Académique Documents
Professionnel Documents
Culture Documents
Greg Hutchings
August 6, 2008
Greg.Hutchings@valtech.fr
Distributed Agile Teams and Contractual Forms
Greg Hutchings
I live in Paris and work for Valtech,
proposing, negotiating, managing and
living with large distributed agile projects.
2
Session outline
Distributed Agile Teams and Alternative Contractual Forms – What works best?
Building and supporting contracts to engage large distributed Agile teams (30 min)
Fixed-bid, T&M and pay for production contract alternatives (20 min)
3
Building and supporting contracts
Acceptance testing
Warranties / Reversability
4
Distributed Agile proposal process
Most new large company clients I talk to have already have had an offshore
experience, often not very positive or with mixed results, and have some fear
Quality, productivity, schedule and budget control are all key issues for a client
After an understanding of the client’s needs is obtained – high level scope, time
frame and budget, and specific constraints, a proposal is prepared
After the proposal is presented, reviewed, revised and agreed in concept, the
contractual development process begins
5
Selecting and defining methods in the contract
In our proposals and contracts, we discuss Agile, Scrum, TDD, CI, and the
importance of collaboration; we incorporate lessons learned from prior projects
Part of the Vendor role of developing business and ultimately writing and signing a
contract is to understand the key drivers, concerns and constraints of the client,
and another is to propose and coach success factors (agile adoption patterns)
I discuss on-shore and off-shore roles, those that might be staffed by the Client and
by the Vendor – with enablement / delivery trade-offs and relative priority
A key role to define at the client is an internal distributed agile champion – who
becomes elemental to the contract negotiation and project execution, and who
often becomes the Scrum Product Owner. This person should be able to travel.
The relative importance of knowledge transfer, cost, time-to-market and scope are
important to clarify, allocate resources to and define acceptance of in the contract
It is important for the proposal and then the contract to provide the team with time
to ramp-up, self-organize, perform an Iteration 0 and develop and validate sufficient
requirements / stories to begin Iteration 1
6
Types and structures of typical agile contracts
Structures
Pre-contract: verbal understanding, “hand shake”, email, letter of intent
Simple contract for professional services – no scope defined, rates and budget
Simple contract for software development – scope defined, total cost, timing
assumptions defined with terms and conditions to protect vendor and client
Hybrid contract – often a phase 1 (T&M or fixed bid) to define release backlog and
high level estimates, assumptions and to produce a phase 2 fixed bid for delivery
Fixed cost per unit of work. E.G. Valtech Software on Demand
In addition to the type and structure, the contract will need to define the term of the
engagement; it is useful to describe, plan for and gain commitment to events
Although we probably don’t have enough precision in our estimates, yet, the client
probably does have a budget in mind. After all, they are talking to you as a vendor
or IT team which likely follows a budget exercise of the previous year.
Based on a very rough sense of the work to be done, a capacity plan with a number
of iterations and a ramp-up in team size can be used to model what level of effort
the budget might cover.
8
Key events in a distributed agile Iteration
Week 1 Week 2
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10
Iteration Requirements
Est Workshop
Development Development
Planning Meeting Workshop
Design
Workshop
Development of test
Deployment of Development of test cases
cases
previous iteration
Retro-
Demo
spective
Testing Testing Prior Iteration
Build Build Build Build Build Build Build Build Build Package Package
Preparation of UAT Env Validate Env
A backlog of high
High level
level requirements
functional and
Release EM, PO, FM, Project is initiated, is prioritized to Release Backlog,
technical PO, BAs, Arch,
Planning EM PM, Arch, all France 1 day key participants permit planning of known definition of EM, PM, PO
requirements, EM coaching
Meeting BAs, Tech Lead are on board the iterations, project success
schedule and
overall release and
budget contraints
major milestones
Enough
information is Previous iteration
PO, FM, PM, Demo of previous available for scope "Done"
Iteration PO, BAs, Arch, High level
PM Arch, all BAs, France 20 days iteration is finalizing priorities status for each PM, FM, BAs
Preparation EM coaching scenarios
Tech Lead complete for the iteration scenario taken up
and agreement to
scope of iteration.
Enough
Proritized
information is
Iteration PO, EM, FM, requirement
High level available for High level
Planning PM PM, Arch, BAs, India 4 Hours FM, BAs backlog with BA + Customer
scenarios preparing scenarios
Meeting Tech Lead, Devs desired scope for
Requirements
the iteration.
Overview
By discussing in the proposal and engaging via the contract the commitment of client
and vendor to communicate regularly face to face, the budget for travel and the
commitment of key staff to the project can be planned.
10
Client and Vendor Roles and Responsibilities
An agile contract should clearly state what is expected of each party, and of
specific roles: who must do what, when, and sometimes how and where.
Normally the client must approve budget, approve requirements and approve
acceptance of delivered software, at a minimum, but it is nearly critical that they
participate in release and iteration planning, retrospectives and as product owner
using the Scrum role metaphor, be available daily, even at a distance, to clarify
requirements.
