navbar 4Resource papers in action research

Group feedback analysis  1


This is a resource file which supports the regular public program "areol" (action research and evaluation on line) offered twice a year beginning in mid-February and mid-July.  For details email Bob Dick  or

...  in which a small group technique, capable of being used participatively, is offered as an alternative to survey methods as a data collection process




Group feedback analysis 2 or GFA, is a small-group alternative to survey feedback.  3 Information is collected from and interpreted by a varied small group in a single session.  The small group may also devise the questions before answering them.

It is also an interactive procedure; you collect the data in a dialogue with informants.  It therefore offers you a more robust alternative to survey feedback, and in a form which can be used by people with relatively little experience.


When to use group feedback analysis

GFA is likely to be an appropriate technique where small-group information collection is called for.  It is suitable for use with numbers of people between about six and about twelve (more with skilled facilitation).

The best group is one which is a microcosm of the community.  It may be chosen in the same manner as a sample for convergent interviewing (described elsewhere 4), or it may be randomly selected as a reference group or jury.

Before you use GFA with a particular group, check that these preconditions have been met...

  • the group is sufficiently representative;
  • there is agreement on goals; and
  • group members understand what constraints apply.

Unless you are a skilled facilitator, GFA is not advised if there are serious unresolved conflicts within the group.  It may also be unproductive if the information you want is unrelated to the issues of most concern to group members.

The procedure is written for use by a group without consulting help.  A member of the group can fill the role of facilitator.  If you have an experienced facilitator you may wish to alter the procedure accordingly.


Facilitator preparation

In some settings the duties of the facilitator may be shared amongst different group members.  For the present description, however, I will assume you are the facilitator.

Your main function as facilitator is to guide the group through the exercise.  The purpose is to achieve an end result...

  • which is one that the group have reached by themselves,
  • and which reflects their real views.

For this to happen, each member of the group must feel able to contribute her views safely.  It is also important for group members to listen to one another (or a genuine group consensus cannot emerge), and to communicate cooperatively rather than competitively.

Here is a useful frame of mind for you to adopt: regard your task as helping the group to generate its own information while it takes responsibility for its own effectiveness in doing so.


The process

There now follows a description of the steps involved.  For many purposes you may want to precede the steps below with a definition of goals and constraints.  A shortened and modified version of search might be used for this purpose.  The constraints may be defined by someone speaking on behalf of the wider community, or from the body which will use the material you gather.

Here is the process.  I've used a decimal numbering system to identify the different layers -- you can get an overview by looking at the headings numbered 1, 2 and so on...


1  Develop questions

You can develop the questions yourself, as is usually done with research-oriented versions of GFA.  But if so, there is some preliminary work to be done, perhaps using convergent interviewing.  In the procedure described below, the group itself decides the questions.

During both this procedure and the later collection of responses, you use cards to allow group members to remain anonymous.  Standard 125mm x 75mm (5" x 3") system cards such as those used in office filing systems are small and convenient.

  1.1  Ask group members to develop, individually and without discussion, two or three questions which would secure the required information from the group.  Ask them to phrase these questions in such a way that they can be answered on a seven-point scale, ranging from highly favourable (7) to highly unfavourable (1).

You can suggest that in the first instance they begin questions with "To what extent ..." or "How much ..." or "How well ...".  This increases the likelihood that questions will be able to be answered on the scale.

  1.2  If the exercise is being done with a large group, this step may be done with small groups of three or so, so that the list of questions is not too large for easy collation.  Otherwise it can be done by individuals.  Each individual (or group) writes two or three questions on cards, one question to a card.  These cards are then passed face down to the front of the room.


2  Collate questions

You can collate the questions with the group watching.  It is usually better, however, if the group is responsible for collation.

  2.1  Ask the group to nominate two or three of their members to group the questions into a smaller number (between ten and twenty).  A representative sub-group is suitable, chosen from the quieter members of the group (they often have a more accurate perception of the situation).

  2.2  Collation can be done conveniently on a large table placed near the front of the room, with room for other group members to gather around it.

  2.3  The collators begin by looking for questions which would obtain the same information.  When such a pair of questions is found, it is placed on a separate cleared part of the table.  Other questions are then found which also ask for the same information, or information which is very closely related.

Questions are grouped only where they serve the same function, and not because they are about related topics or have similar wording.

