Agile evm earned value management in scrum projects

AgileEVM – Earned Value Management in Scrum Projects
Tamara Sulaiman, PMP, CSM Brent Barton, CSM, CST Thomas Blackburn, PMP, CSM
SolutionsIQ SolutionsIQ InfoTech, Inc.
tsulaiman@solutionsiq.com bbarton@solutionsiq.com Thomas.blackburn@infotechfl.com
Abstract
Scrum is an Agile Project Management
Framework. This framework specifically focuses on
maximizing Return on Investment (ROI). Scrum,
however, does not define how to manage and track
costs to evaluate actual ROI against the vision. A
reasonable cost measurement that integrates with
Scrum would help provide an additional feedback
loop.
We adapted Earned Value Management (EVM),
using values defined in Scrum. The result is called
AgileEVM (Agile Earned Value Management) and is
a simplified set of earned value calculations. From
the values in Scrum, we derived a release date
estimate using mean velocity and from this equation,
generated an equivalent equation using traditional
EVM techniques, thus establishing the validity of
using EVM with the Scrum framework. Finally, we
used this technique on two projects to further test our
hypothesis. This investigation also helped us
determine the utility of AgileEVM.
1. Introduction
Scrum is an Agile Project Management
Framework. Many Agile project methods, especially
Scrum, emphasize incremental, multi-level planning
and discourage fully planning software projects down
to the finest level of decomposition in the early
stages. The basis for discouraging big up-front plans
is the amount of inherent ambiguity present in
complex software systems that makes fully defining
the requirements up front nearly impossible. Further,
Scrum focuses on frequently prioritizing
requirements to maximize Return on Investment
(ROI) [1]. In order to evaluate actual ROI, project
costs must evaluated and used to determine whether
release plans have sufficient business value or
additional re-planning is necessary. One challenge of
evaluating ROI is the Scrum framework does not
integrate costs and leaves this up to the implementers
of Scrum.
“Earned Value Management (EVM) [is] a
method for integrating scope, schedule and resources,
and measuring project performance.” [2] This
technique is used in many industries. If we can
validate a way to use this accepted practice on
Scrum, then we have the tools we need to help
provide balanced what-if scenarios for release
planning for scope, schedule and budget. EVM
techniques, however, assume complete planning of a
project to discrete work package levels, then
assigning cost and duration to these packages so
EVM’s utility has been questioned for Agile projects.
To address this, we define Agile Earned Value
Management (AgileEVM), a simplified set of earned
value calculations adapted from traditional EVM
using Scrum metrics. By demonstrating the
mathematical relationship between EVM calculations
and Scrum metrics we intend to prove that Earned
Value Management (EVM) is valid for use on Scrum
projects. It should be noted that while we believe that
these statements hold true for other Agile methods,
we focus only on Scrum projects because that is the
framework used in our research.
To demonstrate the validity and utility of
AgileEVM, we present three main areas: research,
AgileEVM development and implementation on two
projects. First, we review the available literature on
earned value methods to establish the validity of
EVM techniques on software development projects
and to document previous studies applying Earned
Value techniques on Agile projects. Second, we
define terms and a process for using AgileEVM on
Scrum projects and correlated EVM with standard
Scrum burndown methods. Finally, we introduce
AgileEVM in two very different Scrum projects so
we could validate our assumptions.
2. Research
The literature review revealed an abundance of
publications validating the use of EVM for traditional