These may include transferring risk back from vendor to client (e.g. convert to time
and materials, assume capacity was consumed but wasted, etc.) or from client to
vendor (e.g. a quality or scope debt was incurred by the vendor for not delivering)
11
Roles and Organization Chart
Executive
Executive Steering
SteeringCommittee
Committee
Transversal
Transversal Committee
Committee
Operational
Operational Committee
Committee
Client (typical)
Product Owner Program Director
(Owns Initiative budget) France
Feature Manager
Domain experts France
Senior Developers
Development Team
Release Management
Infrastructure
The larger the Client’s delivery involvement is, the less likely that the
terms of a distributed agile contract can be fixed bid, and the more
likely that enablement will be the top priority. 12
Estimating functional and non-functional scope
In all Agile projects, estimation is very important, and is necessary to plan releases and
iterations
Estimation units of a team often need to be translated into terms meaningful to the client
The client’s units of measure vary, from story points, ideal days, actual days, use case points,
function points, use cases, stories, large functional specifications, “a system sufficient to
replace the existing”, etc.
Contractually, if the contract is not time and materials, something other than time must be
delivered and measurable in order to justify payment, and for distributed agile contracts the
more easily and exactly the units can be measured and verified, the more clear communication
will be and generally the better the relationship
Clearly, it is very important to define “What is done” with acceptance tests and to include these
with requirements when doing detailed estimates
Others in this conference, including Mike Cohn in his session yesterday, have treated this
subject in great detail, so I will stay high level
The main point to make for distributed agile contracts is that you should define the unit of
measure in a manner that is clearly described and understood in the agreement, and can be
managed, with incremental acceptance. We have found Use Cases, UCP and function points to
be useful measures in this regard.
Teams often forget to include estimates in what ever units they are using for non-functional
requirements in fixed bid contracts – or sometimes, even to adequately define these
requirements.
13
Acceptance Testing
Agile contracts or their associated proposals should define what acceptance tests are, who
writes them, validates them, performs them, where, and when.
I often contractually suggest that the client participate in the definition of the acceptance tests
during the preparation of stories or use cases, and that they be discussed and validated
consensually in the requirements workshop
Acceptance of the delivery of an iteration is specified contractually to be associated with the
passage of the scope’s related acceptance tests, and in a fixed bid contract or a pay for
production contract, this acceptance is often associated with the release of a payment.
A KPI I recommend tracking is the % of functional tests automated, and I prefer to state in the
contract that feature delivery acceptance is assumed within a short period of time (eg 10 days) if
automated acceptance tests pass and there is no indication by the client or team of other
reasons not to accept the delivery
UAT is usually, at least with our clients, a separate phase of time that includes some additional
risk in terms of time and budget, but which is absolutely necessary. We invite client
representatives to iteration end demos and provide access to the project dashboard to reduce
this risk.
Customer visits to the offshore development site can be coordinated with demos and
acceptance testing of the most recent iteration’s delivery if planned appropriately, and this
practice greatly increases the offshore team’s satisfaction level and the client’s confidence and
trust in the offshore team.
In distributed agile projects, the importance of comprehensive and automated functional and
non-functional testing is critically important to serve as an objective indicator of project quality
and progress. Tools such as Rally Dev, Version 1 and others provide information radiators to
keep client and vendor in sync, and their usage is often stipulated in our contracts.
14
Warranties and Reversability
Most vendor attorneys generally advise limiting the warranties provided by the
vendor in many important ways
However, clients wish to be reassured that the vendor is engaged and committed
and aligned with the client’s business needs.
These terms vary from country to country, and I have found warranties to be
stronger in Europe than in the US in many cases. A common European standard is
to guarantee that a custom app will be free of major or blocking defects for a period
of 3 months after delivery to production, and most often provides correction
services at no charge. There may be penalty clauses for the vendor in the event of
major defects that require liability insurance to be purchased.
Reversability clauses deal with the event that the Client decides to switch vendors
or in-source the application’s continued development or maintenance, by providing
for a knowledge transfer process supported by the vendor for a certain period and
with specific measures for completeness.
15
Case Studies
16
Case 1 : Employment agency application
Fixed bid -> bid per iteration
Re-Negotiation
The initial contract was bid for a fixed fee, with fixed scope, and though a desired time frame
was agreed, there was no penalty for late delivery
The onshore team was initially the sole point of contact with the client. They developed close
rapport, negotiated the contract, provided estimates and built the requirements and
architecture. The client wished to communicate in french and did not have face to face contact
with the offshore team.
The offshore team was not involved in validating the onshore architect’s estimates, and did not
feel committed to the estimates, the contract or to the client
Strained relations developed between the onshore and offshore teams, internally, and
eventually with the client as quality expectations were not met. The client seemed
unreasonable.
A “blow-up” occurred when the quality level was consistently not acceptable.
The parties met, including representatives from offshore, addressed some technical issues
related to human interface requirements, and renegotiated the budget and contract to permit re-
estimation and budget fixing by iteration
The now more collaborative and integrated team went on to successfully deliver the application.
