Systems Analysis I Information Gathering

1
Systems Analysis IInformation Gathering

• ISYS 200
• Glenn Booker

2
Information Gathering

• There are two types of methods for gathering
  information about a system
• Interactive methods, which interact with the
  users of the existing system
• Unobtrusive (passive) methods, which observe
  aspects of the existing system
• Our goal is to understand the good and bad
  aspects of the existing system, to help determine
  requirements for our new system

3
An Aside

• Notice a couple of key assumptions here
• We assume that we are basing our system on some
  existing set of activities and/or associated
  information system(s)
• The existing activities could be automated,
  manual, or (most likely) a mixture of the two
• Part of the objectives for our new system is to
  improve on the existing activities in some way
• In order to understand the needs of the system
  were going to create, we need to understand the
  existing way things are being done

4
Interactive Methods

• Well cover three methods for gathering
  information interactively
• Interviews
• Joint Application Design (JAD)
• Questionnaires

5
Interviews

• Preparation is key to giving good interviews
• Look for your own biases, such as regarding
• The type of legacy system
• The educational level of the users
• The intelligence of the users
• The type of work being done
• And anything else which might be a factor

6
Interviews

• Make sure the purpose and scope of the interview
  are clear
• What kinds of information are you looking for?
• Whom do you need to interview?
• How much time do you need?
• Are the interviewees all in one location?
• Keep in mind that opinions are also important
  information, not just facts

7
Interviews

• Users may have strong feelings about whats good
  and bad about the existing system, which can feed
  the problems and opportunities analysis
• Also gives insight into the organizational
  culture
• Look for the users objectives or goals
• You want to interview to be a comfortable
  experience for the subject, but you need to keep
  control

8
Interview Preparation

• To prepare for an interview
• Read background material
• Establish interviewing objectives
• Decide whom to interview, e.g. people in
  different roles
• Prepare the interviewee call or email in
  advance
• Decide on question types and structure

9
Question Types

• Questions can be open-ended or closed
• An open-ended question encourages a long
  response
• What do you think about ?
• Whats your opinion about ?
• Please explain how you ?
• What are the biggest problems ?
• Could you tell me why ?

10
Question Types

• Closed questions encourage a Yes or No answer, or
  a very short response
• Who receives ?
• Do you like ?
• Do you want ?
• How many times a week ?
• Keep in mind they take little time to answer
  compared to open-ended questions

11
Question Types

• Open questions get more detail and help build
  trust with the interviewee
• Closed questions cover more subjects, keep
  control better, and are easier to manage
• Open questions may be followed up with a probe
  question
• Why?
• Can you give me an example?
• Will you please explain that for me?

12
Organizing Questions

• Just like logic can be inductive or deductive,
  the structure of interview questions can follow
  similar patterns
• The pyramid approach goes from specific questions
  to general ones (inductive)
• The funnel approach goes from general to specific
  (deductive)
• A diamond approach goes from specific, to
  general, then back to specific

13
Organizing Questions

• After the interview, write a report on it to
  capture key information
• Its common for someone to write down responses
  during an interview, the avoid losing key
  observations
• Compare results with the objectives of the
  interview
• If needed, contact interviewee for clarification
  or get missing info from later interviews

14
Joint Application Design

• Joint Application Design (or Development) (JAD)
  is uses a group setting to resolve some key
  aspects of the system often requirements and/or
  high level design
• Good to use when there are strongly conflicting
  views or priorities, or urgent need for a new
  perspective
• Otherwise generally not worth while, since it
  tends to be expensive

15
Joint Application Design

• JAD typically involves 8-12 people representing
  various users and technical specialties,
  including
• Executive sponsor starts and finishes the session
• At least one IS/IT analyst should be present, but
  primarily for a sanity check on possible
  solutions
• The session leader is typically an outside
  consultant, trained in leading JAD sessions
• Observers may attend
• A scribe is used to record results (take minutes)

16
Joint Application Design

• JAD sessions are typically held off-site (away
  from the normal office) to help people focus
• For design work, prototyping tools may be needed
• JAD sessions typically last 2-4 days theyre too
  intense to last much longer
• Attendees need 100 commitment to the session
• The scope of a JAD session needs to be clearly
  defined a formal agenda is common

17
Joint Application Design

• JAD can save time over interviews, reduce the
  cycle time, improve ownership in the decisions
  reached, and finally, it can encourage more
  creative design work
• OTOH, JAD requires a noticeable time commitment,
  it relies on good preparation and follow-up, and
  the organizational culture needs to support its
  approach