project management, particularly on defense and
construction projects. In use since the 1960’s, EVM
is a well recognized project management technique;
included in the Project Management Institute’s
PMBOK [2]. We concentrated on publications
relating earned value to software development
projects in general, and relating earned value
specifically to Agile software projects.
In the Spring, 1995 issue of Acquisition
Quarterly, Major David S. Christensen and Daniel V.
Ferens published a paper, “Using Earned Value for
Performance Measurement on Software Development
Projects”[3]. The processes described for obtaining
and using EVM data are clearly based on a
heavyweight, high-ceremony waterfall software
development life cycle.
An early piece by Quentin W. Fleming and Joel
M. Koppelman, "Earned Value Project Management
– A Powerful Tool for Software Projects" [4],
validated the appropriateness of using EVM on
software development projects. In the
aforementioned article and in Earned Value Project
Management (2nd
edition), they reiterated their view
on the appropriateness of EVM for software
engineering projects [5]. The EVM methodology
described is based on exhaustive initial planning,
formal work breakdown structures, detailed work
packages associated with dates and dollars, and
detailed tracking and analysis.
The earliest mention of EVM together with agile
practices was a paper presented by Steven H. Lett of
Lockheed-Martin, at the European SEPG in 1998.
Mr. Lett’s paper, “An Earned Value Tracking System
for Self-Directed Software Teams” [6] describes an
earned value tracking system (EVTS) for self-
directed work teams (SDWT) that involved
traditional project management planning tasks.
Detailed reports from individual team members were
required on a regular basis and involved a metrics
engineer to collate and analyze the data obtained.
In the February, 2003 issue of PMForum [7],
Glen Alleman described a method for an Agile
project manager to use earned value techniques on
iterative projects. The method advocated the use of
‘testable requirements’ and 100% unit test coverage.
At the June 2003 Agile Conference in Salt Lake
City, Glenn Alleman, Micheal Henderson, and Ray
Seggleke presented a paper entitled “Making Agile
Development Work in a Government Contracting
Environment - Using Earned Value to Measure
Velocity” [8]. This large government-contracted
development project mandated earned value
management (EVM). In this paper they describe
introducing agility by using certain Agile methods,
including XP-like engineering practices.
In his book Crystal Clear – A Human Powered
Methodology for Small Teams, Alistair Cockburn
describes the similarity between a "burn up" chart
and an EVM chart [9]. He warns that EVM credits
tasks completed, not necessarily integrated code. He
concludes that the Agile burn up chart provides more
accurate information as to the state of the project than
the earned value chart.
We reviewed the threads from two very active
discussion groups: http://groups.yahoo.com/group/
scrumdevelopment and http://groups.yahoo.com/
groups/agilemanagement. Opinions generally
centered around two themes. There is much debate
about the meaning of the term “value” and confusion
between EVM, business value and economic value.
The second theme considers that EVM techniques
can be used on Scrum projects, but asserts that there
is no tangible utility to using them.
We found the results of the review encouraging,
validating the use of EVM on traditional software
development projects. We found articles describing
the use of Agile practices on software projects using
EVM. What we did not find was conclusive
evidence demonstrating the applicability and value of
using EVM on Agile projects.
3. Development of AgileEVM
Our implementation of AgileEVM concentrates
on measuring progress at the release level, rather than
at the sprint level or at the product level. We feel this
to be the most appropriate way to use earned value
management formulas on agile projects. It is true that
these formulas can easily be used to measure
progress of a project with multiple releases, but this
would require that the backlog of multiple releases be
identified and estimated. An example of where this
may be a good approach would be when each Sprint
has a production release and we want to use
AgileEVM for helping us with the entire product life
cycle.
In comparison with the requirements of
traditional EVM, AgileEVM leverages work items
that are integral to the Scrum process. We measure
progress at the end of each sprint when actual sprint
velocity and actual costs are known. In five key
terms below (see Table 1) Traditional EVM
applications are compared to AgileEVM applications:

Table 1: Comparison of EVM terms
Performance Measurement Baseline (PMB)
Traditional EVM The sum of all work
package schedule estimates
(duration and effort).
AgileEVM Total number of story
points planned for a release
(PRP)
Schedule Baseline - often integrated in PMB
Traditional EVM The sum of all work
packages for each time
period calculated for the
total duration.
AgileEVM The total number of planned
sprints (PS) multiplied by
sprint length.
Budget at Complete (BAC)
Traditional EVM The planned budget for the
release or project.
AgileEVM The planned budget for the
release.
Planned Percent Complete (PPC)
Traditional EVM What % complete did we
expect to be at this point in
the project? Can be a
subjective estimate, or a
calculation of the dollar
value of the cumulative
tasks planned to be
complete by this point in
time divided by the
performance baseline
AgileEVM The number of the current
sprint (n) divided by the
total number of planned
sprints (PS).
Actual Percent Complete (APC)
Traditional EVM The dollar value of work
packages actually
completed divided by total
dollar value of the budget at
complete.
AgileEVM The total number of story
points completed
(potentially shippable
increments) divided by the
total number of story points
planned.
Initial values needed for AgileEVM are
straightforward. The values in Table 2 provide the
ability to create an initial release baseline to measure
progress against.
Table 2: Initial release parameters
Name Definition
BAC The amount budgeted to be spent on the
release
L The length of time for each sprint
PS Total number of sprints planned for the
release
SD The start date for the release.
PRP Total points planned for the release
At the completion of each Sprint, we capture
four data points. This is sufficient to calculate Scrum
metrics and AgileEVM.
Table 3: Sprint data points
Name Definition
n Sprint number - starts at 1
PC Points Completed - The points of work
completed from the Release Backlog during
the Sprint
PA Points Added - The points added (or
subtracted) to the Release Backlog during
the Sprint
SC Sprint Cost - What was spent
Change control in Scrum is managed and
reviewed each Sprint cycle. Upon acceptance of the
work completed in the Sprint and the adjustments to
the release backlog, this can be considered a new
project baseline. No extra weight or ceremony is
imposed on the Scrum process except we explicitly
review the release plan in the context of the budget.
To calculate earned or planned value we must
have an accurate representation of actual and planned
percent complete. Story points, as described in Mike
Cohn's User Stories Applied [10], represent an
estimate of effort and are therefore a valid measure of
percent effort complete. This is especially true if the
release includes testable requirements (Alleman [9])
in the form of Stories that must pass and be accepted
by the Product Owner. We define actual percent
complete, APC, to be the ratio of Story points
completed to Story points planned. We further define
planned percent complete, PPC, to be the ratio of
Sprints completed to Sprints Planned. These are
summarized in Table 4.

Table 4: AgileEVM definitions
Name Definition
PRP
Planned Release Points. See (18) for
further discussion
RPC
Release Points Completed. See (19) for
further discussion
APC
Actual Percent Complete of Release. This
is the ratio of Points completed to Points
planned
PPC Planned percent complete.
We use the following EVM definitions and
equations from the PMBOK: [2]
Table 5: Standard EVM definitions and
equations
Equation Definition
BAC Budgeted Cost at Completion:
This is the initial budget for
the release
AC Actual Cost: This corresponds
to budgeted costs in PV for
release
PV = PPC * BAC Planned Value: (1)
EV = APC * BAC Earned Value (2)
CV = EV - AC Cost Variance (3)
SV = EV - PV Schedule Variance (4)
CPI = EV / AC Cost Performance Index (5)
SPI = EV/PV Schedule Performance
Index
(6)
ETC = 1/CPI *
(BAC – EV)
Estimate To Complete (7)
EAC = AC + ETC Estimate At Complete (8)
4. Demonstrating a Mathematical
Correlation of Scrum Metrics and
EVM Metrics
There is a correlation of the schedule forecasts
using Scrum compared to Estimate at Completion
(EAC) forecasts using EVM. The logical argument
starts with a presentation of the hypothesis and then
proceeds through a derivation of a Release Date (RD)
function. Following this, we derive the Release Date
using mean velocity of Sprints. Finally, we calculate
a release date using EVM from the mean velocity.
Inspecting this result, we validate a reasonable
measure for Actual Percent Complete using EAC,
thus establishing the correlation.
4.1 Hypothesis
Release date estimates using EAC calculations
provided by EVM correlate to Mean velocity
predictions provided by Scrum.
More specifically, let RDm stand for the
calculated Release Date using method m. If the
projected release date based on the mean velocity
(RDv) equals the projected release date based on
EVM (RDEV), then EVM techniques can be used on
Scrum projects to calculate EAC with equal precision
to traditional plan-driven projects described in the
PMBOK (ANSI/PMI 99-001-2004 standard) [2].
Restated, if
RDv = RDEV (9)
then EVM techniques related to EAC can be used on
Scrum projects.
4.2 General Equation of a Release Date
Deriving a release date RDm using any method m,
needs an initial position, let it be named start date
(SD). Then a release date is simply an offset from the
start date. This offset can be determined by the
number of Sprints and the Sprint length. From this
we can define the offset to be the number of sprints
multiplied by the duration of the Sprints. Thus,
LNSDRDm ⋅+= (10)
where N equals the number of Sprints and L equals
the length of each Sprint in calendar days. Note that
(10) assumes equal Sprint lengths. It is
mathematically simpler to limit this discussion to
equal Sprint lengths. Although unusual, Sprint
lengths are occasionally adjusted to accommodate
environmental situations. For these anomalies, one
can, in most cases, redefine the Sprint Length to be
equal to the Greatest Common Factor of the different
Sprint lengths without invalidating this method.
Upon inspection of (10), N is the only variable,
where SD and L are initial parameters that do not
change for a release. Thus, N is the single
independent variable. See Fig. 1: Graph of a release
date.
RDm = f(N) (11)
So RDm is a linear function dependent on N.
RDm = SD +L·N (12)

