Time Tracking in Scrum

Question

Note : Before asking this question I did an exhaustive search, and found little bits of the answer in various other questions, for example:

• What is the best resource for learning Scrum?
• Scrum Process Management - tips, pitfalls, ideas
• Two questions regarding Scrum

However, I feel like this question hasn't been directly addressed (if it has, please let me know).

Do you track time in Scrum as a function of hours/days spent on a task, or simply whether that task is complete or not? Can you adjust those tasks and estimates?

Background: Our new VP of development came from a Scrum environment, and so we're all learning about the process, but one of the things he has brought with him is the concept of very carefully quoting estimates of actual hours each task should require to complete, with the intention of getting more accurate with our estimates over time: thus once a project has started we cannot add new tasks or adjust the hourly estimates on those tasks.

But it was my understanding that agile practices, specifically Scrum, were based upon the concept of tasks being buckets that store individual deliverable goals, and you add/remove/adjust them as the clients' needs evolve after each sprint.

I realize this could potentially be argumentative, but I assume that viewing Scrum as a process, only one of those concepts is the "correct" philosophy for that system.

Answer 1

Do you track time in Scrum as a function of hours/days spent on a task, or simply whether that task is complete or not?

I track the estimated remaining work. This is a must have information. Without this, you can't draw the Burndown Chart. Without the Burndown Chart, you don't know "where" you are, you don't know if your Sprint is still on track or not. This would make this decision tool pretty useless. Yes, the Burndown Chart is not a tracking tool, it's a decision tool.

Can you adjust those tasks and estimates?

Sure!

Actually, the team owns the estimates, nobody else, and it is the job of the ScrumMaster to guaranty that this principle is applied. This should already answer the question. But there are other reasons.

As I said, a Sprint Backlog and a Burndown chart are decision tools and should thus be representative of where you really are. If you hide the reality, if you are not transparent, these tools won't help you to take any valuable decision, they will be useless. Think about it, what's the point of having good looking numbers if they are useless? What's the point of having a "nice looking burndown" if it doesn't reflect the reality.

So, during a Sprint, team members should obviously update the estimations of the remaining work as soon as they can do it (upward or downward). If a task estimate was initially 6h but the team discovers that more work has to be done and that the task will actually take 8h, the team should update the Sprint Backlog accordingly. If someone spent 4h hours on a task that was initially estimated at 4h but still need 2h work, these 2h should be reported on the Sprint Backlog. If the team discovers a task that has to be done but that wasn't identified, the team must add this task and its estimate to the Sprint Backlog. And being not accurate in the start is not a problem, as long as you update the backlog with the knowledge gathered over time. The sooner you make these updates, the sooner you'll be able to adapt and take decisions.

That said, it can be useful to keep the "initial estimate" and to compare it to the "actual time spent to complete". But not for tracking purpose, only to help the team to make better estimates. Actually, I would advice to not do this if you are transitioning to Scrum. There are often many other impediments to solve, many other things to improve first when you are learning Scrum values and principles. And if you do it, beware of the Waterfall daemons. Be ready to fight them, they may come back very fast.

Answer 2

We track both the time spent working on the tasks, and the time remaining to complete them. The remaining time allows to determine the progress made during the Sprint, and to anticipate whether we will be able to achieve the Sprint goal. We update the remaining time for the tasks, adjusting it (sometimes increasing it) on a daily basis.

The time spent is - supposedly- for micro management. It also gives the team a chance to get some feedback on the accuracy of the estimates - and to get better at estimating - and to show how interruptions prevent the team to work on the Sprint backlog and therefore, slow it down.

In the Scrum process, individual deliverable goals are called Backlog Items, and can be seen as bucket of tasks. The Backlog Items are prioritized by the Product Owner, estimated by the Team, first as a whole and then task by task. Content, scope, priority and estimation of the Backlog Items can be revised.

We estimate both the Backlog Items and the tasks in time units (days or weeks for the Backlog Items, hours for the tasks) and we apply a focus factor (ratio of time dedicated to work solely on the Sprint tasks) to account for time not spent working on tasks to achieve the Sprint goal.

Answer 3

I don't know if our implementation is "correct", but what we do is:

• Have Backlog Items added, which we put an estimated complexity number on (in relation to other backlog items).
• Before each sprint, we go through the backlog items in priority order (prioritized by the product owner), break them down into tasks for which we make a time estimate (in hours).
• When the number of available hours in the sprint are used up, the sprint is full

Then, during the sprint after each day of work we adjust the times on the tasks that we have been working on, so that they show the number of hours that we think is left before the task is done. This means that if I have a 6 hour task, work on it for a full day (we consider 6 hours a full day) and then feel that I still have 2 hours left before it's done, then I take down the "hours left" from 6 to 2. In case the task is time-boxed we need to check actual hours used instead, of course.