18
Questionnaires

• Questionnaires (surveys) can help determine how
  people in the organization are affected by the
  existing system
• Helps determine if a view found in an interview
  is widespread or not
• Or could be used to help identify problem areas
  before interviews are conducted
• If you use closed questions, the responses can be
  quantified

19
Questionnaires

• Questionnaires are good to use
• If the respondents are widely distributed
  geographically
• You want statistics on certain viewpoints
• You want an overview of the situation before
  further investigation, or
• If you want a survey of problems with the
  existing system to follow up with interviews

20
Questionnaires

• The questions used for questionnaires need to be
  more carefully thought out than for an interview
• You need to decide if open-ended questions should
  have some structure (e.g. the top three
  problems), instead of just an open question
  about problems
• If you want quantifiable information, you need
  to structure the response section to give
  consistent responses

21
Questionnaires

• Closed questions typically need to be answered by
  checking a box, circling a response, or entering
  numbers
• The choice of wording is very important
• Use local terminology
• Keep wording simple, but dont talk down to the
  audience
• Keep questions short

22
Questionnaires

• Avoid bias in wording
• Why do you hate the existing system?Like a
  psychiatrist asking Why do you hate your
  mother?
• Make sure wording is accurate
• Make sure the audience can answer the questions
  is it in the scope of their knowledge?

23
Scales in Questionnaires

• Consider how you are measuring responses
• What kind of scale are you creating?
• Do you find the help desk staff courteous?1
  Yes, 2 No
• How courteous do you find the help desk staff?
  (Rate on 1-100, 1 not at all, 100 extremely)
• Between those extremes, the 5-point Likert scale
  is common (e.g. 1 not at all, 5 extremely)

24
Designing Questionnaires

• The approaches for designing a questionnaire can
  follow the same principles as an interview you
  just have no flexibility to adapt it based on
  the responses
• Put questions important to the respondent first
  (why?)
• Cluster similar topics together
• Introduce controversial questions later

25
Questionnaires

• Administering (giving) a questionnaire can be
  done many ways
• Hand them out at a meeting, and collect them
  immediately
• Collect finished questionnaires in a drop box
• Provide instructions and return postage so they
  can mail the completed questionnaire back
• Email the questionnaire
• Post the questionnaire on a web site

26
Questionnaires

• Keep in mind that a questionnaire could be more
  valuable (objective) if its anonymous
• Dont include questions that would compromise
  that objectivity
• Who is your manager?
• How long have you been working here?
• What is your age?
• What is your gender?

27
Unobtrusive Methods

• Unobtrusive methods for gathering information
  include
• Sampling
• Investigation
• Observation
• These should be used in conjunction with one or
  more interactive methods

28
Sampling

• Sampling refers to selecting a subset of the
  whole subject (the population) in order to deduce
  information about the population
• In this context, we might want to know
• Which of our systems inputs and outputs we
  should examine? or
• Which people we should interview?
• We could also sample existing data to use for
  test cases

29
Sampling

• Sampling helps speed up data collection, thereby
  reducing cost but can also reduce bias and
  improve effectiveness
• To create a sample, need to
• Determine data to be collected
• Determine population to be sampled
• Choose type of sample
• Calculate sample size

30
Types of Samples

• A systematic sample just takes every nth person
  or record
• A simple random sample is the most common kind
  for polling
• Use some objective means to generate a random
  number, and include that person or record in your
  sample

31
Types of Samples

• Stratified samples look for types of groups in
  the population (e.g. Windows, Mac, and Unix
  users), then takes a simple random sample within
  each stratum
• A clustered sample is based on choosing some
  number of clusters of information (e.g. sites) to
  represent all clusters, and taking a random
  sample within the selected clusters

32
Sample Size

• Determine what attribute you want to measure
• Finding a sample size is weird, since you have to
  start by guessing what the answer is
• Select a value for p, the proportion of things
  which will have the thing you want (e.g. percent
  of people who will vote for X)
• For the worst case (and largest sample size),
  choose p 0.50

33
Sample Size

• Choose the acceptable margin for error, i
  (typically 3-5, a.k.a. interval estimate)
• Choose the confidence level in your result, and
  look up the corresponding z value
• Let sP i/z
• Then the sample size, n, is n p(1-p) /
  sP2 1

34
Critical Z values
35
Sample Size Example

• So if we want a worst case sample size for 3
  confidence interval, and 95 level of confidence,
  the simple random sample size is
• P 0.50
• i 0.03
• Z 1.96
• sP i/z 0.03/1.96 0.01531
• n p(1-p) / sP2 1 0.25 / (0.01531)2 1
  1068We need a sample size of 1068

