Samstag, 18. Februar 2017

Blended Planning - The Secret Of Keeping Fine-Grained Short-Term Planning and Coarse-Grained Long-Term Planning Up to Date and Consistent


Dear People!

Since beginning of last year I have been acting as a project manager in a project creating industry's most complex integrated electronic control unit (ECU) to support level 3 autonomous driving in series vehicles. 'Level 3' means you can check your emails while your car pilots you - no need to keep your hands on wheel and your eyes on the street.

As autonomous driving and its technical challenges are paramount on their own, managing big projects brings its own challenges - this is what I want to talk about in this article.

Note: If you want to take he quick tour just read the section summaries next to the pictures and the Summary at the end of the article!


The classical way

As in any big project project management gets increasingly complex with the number of project team members, work packages, teams and project duration. And as in many projects we started to create our project plans using MS Project©. We planned about 1000 tasks in each project plan of each team. Right - one project plan per team, not one single plan for the whole project. The reason for this was most probably that we could not make use of the MS Project Server©. As in any Excel© sheet, simultaneous editing would have been impossible in a single plan.


Section Summary:

Classical project planning does not regard the fact that there is no sense in doing fine-grained long-term planning as there is a high risk of changes for work packages in the far future anyway. A more detailed short-term planning on the other hand is often omitted to not end up with a plan of different level of detail for short- and long term planning which could be regarded as quality problem of the plan.


With about thousand tasks per sub-team covering about two years of remaining project time we got a - so to say - reasonable long-term plan. Also the project cost per year could be computed directly out of the tool. So far so good. The main problem we ended up with was that crucial questions by the management could not be answered:

"What are your project team members working on right now and when will that be finished?"

Or:

"What did we accomplish in the last month / the last two weeks?"
Not only management asked those questions, also the project managers of the single teams had no concise current view of what the other teams were currently working on and the project team members themselves could not deduce an unambiguous work item list from the plan and were uncertain of what to do next.

Adding all fine-grained work packages for each team member would have spoiled the project plan and would maybe have made it impossible to maintain. And no interconnections to other work packages could have been modeled because of the 'one plan per team' approach.

So we ended up here with the following situation:
  1. We had a - at least somehow maintainable - long-term plan with milestones and the possibilities to deduce cost per month and year.
  2. We had no clear picture of the two, four or eight weeks ahead of us: What will we precisely do in this short-term time frame? What will the others be doing so that we can come up with a collaboratively created plan?
The real mess of this situation is the following paradox: We missed a fine-grained short time planning having at the same time a kind of fine-grained long-term planning. The long-term plan was subject to change with every project scope change - and needless to say: we had a lot of those scope changes. So we continuously adapted the long-term plan which we did not need in our daily work by at the same time missing the short-term plan we were in urgent need of - what a unfavorable situation!

After some time of discussion and the strong feeling something had to happen quickly we decided for the disruptive, the modern, the brave way - let's go agile!


The agile way

Over Christmas we changed to a two-week sprint approach based on Scrum© - a present if you will to the whole project team. We used the newly provided JIRA Agile© tool and - beginning with strategically chosen high level sprint goals - each project manager started crating a sprint plan for his team members. This was a huge challenge but went quite well - after about five sprints the new approach was calibrated we ended up with the following new situation:

  1. We had gained a fairly precise short-term sprint plan where inter-dependencies between teams were aligned via the 'high-level sprint goals' defined in collaboration with all major stakeholders at the beginning of the sprint planning.
  2. We could immediately make use of the JIRA Agile provided sprint reports - showing e.g. how many stories we completed and how many stories stayed open (i.e. our short-term performance).
  3. We could answer the questions 'What has been finished last sprint?' and 'What will we finish by end of this sprint?'
We did re-estimate the full backlog of the project and bulk imported thousands of stories into JIRA. So we had the full backlog in JIRA and could now again answer the question 'How many effort/cost we face until finishing the project'. So far so good!



Section Summary:

Sprint planning is a very good tool to plan the near future but maintaining a detailed sprint plan for the far future (e.g. for upcoming one to even three years) is wasted effort because any agile expert knows that the far future is prone to the 'cone of uncertainty' meaning that changes of today will tremendously shift the way we take in the future. Thus, maintenance of a fine-grained long-term plan will most probably never pay off.