∑=
+=
n
k
kn PAPRPPRP
1
0
∑=
P
ipPRP
10/25/2005
12/14/2005
2/2/2006
3/24/2006
5/13/2006
7/2/2006
1 2 3 4 5 6 7 8 9 10 11 12 13Number of Sprints (N)
ReleaseDate(RD)
SD
0
f(n) = SD +L·N
Fig. 1: Graph of release date
4.3 Using Mean Velocity to Define the
Equation of a Release Date
Considering (12), we need to define N in terms
of an estimate of how much work can be done per
Sprint. Let v equal the average work completed per
Sprint, which is known as the mean velocity, or often
just velocity. The total number of Sprints N
decreases as velocity increases, so it is inversely
proportional to N. Let W = the total work planned to
be done in the release. Then the total work factored
by the mean velocity equals the total iterations N for
v>0.
v
W
N = (13)
For Sprint n, where n ≥ 1, let
N = n + n' (14)
where n' is the estimated Sprints after n. Also, let W'
be the work to be done in sprints n'. Then (13) can be
rewritten
v
W
n
′
=′ (15)
Combining (14) and (15), then substituting into
(12), we have





 ′
+⋅+=
v
W
nLSDRDv (16)
4.4 Using Mean Velocity to Determine
Sprints Left
Clearly, if there have been no completed Sprints,
an estimate of an initial velocity must be made. Once
underway, measures of actual velocity can be used to
estimate future velocities. Empirical measures, such
as "yesterday's weather," are effective for evaluating
what can be accomplished in Sprint Planning. For
release planning, however, applying the arithmetic
mean to determine the average velocity (work done
per Sprint) is a common practice.
To provide measures for work, we need to
introduce a valuation for sizing Release Backlog
Items (RBI). This allows us to measure how much
work is completed for velocity calculations. The
method for sizing these can be any numerical value.
A few examples are: Story Points, T-Shirt Sizes with
Ideal Team Days applied, Function Points and
Working Days
1
.
For convenience, let us define the units for these
numerical values to be points, abbreviated p. Thus,
for the items in the Release Backlog, there exists a
same-sized set P of point values, P = {p1, p2, p3,…; pi
≥ 0} where the total point value, named Planned
Release Points (PRP) can be calculated as
(17)
Before the inception of any Sprint, when n = 0,
denote the initial Planned Release Points: PRP0. For
each Sprint n, define any new RBI points added PAn
and the points completed PCn. Thus, the total release
points from inception through Sprint n can be
calculated as
(18)
Further, define RPCn as the total release points
completed through Sprint n. This can be calculated
as
∑=
=
n
k
kn PCRPC
1
(19)
Finally, let the number of points remaining at
Sprint n be denoted PRn. This can be found by
subtracting the Completed Release Points from the
Planned Release Points. Thus,
nnn RPCPRPPR −= (20)
We can now redefine W '≡ PRn from (15) , so.
v
PR
n n
=′ (21)
To calculate the number of Sprints left we need
to calculate N. Substituting (21) in (14), we have
v
PR
nN n
+= (22)
1 One needs to be cautious when using values with units such as
time or cost. These can lead to confusion and potential
misinterpretation or miscalculation.

