How To Implement Scrum in 10 Easy Steps - Step #2: How To Estimate Your Product Backlog
by Kelly Waters
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
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
28 September 2007 09:32
I prefer pseudo-Fibonacci set: 1, 2, 3, 5, 8, 13, 20, 40, 100. Somehow the big difference between 20 and 40, between 40 and 100 more clearly indicates that if item is that big, it probably has to be split up before actually starting to work on it
28 September 2007 15:23
I prefer the pseudo-Fibonacci set too because it suggests that the bigger the number, the less precision it has. I feel it makes more sense think about a 40 than a 34 or 55.
And to anyone interested in
learn more about planning poker I suggest this link.
11 December 2007 13:44
Mike Cohn's book, Agile Estimating and Planning, changed my life. We now do estimates in a fraction of the time, with more accuracy. Why? Because more minds uncover more risks. Perhaps the biggest advantage of planning poker is that the client (product owner) offers a very valuable overview of all of the features considered for an iteration up front, so everyone starts with the same vision. Then details, of course, are discussed between the product owner and the developer coding the feature.
27 February 2009 11:14
What is kindof confusing here is that in agile projects all the teams :Dev,QA and Doc and other teams if involved will decide the size of the feature (Not Only Dev and Product Owner) it is quite unlikely that the size decided by each team (Planning poker) would logically be the same as each team will think from their perpective , like dev thinks how much time it will take for coding so they come out with a size XXL(in this case it gets 21 points) as multiple layers of code\design patterns may be involved and QA thinks about QAing which should not be much and so on..so estaimation of a feature might become like a share market instead and I have personally seen this.
10 September 2009 14:28
Priorotizing backlog can be tricky though if one doesn't keep ones' mind open. Consider this: You have a list of priorotised backlog from the product owner (initial list). After estimating work item tasks usin Fabonacci numbers product owner decide to change priorities to have the small ones to go first. Now, what happens if the delivery of the items at the top of the list is dependent on the ones at the bottom of the list? Can we have a so called dependency backlog for the project?