[bind10-dev] Estimates for current sprint (ending 2011-07-12)

[Stephen Morris]

On 06/07/2011 11:50, Shane Kerr wrote:
> Jinmei,
> 
> On Tue, 2011-07-05 at 00:30 -0700, JINMEI Tatuya / 神明達哉 wrote:
>> At Thu, 30 Jun 2011 14:49:55 +0100,
>> Stephen Morris <stephen at isc.org> wrote:
>>
>>> With another sprint comes another request for estimates.  However, be
>>> encouraged: as a result of previous rounds of estimating, over half the
>>> tasks on the current sprint already have estimates associated with them.
>>>
>>> Could you please take a look at the attached tasks and send me your
>>> estimates by the end of Tuesday 5 July.
>>
>> I've just sent my estimates to Stephen, but I'm afraid I really don't
>> remember (or even understand perhaps) the consensus on the biweekly
>> call discussion on the effectiveness of the estimates.  So, please let
>> me ask a stupid question: what's the purpose of these estimates?  We
>> still don't seem to use them for planning (at least we didn't use them
>> for the most recent planning session).  Are the estimates now just for
>> records?  Or do we have a plan to (retry to) use them for planning as
>> well?  If so, how?
> 
> We do track our productivity:
> 
> http://bind10.isc.org/wiki/Burndown
> 
> We have used these in the past. I know when I was trying to fill in for
> Stephen I used the estimates to try to get a rough idea of how much
> stuff we could put in the sprint.

I agree with Shane.

However, it has to be admitted, at the moment we are mostly just
collecting data.  Ideally we should be doing the task breakdown and
estimates at the estimates at the start of the sprint, during the
planning meeting.  This would allow us to be realistic in what we can
achieve.  Unfortunately, previous experience has shown that this is not
easy when the team is spread around the world; it just takes too long.

instead, we have adopted a more pragmatic approach of trying to do a
task breakdown for the next few releases when we have a face-to-face
meeting and can discuss what needs to be done.  As estimating for that
number of tasks is something that we've found difficult in these
meetings (i.e. the team find it mind-numbingly boring after 15 minutes
:-), we've tried to break up the estimating into small chunks.  So far
we just been keeping head above water by estimating for the tasks in the
current sprint.  But at a recent meeting we agreed to try to be more
pro-active, which is why I'm now trying to get out two rounds of
estimating in each sprint. In this way I hope that we will accumulate
enough estimates so that we can use the information in our planning.

The "burndown" table has shown that our "velocity" is fairly constant,
ranging between 60 and 80 estimate points per sprint and with a figure
in the mid 70s being something we could take as being characteristic.
With this information, if we have enough task estimates up-front, we can
have a fair idea as to whether what we want to achieve in a sprint is
realistic.  This will give an early indication as to whether we can meet
release goals, which in turn allows us to set user (and management)
expectations appropriately.

Stephen