Next, we need a more explicit way to express v
in terms of data we capture. Let n define the nth
Sprint where n>0 and let vn be the velocity of Sprint
n. Then the velocity for Sprint n is the number of
points completed in Sprint n. Thus,
nn PCv = (23)
The arithmetic mean of the velocity v can then be
calculated using the standard equation:
∑=
⋅=
n
k
kv
n
v
1
1
(24)
where n>0.
Substituting (23) into (24) yields:
∑=
⋅=
n
k
kPC
n
v
1
1
(25)
Substituting (20) and (25) into (22) generates the
following equation.
∑⋅
−
+=
=
n
k
k
nn
PC
n
RPCPRP
nN
1
1 (26)
We can simplify this equation by factoring n
from each operand and substituting RPCn from (19).






−+⋅=
n
n
n
n
RPC
RPC
RPC
PRP
nN 1 (27)
This can be immediately simplified to:






⋅=
n
n
RPC
PRP
nN (28)
From the description of Actual Percent Complete
(APC) in Table 4, we can construct APCn as the ratio
of release points completed (RPCn) and total points
(PRPn) for sprint n. Thus, let
n
n
n
PRP
RPC
APC = (29)
or
n
n
n RPC
PRP
APC
=
1
(30)
Substituting (30) into (28) causes the equation to
simplify to:
nAPC
n
N = (31)
Finally, we can calculate the Release Date from
(12) using this simplified formula:






⋅+=
n
v
APC
n
LSDRD (32)
4.5 Deriving a Release Date based on EAC
from RDv
If we can derive an equation in the form of (12)
from (32) for a release date based only on EVM
terms, we will have correlated it to mean velocity.
Considering (31), we can manipulate this equation as
follows.






−+⋅=
n
n
nn APC
APC
APC
n
APC
n 1
1 (33)
and













 −
⋅+⋅=
n
n
n APC
APC
BAC
BAC
n
APC
n 1
1 (34)
But from (2), EV = APC·BAC, so





 −
+⋅=
EV
EVBAC
n
APC
n
n
1 (35)
From (5), CPI = EV / ACn, or EV = CPI · ACn so






⋅
−
+⋅=
nn ACCPI
EVBAC
n
APC
n
1 (36)
From (7), ETC = 1/CPI * (BAC – EV), so






⋅+⋅= ETC
AC
n
APC
n
nn
1
1 (37)
or
( )ETCAC
AC
n
APC
n
n
nn
+⋅= (38)
But, from (8), EAC = AC + ETC, so
nn AC
EAC
n
APC
n
⋅= (39)
That is both (n / APCn) and (n * EAC / ACn) will
calculate total Sprints. Also, for Sprint n we have
equivalent values for determining APCn.
EAC
AC
PRP
RPC
APC n
n
n
n == (40)
Substituting (39) into (32),






⋅⋅+=
n
v
AC
EAC
nLSDRD (41)
Thus we have derived RDv solely in EVM terms
so, let us define RDEV such that RDv = RDEV. Finally,






⋅⋅+=
n
EV
AC
EAC
nLSDRD (42)
5. AgileEVM Test Implementation
With AgileEVM defined, and equipped with a
spreadsheet for tracking purposes, we were ready to
try this out on some test projects. We chose two very
different projects for the initial test set. The goals for
the test were:

1. Validating the correspondence of the
projected release dates by comparing mean
velocity and EVM analysis.
2. Determine if the EVM metrics provided
additional value beyond the established
Scrum metrics for making project decisions.
3. Analyze whether the impact of cost analysis
adds value to technical teams for making
good decisions.
Two projects using the Scrum methodology were
tracked using both the AgileEVM and the burndown
method. These projects have very different
adaptations to the Scrum framework, one utilizing
very short Sprints with a small team, and the other
being a single large team with monthly sprints. Both
projects used story points for estimation purposes. It
is important to note the larger team project is still
active at the time of this writing, however, we feel
the results of tracking the projects for nine months is
sufficiently demonstrative to report. The smaller
team, Project A, has recently been released to
production.
For both projects, a data gathering spreadsheet
was developed to collect the data and perform both
the velocity trend analysis and the AgileEVM
calculations on a Sprint-by-Sprint basis. This method
provided for lightweight data collection along with
easy comparison of the results between the burndown
metrics and the AgileEVM metrics.
Project A is a one-half million dollar web
application project. The team is developing a product
enhancement and therefore is closely monitored by
company management. The collocated team has five
to seven members and is experienced with the Scrum
process. Due to dynamic business requirements,
sprints are one week long. Both the AgileEVM and
Scrum metrics are communicated to the entire team
at the end of each sprint. Key management
stakeholders are presented with the Burndown charts
and AgileEVM metrics on a monthly basis.
Project B is a very large, externally funded
project. The project size is about three million
dollars for the first eighteen month phase. The
Project B team is much larger than usual with 15-18
members, and continues to add team members every
couple of sprints. This team is new to Scrum and
Agile development methods. They are also collocated
for the majority of the day. Sprints are 30 calendar
days long. Unlike Project A, the primary project
guide on Project B is the Scrum Burndown chart, not
the AgileEVM metrics. We were still able to validate
the correspondence of the data but circumstances
limited our ability to interpret decision support value.
5.1 AgileEVM Test Results
Our analysis of the data from both projects
demonstrated indistinguishable results between the
predicted release dates using the mean velocity
method and the earned value method (see Fig. 2 and
Fig. 3). This result was as expected; following the
mathematical analysis proving that these two
formulas are equivalent. We also tracked release date
projections based upon atypical calculations for EAC
[2]. We noticed that in the early Sprints, the atypical
calculations provided more realistic predictions. This
is consistent with the nature of discovery inherent in
Agile projects.
The velocity of both projects fluctuated each
Sprint. Project A had cyclic velocity with several
highs and lows (see Fig 4: Project A velocity).
Project B started stabilizing about Sprint 5 (see Fig.
5: Project B velocity). Because of the fluctuations in
velocity, the mean varies considerably from the
actual velocity for some Sprints; though comparing
the predicted release dates to the burndown charts
demonstrates a close relationship.
Project A Release Date Projection Com parisons
7/17/2005
9/5/2005
10/25/2005
12/14/2005
2/2/2006
3/24/2006
5/13/2006
7/2/2006
8/21/2006
10/10/2006
11/29/2006
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Sprints Com pleted for RD estim ate
EstimatedRD
Plan
Mean Velocity
Estim ate at Complete
EAC (atypical)
Fig. 2: Project A release date projections
Project B Release Date Projection Com parisons
3/24/2006
7/2/2006
10/10/2006
1/18/2007
4/28/2007
8/6/2007
11/14/2007
2/22/2008
6/1/2008
9/9/2008
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Sprints Com pleted for RD estim ate
EstimatedRD
Plan
Mean Velocity
Estim ate at Com plete
EAC (atypical)
Fig. 3: Project B release date projections
The combined burndown and performance index
charts provide evidence of the additional value
provided by AgileEVM. Consider Project A (see
Fig. 6: Project A Burndown, CPI, and SPI). This