18
Case 2 : eBusiness Catalog, eCommerce and POS
Agile contract evolution in large distributed agile program
Project was to implement an integrated multi-channel eCommerce solution with a new POS in 220 stores
Progressively replace features of 7 legacy apps with a custom Java / WebSphere Commerce Server solution
Distributed agile program began in France and shifted over time to large distributed team
Project size : 15 000 person days. Duration: 3 years
4 contractual modes :
• V1.C – Product Catalog, fixed bid, fixed scope – but no time boxing and 8 month delay!
• V1.S - Sales (Point of sale) with integrated Catalog and eCommerce site – T & M with bonus /penalty.
• V2 - Evolved Catalog and Sales for full production roll out – pay per productivity with velocity assumption
• V3 - Time and materials support agreement, capped budget, prioritized backlog
V1.C
Fixed bid duo-shore « agile » Time and materials with KPI – strong offshore T&M UAT
POC
V1.S
Negotiation
Negotiation
• Change project management, France -> India
• Rebalance French/India team and roles • V1.C delivered
• Create direct India/customer communication • V1.S negotiated, POC begun
• Quality and productivity indicators put in place • Revamped process, ramped team19
Case 2 : eBusiness Catalog, eCommerce and POS
Agile Contract Evolution…
V1.S V2 V3
BIG ramp up, T&M +/- Pay per UCP, Rising Velocity assumption T & M with a cap
20
Case 2 : eBusiness Catalog, eCommerce and POS
Agile Contract Evolution…
V1.S the T&M contract with bonus / penalty clause was in response to client demand for fixed bid and
vendor concern about estimation risk and acceptance speed.
This contract form had an effect on the team of very strong motivation to deliver completed use cases
(use cases which passed their functional acceptance tests), but sustainable pace was not maintained.
Morale was none-the-less very high.
During this period, January – March 2007, the team ramped up from 18 to 90+ members, from 1 to 4
delivery locations.
The team earned the 5% productivity bonus by delivering 99,5% of Use Cases, but had a 2.5% penalty for
quality based on integration tests
The planned 2.5 month UAT and Beta test period was extended to 3.5 months, and a larger team than
planned was maintained.
During the release retrospective workshop we determined to broaden integration testing and define
collections of use cases that together delivered potentially shippable increments and real customer
value, and focus on ways to increase efficiency and production. We also agreed to discontinue the
bonus / penalty clause.
Quality OK 2.5% 5%
Quality Productivity Productivity Penalty Penalty
5% Bonus +5% +2.5% 0
Defect
Person Day
Rate and
In-Budget Neutral 0 -2.5% -5%
Severity
Variance
KPI
5% Penalty -5% -5% -5%
21
Keep it simple?
V1.S
Formule de calcul :
∑ ((UC A / (UC A+UC Raf)) x UC Init)
V1A =
∑ (UC Init)
Inférieur ou égal à
V1 Feature Complete (V1 F) 70% 80% 90% 95% 100%
Bonus / Malus au 31/3/2007 -5% -2,5% 0% 2,5% 5%
V1 Acceptance (V1 A) 70% 80% 90% 95% 100%
Bonus / Malus au 15/6/2007 -5% -2,5% 0% 2,5% 5%
22
Case 2 : eBusiness Catalog, eCommerce and POS
Agile Contract Evolution…
V2
The client was ultimately satisified with the team’s delivery in V1.S, but felt that
productivity could be improved.
We agreed and together defined a budget for feature development estimated in Use Case
Points (UCP).
After substantial debate on whether velocity could be predicted, for commercial reasons
and based on some thin research on theoretical UCP productivity, a new contract was
agreed.
The contract stipulated that we would bill the client for UCP delivered, and in effect, over a
six month period, (6 four week iterations), deliver to the client a UCP productivity per
person that corresponded to the research.
The total UCP (plus some) engaged by this contract were delivered, but only after 9
iterations. The team became more efficient but not as much as predicted.
Substantial collaboration was necessary between client, business analyst, technical leads
and project management to permit planning and accepting the UCP for each iteration
This ultimately resulted in a high trust relationship – and the contractual eventually
evolved again, into… Time and materials.
The project was deemed a success – and also a substantial learning experience.
23
New contractual forms to consider?
- Valtech has introduced Software on Demand
24
What did the manifesto say about contracts?
© 2001 http://agilemanifesto.org
25
Conclusion
Agile contracts are especially important for clients and vendors when large
commitments are at stake
It is better for both parties to gain some experience with a smaller commit level,
and to inspect and adapt the contractual form
Direct communication between the team and client are essential to optimize the
learning and evolution of the relationship and related contract
26
Questions and Discussion?
27
Thanks!
Greg
GregHutchings
Hutchings
E-mail
E-mail gregoryhutchings@gmail.com
gregoryhutchings@gmail.com
Ligne
Lignedirecte
directe +33
+33(0)1
(0)153
5357
5773
7356
56
Mobile
Mobile +33
+33(0)6
(0)687
8725
2500
0058
58
28