36
Sample Size Caveat

• The previous formula for n assumes a very
  large population (in the 1000s)
• To correct for a small population, N, or if
  i is extremely small, use this
• n NZ2p(1-p) / Ni2 Z2p(1-p)

37
Investigation

• Investigation in this context refers to examining
  the existing system, both quantitative and
  qualitative documents
• Quantitative Investigation
• Qualitative Investigation

38
Quantitative Investigation

• Existing reports may be examined to determine
  several things
• What existing reports will need to be created by
  the new system?
• How has data been entered into the system?
• What are the current performance and process
  goals or objectives?
• How well have they been doing in meeting those
  goals?

39
Quantitative Investigation

• Examples of quantitative reports might include
  reports used for decision making
• Status of inventory
• Production quality
• Amount of sales
• Profitability
• Staffing reports
• These typically look for trends and problems

40
Quantitative Investigation

• Performance reports might compare actual
  performance to the planned or goal values
• Could look for completeness of such reports are
  all areas reporting their performance?
• General records can show individual data
  updates
• Such as payment records, invoices, timesheets
• Check for completeness, accuracy, and
  opportunities for improvements

41
Quantitative Investigation

• Data capture forms are generally the lowest
  level of data
• Great for understanding the start of business
  processes
• Look for unofficial and informal forms in use
• Document how the forms are used
• What is the business process flow?
• Who fills out each form?
• Who approves each form?
• Is their use mandatory?

42
Quantitative Investigation

• Check for consistency between the forms and the
  existing database
• Are some forms never used?
• Are there variations of the same form used?
• Are there multiple places to find a form?
• Data capture forms are often the most frequently
  used work products, so improvements here can pay
  off a lot

43
Qualitative Investigation

• Qualitative documents include any kind of
  relatively unstructured text memos, letters,
  procedures, fliers, web sites, etc.
• These often tell a lot about the company culture
  and document expectations
• Is the company portrayed as a machine? Family?
  Corporation? Team?
• Is there a sense of humor?

44
Qualitative Investigation

• Memos, for example, can tell a lot from their
  distribution and tone
• Whats most important to the organization?
• Safety? Profit? Happy customers? Happy
  stockholders? Quality? Market share?
• Signs posted in common areas can answer this

45
Qualitative Investigation

• What about their web site?
• What kind of metaphors does it use?
• Is it clear?
• Is it interactive?
• Look at their manuals, procedures policies
• Are they current?
• Are they used by anyone?

46
Observing Behavior

• Another key way to obtain information
  unobtrusively is to observe people doing their
  jobs
• Need to get their managers approval, and explain
  what youre doing
• Can observe people at different levels data
  entry people, analysts, managers, etc.
• Note this differs from the text

47
Observing Behavior

• At low levels of the organization, observation
  can help answer key questions
• How often are different types of tasks done?
• Are they following the stated procedures?
• What do they do when something goes wrong?
• In what order do they do things (e.g. data
  entry)?
• For managers, how do they make decisions?
• This can help outline procedures as they occur

48
Observing Behavior

• Can model behavior by treating each person as an
  actor, and summarize the steps each follows in
  their script
• How do they gather data? Read, observe, ask?
• What decisions do they make? Whom do they tell
  about those decisions?

49
Observing Environment

• Another aspect of observation is to look at the
  work environment
• The STROBE method refers to STRuctured
  Observation of the Environment
• Decision makers often define the work
  environment, so its helpful to see what kind of
  world they have created
• Like the director of a movie

50
Observing Environment

• What does their office location say?
• Do they have corner offices?
• Are they located with other managers, or with
  their staff?
• Does their desk placement encourage
  communication? Demonstrate power?
• Where do they store data? Do they have a lot of
  it or a little?

51
Observing Environment

• Do they appear to use their PC and/or PDA often?
• Do they get information from outside publications
  or the Web?
• Do the offices lighting and colors encourage
  communication, or detailed work?
• What do their clothes say about them? Is there a
  clear hierarchy based on clothes?

52
Observing Environment

• Using STROBE can help understand the priorities
  of a user
• If they dont use a PC often, maybe reports
  should be formatted for paper instead of Web
• If they dont care about looking at data, then
  the ability to drill down in reports might not
  be urgent
• If they are very hierarchical, that could imply a
  limit on the distribution of reports
• And so on