Design of Study |
Recognising Need for a Study |
Defining a Research Question |
Data Access and Gathering + Data Cleanup |
Threat Identification |
Validation of Results |
Interpretation of Results + Transfer of Results |
Getting Industry Commitment |
|
Alignment with Business Goals |
1.
Study design
must fit company's reality. 2.
ROI ->
this is all that matters in industry. 3.
You need to
be open to adapt to company needs even if out of your expertise. 4.
Compromise:
industry needs vs. research needs. |
1.
The
relationship of empircal studies of software with
long term business goals. Study must be of interest to business. 2.
How to
'sell' an exploratory study e.g., survey state of practice to possible
participants. |
1.
A research
question should be aligned with at least one business goal. 2.
If you have
a research question find a business goal that answering the question meets
the goal. 3.
RQ should
describe feasible business challenges. |
1.
Sell the
study better; get more data. |
1.
Observe
business restrictions. 2.
Risk
management -> increased chance of ROI -> industrial trust. 3.
Can you
identify a threat and its solution? 4.
Industry
focuses on survival not clever and fancy. |
1.
Include
observations regarding business. 2.
Find a
process being repeated where method can be changed. |
1.
What
principles might be deduced to argue on behalf of interpretation? 2.
ARTC Methods
need to be Adoptable, repeatable, teachable, costable/use. |
|
Stakeholder Motivation & Commitment |
1.
Establishes
language of doing and real problem. 2.
Necessary to
avoid Validity threats. 3.
Manageable
study length especially if for a PhD. 4.
Are the
primary, secondary and tertiary stakeholders; are the latter two being
ignored? |
1.
Assess
company's Reality. 2.
See comments
on column on right |
1. Deal with real problem. Learn about target and their
technology in market - identify their needs. See my techico commercial
drivers. |
1. Danger of missing values.
|
1.
Will the
company collapse before the study ends? Personnel change in company can harm
study. |
1.
KISS ->
keep it simple for solutions for industry. |
1.
Technico-business
drivers: how to relate business needs to research. |
|
Challenges, Barriers, etc. |
1.
Consider the
difficulty factors. 2.
When reasearch should be done by outsiders? 3.
Technology
still mentioned as an issue year after year; who submits a paper next year on
how to overcome? |
1.
Trustability short time
for assessments. 2.
First thing
to do with a company build trust. |
1.
To find the
real problem. |
1.
Finding the
right person to talk to. 2.
So much
work! Often resources are lacking. 3.
Find
non-intrusive methods of data collection; use GQM! 4.
Dealing with
confidential data. 5.
Motivating
participants. 6.
How much
does using the correct statistical procedures matter? |
1.
Tradeoff
analysis. 2.
Building
'communities of trust'. 3.
To what
extent and in way should researchers' predispositions be acknowledged? 4.
Recognition
of conventional change management scenarios and practices. 5.
Vocabulary:
are we talking about the same thing? |
1.
Need to
conduct industry trials that are non-intrusive. 2.
May need to
find industry to do trials. 3.
Hard to
control variables in replication studies. |
1.
How will an
experienced practitioner see results: too complex, trivial, threatening. 2.
How to
publish results in academia. 3.
How to
consider confidential information? 4.
How to
foster tech transfer to professional, students? 5.
Problems may
be complext but solutions must be simple. 6.
'Good
enough' in inudstry is perfect, more than 'good
enough' is too costly. |
1. Learning industry's needs. |
Tips, Lessons Learnt, Solutions |
1.
Cannot
capture all context; that's ok! 2.
General vs.
particular; obstrusive vs. non-obstrusive.
3.
Older,
experienced researchers can often see a problem. |
1.
Fragment the
study in more platable parts. |
1.
Take into conideration technology is volatile. |
1.
If you have
a result that is better and a respected champion that uses it. 2.
How to
explain the area of interest to the interviewee without influencing the
answers. 3.
How to
record, make them available and cite interviews without transcribing. 4.
Recognise it will be
hard to predict adoption. |
1.
Continuous
industrial communication and feedback. 2.
Have more
than one 'champion' in company. |
1.
Add Bayesian stuff to
method to make it suitable for academic publications. |
1.
Try research
disclosure sessions; What we are doing, why it might
be useful. |
|
Feedback to/from Stakeholders |
1. Feasible and real design. |
1.
Problem
identification in industry, observations, analyses as from keynotes. 2.
Assessment
reports and issues/challenegs perspectives. |
1. When framing research questions ,
what ethical issues must be considered. Frame questions
if possible
in terms of industrial outcome. |
1.
Provide a
complete set of reports
also produce reports written for new research. |
1.
Be upfront
about threats of study to industry (from paper this morning. 2.
Should contextualise findings.
|
1.
Independent
industrial integration; they use 'it' even
when you leave. |
1.
Should be
easy to understand. 2.
ARTC. |
|
Industry Setting |
1.
Context
defines study not the opposite. 2.
Industry
settings should capture context and understand its effects. |
1.
Q==F(P;C) Question, problem, effectiveness, context. |
1.
Try to make
as much data operational as possible (with permission). |
1.
Recognise your target
may already be successful. 2.
The worst
threat is not matching industry setting. 3.
Researcher's
pet idea might not always be appropriate. |
1.
Find a
company committed to R&d.
2.
How to
decide what is proprietary research results and what is stakeholder's.
|
|||
Principles and Fundamnetals |
1.
See ISERN
Related reports. 2.
How about a 3.
handbook of SE esearch processes? 4.
Epistimiology in SE;
research needs attention 5.
Pizza
framework (see Runkel and McGrath 1972) helpful particularly for
teaching. 6.
Systematic
literature reviews; evidence-based SE (not much here) |
1.
For those seeking
venture capital isn't the first round seeking money for operational study |
1.
Problems in
the real world should lead to a theory which will help to identify RQs. |
1.
Research Q's
should be assured through different types of studies/ methods. |
1.
Study
results should be interpreted in context of a working" theory to add to
knowledge base. |
|||
Tool Support |
1.
Tool support
for study design is a need. 2.
I don't know
of any tool to help study design that would help. 3.
Can you use
existing tools? |
1.
Tool support
for data gathering and cleanup is a need. 2.
Essential in
order to have data analysed on time. 3.
Any
experience performing interviews involving text chat
tools like skype. |
1.
Prepositions
of case study dots + results. |