project had a 30% budget buffer and the CPI index
clearly shows the erosion of this buffer that the
burndown doesn't recognize. Adjusting the planned
release date in Project A at the end of Sprint 21
caused the SPI to move from an index value of 0.7 to
0.97 (see Fig. 2 and Fig. 6). This aligned the planned
and projected release dates along with the AgileEVM
indicators.
Project B has been adjusting scope down. The
SPI and CPI is registering the potential ROI of
reducing scope in Sprint nine (see Fig. 7: Project B
Burndown, CPI, and SPI). This scope reduction is
demonstrated by comparing the flattening velocity of
Sprints six through nine to the indices (see Fig. 5:
Project B velocity). Note that the same result can
also be interpreted by evaluating the burndown graph
and the velocity. This action of reducing scope has
kept the cost and schedule constraints within current
expectations.
The data also showed strong correlation in the
cost information provided by AgileEVM and manual
planning calculations. Given the expected behavior
of the indices in traditional EVM, it is clear that the
CPI, SPI and EAC behaved as expected given the
budget, schedule, and scope.
Pro ject A Velocity
0
20
40
60
80
100
120
140
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Sprint
PointsCompleted
Velocity
Mean Velocity
Fig. 4: Project A velocity
Project B Velocity
0
50
100
150
200
250
300
350
400
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
S print
PointsCompleted
Velocity
M ean V elocity
Fig. 5: Project B velocity
An interesting aspect is the way the information
was used on Project A. The metrics providing the
most value to team members and management were
the CPI, SPI, and EAC. These metrics were
requested by both executive management and team
members alike. Given that significant change is
expected on an Agile project, it was necessary to
remind team members and management that the
forecast release date, percent complete, Effort To
Complete and Estimate At Complete are based on
what is actual “right now”. This is exactly the same
as with the burndown method.
Executive management was comfortable with the
AgileEVM metrics and found them useful in making
decisions concerning the direction of the product.
Reviewing this data helped them validate decisions
and the need for process change.
5.2 The Agility of AgileEVM
A key question of the usefulness of AgileEVM is
the level of agility. Would the implementation of the
method add "drag" to a Scrum team’s velocity? If so,
would the impact to velocity outweigh the added
value of the information?
Project A B urn do w n an d Ind ices
0
100
200
300
400
500
600
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Sprint
PointsRemaining
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
IndexValue
Burndown
CPI
SPI
Fig. 6: Project A Burndown, CPI, SPI
Project B B urn do w n an d Ind ices
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Sprint
PointsRemaining
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
IndexValue
Burndown
CPI
SPI
Fig. 7: Project B Burndown, CPI, SPI
The question was posed to the team on Project
A: "Did the process of collecting or calculating the
AgileEVM metrics add weight to the project?" The
answer was a resounding, "No! The data used in
AgileEVM is data that is already available. The
calculation and tracking of the metrics was automated

in the spreadsheet that the ScrumMaster maintained.
Communication of the metrics to the project team
was as easy as posting the results on the wall, and
having the information available during the project
retrospective discussions."
Most of the developers reacted positively to the
availability of the data, feeling that if an analysis was
needed in order to determine a cause of a certain
issue, like a sudden loss in velocity, the data is
readily available to review. The team was able to
quickly identify changes in key metrics, and this
aided in the continuous improvement of the team.
The Product Owners for both projects felt that
the AgileEVM metrics did not provide any more
schedule insight than the burndown chart provided.
It is interesting to note that the Product Owners for
these projects do not have budget responsibility. We
agree that without the need to manage cost
performance, AgileEVM does not add significant
value above traditional burndown methods.
The ScrumMaster/project manager for Project A,
who does have budget responsibility, found the data
particularly useful for validating project status,
tracking and forecasting potential budget issues,
identifying alternative strategies by running "what if"
scenarios using AgileEVM to predict results, as well
as communicating status and issues with the product
owner and management. .
The ScrumMaster/project manager on Project B
did not use or report the AgileEVM metrics, nor were
the metrics shared with team members or
management stakeholders. Data was collected,
analyzed and compared to actual results without
impacting the team. The very fact that the process of
collecting and calculating the data did not impose on
the team at all gives further credence to the view that
the AgileEVM process passes the agility test.
6. Conclusions
In researching the validity and value of using
EVM on scrum projects several key questions needed
to be answered. First, are the metrics valid for Scrum
projects? Second, is the process lightweight? Third,
do the resulting metrics add value?
Assuming that the burndown trend analysis is
valid for Agile projects, we have shown
mathematically that the calculations for Release date
using Estimate At Complete and calculations for
Release date using the burndown trend analysis are
the same. Empirically, we have tracked the results of
two projects and have shown the indistinguishable
results of the data generated from both methods. We
feel that the validity of the AgileEVM techniques is
established.
The implementation of the AgileEVM process
has no noticeable impact on a Scrum team’s velocity.
Also, the value of the data was confirmed by the team
who had access to the metrics, as well as the
ScrumMaster and management stakeholders for the
project. Thus, we are encouraged that the AgileEVM
metrics do indeed add value to Scrum projects.
For the ScrumMaster, it is clear that metrics that
are familiar go a long way to ease the discomfort that
new, unfamiliar methodologies can induce. The
analysis that AgileEVM provides, along with the
burndown method, helps to substantiate intuition and
provides executives with quantitative data in a
consistent manner. The cost analysis, with its
forecast Estimate at Complete and Estimate to
Complete are valuable to Agile stakeholders
calculating estimated ROI. Agile stakeholders who
are responsible for making budget decisions find this
information extremely valuable.
Our recommendation is that AgileEVM be used
in conjunction with the Burndown chart and team
velocity as supporting data. One important caveat is
that change is expected on Agile projects and so the
AgileEVM metrics are derived from what is true at
each Sprint boundary.
Providing the team and Agile stakeholders with
useful and understandable data is vital to the "rudder"
with which the Scrum team steers toward better
processes and continuous improvement. By
providing the burndown and AgileEVM metrics
together, the team is better equipped to succeed.
References
[1] Schwaber Ken, Agile Project Management with
Scrum, Microsoft Press, Redmond, WA, 2004.
[2] “Guide to the Project Management Body of
Knowledge - PMBOK® Guide 2003 Edition”, Project
Management Institute, Pennsylvania, 2003.
[3] David S. Christiansen, Daniel V. Ferens, “Using
Earned Value for Performance Measurement on Software
Development Projects, Acquisition Review Quarterly,
DAU Press, Fort Belvoir, Virginia, Spring 1995, pgs 155 –
170,
http://www.suu.edu/faculty/christensend/EVonSWprojects.
pdf.
[4] Fleming, C.D., Koppelman, Joel M. Koppelman,
Earned Value Project Management – 2nd Edition, Project
Management Institute, Newtown Square, Pennsylvania,
2000.

