.

Sizing up your backlog using points in one thing I would encourage you not to adapt until you've tried it, and tried it over a period of several Sprints so you can start to see the effects.

Next in the series: Step #3: Sprint Planning/Requirements

Follow this series by email

See also:
How to implement Scrum in 10 easy steps:
- Step #1: Get your backlog in order!
- Step #2: How to estimate your product backlog
- Step #3: Sprint Planning/clarify requirements
- Step #4: Sprint Planning/estimate tasks
- Step #5: Create a collaborative workspace
- Step #6: Sprint!
- Step #7: Stand up and be counted!
- Step #8: Track progress with a daily burndown chart
- Step #9: Finish when you said you would
- Step #10: Review, reflect, repeat...

'Implementing Scrum' PowerPoint Presentation

10 Key Principles of Agile Software Development

read more