After some time of double-maintaining the short-term sprint plan and the long-term MS project plan we decided to drop the MS project plan and only use JIRA as single source of data (or truth as I often put it). But this had negative consequences as you can imagine as creating a long-term plan in JIRA was no easy venture:
  1. We could not come up with a sprint plan until end of the project because it is a) huge effort to assign all the thousands of stories to sprints and b) even huger effort to manually reflect priority and scope changes in the sprint planning of sprints one year in the future.
  2. Without having the complete sprint plan we could not deduce milestones from our tooling where 80% of the stories were lingering in the backlog without a sprint assignment.
Quickly it became clear that we lost our long-term plan with the ability to forecast because manually maintaining a fine-grained long-term sprint plan was simply not possible - and of course not economic. We thought about stepping back to the MS project plan as a long-term plan but - having the modern JIRA Agile now in front of me - another solution came into my mind.


I did it my way

In deepest distress, I decided to go for - what I call - 'blended planning'. Why should it not be possible to achieve both - fine-grained short-term planning to handle the challenges of today and tomorrow and - at the same time - a coarse-grained long-term planning to deduce milestones and cost and make the necessary forecasts about the future (which is known to not be very predictable these days anyway).

So the approach I came up with is the following:
  1. For the next one to three Sprints the project manager (or product owner together with the Scrum team) knows exactly what his team is supposed to do - if not, we are facing another problem. So you project managers execute a normal sprint planning using e.g. drag and drop in JIRA Agile which is a feature I appreciate very much.
  2. The rest of the stories stays in the backlog, not assigned to sprints but assigned to your team members. Stories not assigned to resources are not allowed - of course you can create virtual resources if you need so.
  3. The secret now is to prioritize all work packages for all your team members - nothing more must be done.
  4. Then let a tool a) create a project plan and/or b) assign the stories to future Sprints automatically. 
  5. The JAVA© tool I implemented is using the JIRA REST interface and can a) generate a MS project plan with all the future milestones and efforts and b) assign all stories in the backlog automatically to a Sprint. The tool can be run after any change to automatically update the plan and sprint assignment again. The result is an always up to date and consistent short- and long-term plan.

Section Summary:

To gain a fine-grained short term plan AND a coarse-grained long-term plan without wasting maintenance effort we combine human-executed story-assignment for the short-term and tool-automated story assignment based on human-defined priorities for the long-term in a blended approach - everything kept in the same tool to have a single source of truth as basis of a consistent management information system.


This is what we gain with this approach:
  1. The short-term plan is human generated, precise and meaningful, It can be easily adapted to any scope change. It reflects the plan according to the current reality in the best way possible.
  2. The long-term plan is automatically generated based on meaningful priorities based on human judgement. We can live with the automated generation because the long-term plan will most probably change anyway - thus too much detail here won't bring us any further. Even worse, planning the far future in too much detail pretends that we can exactly know what will happen which is - most probably - not the case.

In case of any scope and/or priority changes we can easily update the backlog, re-prioritize the stories and re-generate the long-term plan.

Voila! Welcome to my world!

Summary

In this article, I showed the power of blended planning - which is always having a fine-grained up to date short-term plan and a coarse-grained up to date long-term plan with the minimal maintenance effort possible. In case of any changes regarding your project/product scope you manually adapt the short-term plan, fix the priorities for the long-term work packages/stories and re-generate the long-term plan.

From my experience this is the most productive way to get an always up to date plan with which you are able to answer all the questions usually stated by interested stakeholders:
  1. What are your doing right now and when will that be finished?
  2. What was finished in the last iteration (classical: month, quarter; agile: sprint)?
  3. What are the relevant milestones?
  4. What is the effort/cost per milestone/until project end?
  5. What is the work load of your people per milestone/until project end?
  6. Until when can you provide an up to date plan for the next steering meeting? ;-) 

If you did like this read please leave me a comment below!
If you want to see the tool in action write an email to zt@jwr.at.

Thanks for reading,
Johannes Weidl-Rektenwald.


*All shown pictures in this article are free from licenses following Creative Commons CC0.
*Picture source: https://pixabay.com



Keine Kommentare:

Kommentar veröffentlichen