[5] Quentin Fleming and Joel Koppelman, “Earned Value
Project Management – A Powerful Tool for Software
Projects”, Crosstalk - The Journal of Defense Software
Engineering, Ogden, UT, July 1998.
http://www.stsc.hill.af.mil/crosstalk/1998/07/value.asp
[6] Stephen H. Lett, “An Earned Value Tracking System
for Self-Directed Software Teams", European SEPG 98,
Conference Proceedings, 1998,
http://www.benchmarkqa.com/PDFs/software_teams.pdf.
[7] Glen B. Alleman, “Project Management == Herding
Cats, A Field Report, Agile Project Management”’
PMFORUM, PMFORUM.ORG, February 2003,
http://www.pmforum.org/viewpoints/2003/0203agilepm.ht
m
[8] Glen B Alleman and Michael Henderson, “Making
Agile Development Work in a Government Contracting
Environment – Using Earned Value to Measure Velocity,
Agile Development , Salt Lake City, Utah, June 25 – 27,
2003.
http://www.niwotridge.com/PDFs/ADC%20Final.pdf
[9] Cockburn, Alastair, Crystal Clear: A Human-Powered
Methodology for Small Teams, Pearson Education Inc.
Upper Saddle River, NJ, 2005
[10] Cohn Mike, User Stories Applied: For Agile Software
Development, Pearson Education, Inc. Boston, MA, 2004.
[11] Cohn Mike, Agile Estimating and Planning, Pearson
Education, Inc. Boston, MA, November 2006.
[12] Glen B Alleman, Nine Best Practices, The Software
Management Framework, 3 June 1999.
[13] David Christensen, “The Estimate At Completion
Problem: A Review of Three Studies,” Project
Management Journal, Project Management Institute, March
1993 pp24:37-42.
[14] David S Christensen, Ph.D., “Determining An
Accurate Estimate At Completion”, National Contract
Management Journal, Arlington MA, 1993, pp 25:17-25.

Agile evm earned value management in scrum projects

More Related Content

What's hot (20)

Similar to Agile evm earned value management in scrum projects (20)

More from JULIO GONZALEZ SANZ (20)

Recently uploaded (20)

Agile evm earned value management in scrum projects