My first Publication Agile-Data-Warehouse-Design-eBook | Page 159
138
Chapter 5
Model Review and Sprint Planning
Use the data
profiling results to
rank the data issues
and estimate the
ETL tasks
Once the profiling results are in and have been added to the model it’s time to hold
an initial planning meeting (Figure 5-4) with the DW/BI team prior to stakeholder
model review. Armed with their new-found knowledge from running the data
profile checks, the team should rank the data issues by severity (see Table 5-2) and
provide ETL task estimates, in man-days, for loading the viable events.
Figure 5-4
Initial planning
meeting
Team Estimating
Estimating is an agile DW/BI team activity, every team member should be involved
to bring them up to speed with the emerging design. Everyone can usefully con-
tribute; e.g., BI developers can often help with ETL estimates if they are familiar
with the data sources, from having had to report directly off them in the past.
Play planning poker
to get unbiased
team estimates
Dimension and
event estimates
are added to the
event matrix
A downside of team estimating is that one person, who “knows best” or has the
loudest voice, can influence everyone’s estimate. A great way to avoid this is to play
planning poker: using a special deck of cards, everyone reveals their estimate for a
task simultaneously, and the team learns a lot from the differing opinions.
When task estimates have been agreed, the totals for each table are added to the
event matrix so that star schema estimates can be calculated by summing the
relevant dimension and event totals. These estimates, used in conjunction with the
team’s velocity (work delivered per iteration), will give stakeholders an idea of what
could be prototyped after the next sprint or delivered in the next release.
For information on calculating team velocity and estimating by playing planning
poker with agile teams, read Scrum and XP from the Trenches, Henrik Kniberg
(InfoQ.com 2007) Chapter 4: How we do Sprint Planning.
Review the annotated data model and task estimates with stakeholders as soon as
possible. Delaying the review can allow unrealistic expectations for the data
warehouse to grow. Don’t let the stakeholders dream for too long!