Does EVM apply in Agile

One of the best things that I love is interacting with people. When they ask questions, I always go back to the basic principles! That's what happened to me when a few learners in the corporate training I delivered asked if traditional EVM principles are even required in Agile principles. Here are some thoughts, first explaining the EVM concepts and then extending them to Agile.

Earned Value Management has two parts. The first part is lagging indicators that tell how the project is performing at a snapshot in time. The second part is the leading indicators that forecast how much more time and money will be required to finish the project! 

Lagging Indicators

This involves the Schedule Variance (SV) and Cost Variance (CV) expressed in currency units and Schedule Performance Index (SPI) and Cost Performance Index expressed as ratios. In order to compute any of these four measures, there are three things that are required. The first is the planned value (PV, sometimes also called budget at completion (BAC) or budgeted cost of work performed (BCWP)) of what we expected to compete, the second is the earned value (EV) of what we have earned at a snapshot in time and the actual costs (AC, sometimes also called the actual cost of work performed) incurred to generate the EV. Then, the lagging indicators can be computed as follows: 

Formula	Explanation
SV = EV - PV	If SV is zero, progress is on schedule If SV is -ve, progress is behind schedule If SV is +ve, progress is ahead of schedule
CV = EV - AC	If CV is zero, progress is on track If CV is -ve, progress is under budget If CV is +ve, progress is over budget
SPI = EV / PV	If SPI = 1, progress is on schedule If SPI < 0, progress is behind schedule If SPI > 1, progress is ahead of schedule
CPI = EV / AC	If CPI = 1, progress is on track If CPI < 1, progress is under budget If CV > 1, progress is over budget

Now, let us think about the applicability of lagging indicators in Agile. While the focus of Agility is on the team to manage itself, we need to ensure that the team is able to commit reasonably (and no external influence exists on the team to pressed for commitment) and self-organize themselves to deliver on the commitments (of course, there are always deviations). In the daily standup, when the team is marking completion of stories (task) for the user story to be ready for testing and marking validation complete (test cases) and the stories are accepted incrementally through the iteration, they are demonstrating earned value (velocity based on stories completed) towards the planned value (planned velocity).  There is always a cost to the iteration (Rajagopalan, 2019) representing the actual costs. 

So, if we have all the elements of planned velocity, earned value (actual velocity at the end of the iteration), and the actual cost (incurred at the end of the iteration), can EVM lagging indicators? The answer is Yes. But, the goal in Agile should be drive team's health, estimation accuracy, environment, and commitment for them to deliver.

Now, there is a question on the leading indicators. In traditional projects, there is a known scope of work and depending upon the lagging indicators, the estimate to complete (ETC) and the estimate at completion (EAC) can be evaluated. In agile projects, the product backlog is emerging constantly and the fixed scope of work is not always known. Nevertheless, if one looks at the agreed scope of minimum viable product for the release according the product roadmap, there is some anchor in the known scope of work within the MVP. If that MVP level stories are reasonably estimated (even if they were in the Affinity Estimation (i.e., T-Shirt or Coffee Cup)), then, it is possible to project the MVP backlog size. Then, for this known scope of MVP size, it is possible to apply the leading indicator concept from EVM.    

First, let us look at the leading indicators. There are four formulas depending upon the level of accuracy and confidence required. 

Formula	Explanation
EAC = AC + ETC	In this approach, accuracy and confidence are low. The project manager often approxmiates o the additional level of effort and cost. Sometimes, it could involve parametric estimates or expert judgment to come up with an estimate to complete.
EAC = BAC / CPI	This approach is slightly better because the focus is on the cost more than the scheule. The project manager uses the current rate at which costs accrue and uses this rate against the budgeted planned value (BAC) and projects the amount required. Since CPI will be less than 1, the EAC will be more than BAC and the difference is the additional amount required to deliver on the remaining scope. But, it doesn't take into account work already delivered (EV). It is possible to increase accuracy by using the slight revision existing work delivered must be removed. EAC = (BAC-EV)/CPI
EAC = AC + (BAC - EV)/SPI	Here the focus is on the team's ability to deliver on the remaining scope. Cost is not an issue, perhaps because the team members are salaried employees. In this case, the team's ability to deliver on the remaining scope of work at the rate the team delivers (BAC-EV)/SPI is computed and the actual costs already incurred are included.
EAC = AC + (BAC - EV) / SPI * CPI	Here, the focus involves both the rate at the rate costs have accrued as well as the rate at which the team can deliver on the remaining scope. The only difference from the previous formula you will notice is that the denominator includes a product of both CPI and SPI.

As you can see in the above explanation, forecasting (i.e., leading indicators) is a business need. Regardless of whether a team is focused on plan-driven or change-driven approach, the business always has the question of how long it will take for you to deliver on the remaining scope or MVP. 

If the lagging indicators (SPI, CPI, EV, AC) exist, and BAC (PV) is focused more on the remaining backlog size for the MVP, is there a reason why EAC calculations do not apply in Agile? The answer is Yes. In fact, Agile uses the burn rate to compute the rate at which the team burns through the backlog. Furthermore, there is one more leading indicator that people compute. That is To-Complete-Performance-Index (TCPI). This is computed as follows. Does the work and cost remaining relevant to Agile? Then, why question whether EVM extends to Agile or not? 

• TCPI = Work Remaining / Cost Remaining
• TCPI = BAC - EV / BAC - AC

The following diagram further illustrates graphically the thoughts. If you look at the abscissa (X-Axis), I have put both the timeline (T5, T10, T15, etc.) for a phased traditional approach and R1, R2, and R3 indicate the releases that may be comprised of individual iterations (single cadence where increments have to be consolidated and delivered). So, Agile or Traditional, don't get mixed up on terms! Focus on the principles of delivering customer and business value. Use EVM as it applies in your own context. 

Dr. Sriram Rajagopalan's approach to EVM in Agile

It should, however, be pointed out very clearly that in Agile, we should measure a few more things. These include cycle time, lead time, WIP, and Backlog Size (Remaining work). This is because, as mentioned before, the scope of work is constantly changing in Agile because of the constant feedback mechanism and the notion of embracing change.

References

Rajagopalan, S. (2019). Agile iterations also involve cost. https://agilesriram.blogspot.com/2019/04/test-post.html