Answer 4

Estimate time, but don't really care if it's spot on

Just make sure you are careful and estimate tasks thoroughly. Basically you don't really measure time, because it's more error prone. The best way is to use tasks' time estimates as story points. This way you will gain:

1. If your time estimates are off, research shows that they tend to be consistently off (accuracy factor doesn't change too much), so time estimates can easily be used for story points calculation.
2. If you empirically managed to do x number of story points in previous sprint, you'll probably achieve similar results this time round even though your time estimates are incorrect.
3. You will have to be rather good at estimating all story tasks. Otherwise your sprint story points tend to grow during execution and you won't meet your deadline - even though your velocity will remain practically the same
4. Estimates can change but similar to #3, keep some sprint slack time for these changes to meet sprint deadlines (demo day).

But keep time estimates to actually see which tasks must split or join.

Answer 5

The answers I see here aren't wrong, but I don't think they've really addressed your question.

I think you're asking, "Should I track the total hours actually spent on a certain task?" The answer is, "You can if you need to, but it isn't part of Scrum."

Scrum is a very lightweight process. It defines/requires only what is needed to make Scrum work. You can (and, in many cases, probably should) overlay other processes on top of Scrum in order to suit your organizational needs. For example, if tracking the total hours actually spent on a task enables you to better estimate similar tasks in the future (as it seems your VP wants), then that might be a good reason to track total hours, provided that it doesn't interfere with productive work too much. Or, perhaps you need to know the total hours for billing purposes. So just because Scrum doesn't require something doesn't mean you shouldn't do it.

However, for the purposes of Scrum itself, there is no need to track the total hours actually spent on a task. It is not needed for any of the Scrum artifacts, which only track the estimated amount of time remaining.

Answer 6

With respect to time tracking, what you're looking for is a burndown chart.

Fredrik explained what a burn down is, without using the term. Essentially, you regularly reestimate the time remaining for a particular activity.

So to your question of whether or not we track time spent, not necessarily. Scrum likes to work with time remaining instead. (You could substitute hours with story points, the principle is the same, as Robert explained.)

To your second question of whether you can adjust your tasks and estimates, most definitely yes. Agile follows the 'reactive to change' philosophy; you prioritize what's most important to the customer.

However, some teams to prefer not to add/delete/re-prioritize tasks in a particular sprint once it's begun, since that is almost an ad-hoc way of working, and even scrum requires some structure and discipline.

The statement "thus once a project has started we cannot add new tasks or adjust the hourly estimates on those tasks." is almost certainly not in the spirit of agile.

Answer 7

We use the Pomodoro Technique to track the time remaining. One of its advantages is that the amount of time spent is recorded in a disciplined way.

After estimating stories in story points, we estimate tasks in terms of pomodori, and use this estimate (which may be reestimated ad hoc) to judge the amount of time remaining. At the end of the sprint it's easy to see which tasks we originally estimated the least accurately and improve how we estimate in the future, due to the way we mark the number of pomodori estimated and completed on each post-it.

In terms of the sprint, the estimated hours remaining are just a measure of progress so we can see where we are burndown-wise. They're a clue to whether we're on track or not. The score that matters is story points completed.

Answer 8

I have to add something here because

but one of the things he has brought with him is the concept of very carefully quoting estimates of actual hours each task should require to complete, with the intention of getting more accurate with our estimates over time: thus once a project has started we cannot add new tasks or adjust the hourly estimates on those tasks.

Is just plain not scrum so I don't know where your VP got his info. Tasks (know as Sprint Backlog Items) are not created until Planning the next sprint. They are created just in time and certainly not before the project starts. Before the project starts (Sprint 0), the Product Owner creates the Product Backlog and fills it with stories. He can add to it at ANY time during the project. It is his to manage. The team estimates these stories roughly against one another in story points or some other relative measure (ideal days?).

The estimating of tasks in hours is only a tool the team uses to figure out how many stories to commit to in the sprint and then to plot progress to predict success (burndown). Once a team has gelled and has a historical velocity; it may decide to not do any tracking in hours at all and just track their burndown in story points or # of stories. Estimating in hours is a form of waste in itself if the team does not need it to achieve commitment to the sprint goals.

I would ask the VP what these "very careful" estimates are going to accomplish.

Answer 9

By definition, an item is done when all of the tasks that need to be completed in order to fully implement that item have 0 hours left. What you need to track inside the sprint is remaining hours on remaining tasks. Not hours spent on a task. Why? Because our knowledge of how long something will take is imperfect and we gain little by trying to come up with a super-accurate estimate when we should be working on the product.

You are always allowed to add tasks under a sprint backlog item as you identify more work that must be done to fully implement the item, and you should update the remaining hours to completion daily (or set them to 0 once you've completed the task).

You should tell your VP that knowing when you're going to ship the product based upon your most accurate information (today) is far better than setting a number/making an estimate in the past and never updating it. This doesn't mean re-estimating user stories (don't do that until the end of the release), it means updating the sprint backlog with new tasks, and the best estimate as to when active tasks will be complete in remaining hours.

BTW, the way to work on accurate estimates is to plan your release using story points, create an iteration plan based upon your estimated team velocity, and then to continually update the iteration plan based upon the output at the end of each sprint. After a very few sprints you will get a very accurate idea of the actual team velocity, making it easy to forecast when you will ship your release with the desired scope... or what scope should be completed by the original ship date. Using actual project data from your current project to predict project completion is a software engineering best practice, because it is the most accurate way to make a prediction.