  2.4  The collators then choose from each group of questions one which captures the theme, and which can be answered on the scale used.  If the best question is one which would attract a "yes/no" answer, it can often be converted by adding "To what extent" to the beginning of it.

  2.5  The questions are written up on a chalkboard or piece of butcher paper.  Space is left at the right for measures of average and spread.

  2.6  At the end of the questions are added a number of open-ended questions to be answered with a few words or a sentence.  These are "catch-all" questions to capture information not covered by the more specific questions.

The wording depends upon the type of information being sought.  As an example, suppose the GFA is being used to gather reactions to a proposal.  The following questions might then be useful...

  1. Write down the three major strengths of the proposal
  2. Write down the three major weaknesses of the proposal
  3. Write down the two or three specific changes which you think would do the most to increase the usefulness of the proposal.


3  Answer the questions

Group members answer questions individually.  Summaries are then written up and used as the basis for the later discussion.  (The summaries are not written up until all questions are answered; otherwise the responses to the earlier questions may influence the later responses.) In more detail...

  3.1  Read out the first (or next) question to the group.  Group members each take a 125mm x 75mm card.  In the top left hand corner they write the question number preceded by "Q" (otherwise someone will confuse left and right, and there will be problems).  In the top right hand corner they write their response to that question.  If reasons for the response are also to be collected, these can be written below.  The card then resembles the illustration.

  3.2  The cards are collected face-down by one of the collators, and given to a second collator who compiles the summaries.  (The collators also record their own responses on cards.  So does the facilitator if she is a member of the group.)

  3.3  The second collator sorts the card into rank order, with all the "l" responses first, then the "2" responses, and so on.  The cards are sorted into two equal piles; and each of those piles is itself further divided into two.

The middle card (or the average of the middle two cards) provides an average (technically, the "median").  The difference between the quarter and three-quarter marks gives a measure of the spread (technically, the "interquartile range") --

Original order

3    5    3    7    1    3    5    5    4    2    3    2    4

Arranged in rank order

1    2    2    3    3    3    3    4    4    5    5    5    6

Sorted into four equal stacks

1   2   2      3   3   3      3      4   4   5      5   5   6
         Q1=2.5        Mdn=3            Q3=5

The median is 3 (the midpoint) .  The interquartile range = 5 &endash; 2.5 = 2.5

The procedure of step 2 is then repeated for each of the scaled questions.  For the open-ended questions, the responses are recorded one to a card.  They can be collated on the same table used to prepare the questions, and a summary of the main themes recorded.


4  Prepare for the discussion

This and the following steps form the core of the procedure.  Group members use their previous responses as a means of knowing if there is close to a group consensus on some items.  They reach agreement on a summary statement of the views of their group.

In the interests of free and open discussion, it is better if this part of the procedure is conducted with the understanding that it is confidential.  The task of deciding finally what will be communicated to others outside the group is then left to become a separate exercise.

In more detail...

  4.1  The facilitator suggests that the rest of the discussion might be anonymous and confidential.

The facilitator reminds group members that they are using a collaborative process.  Their goal is to understand and communicate the view or views of the community as a whole.

During the discussion which follows any points made are written up on a second sheet of butcher paper.  This written record, like the discussion, is confidential.

  4.2  The items are worked through in an order determined by their average, from most negative to most positive.  For each, group members are given individual thinking time to prepare their interpretation of the results.  These are then compared.  Either a consensus is identified, or several views are recorded for reporting.


5  The report

All information so far generated and recorded is private information: confidential to the group.  The final session of the procedure is used to develop public information, or information which can go outside the group.  In more detail...

  5.1  Taking each agenda item in turn, group members decide if they want to communicate it to others.  If they do, they then agree on the wording with which it can be communicated.  This can often be done most easily by editing the summary developed when the item was first discussed.  When a final wording is agreed on it can be transferred to an appropriate action sheet.



  1. Modified from my 1990 resource document Processes for community consultation (Brisbane: Interchange and Department of Transport).back ]
  2. Group feedback analysis was first described by Frank Heller, who designed it as an action research technique.  Descriptions are available in a number of articles.  The original article was: Heller, F.  (1969) Group feedback analysis: a method of action research.  Psychological Bulletin, 72, 108-117.As Heller described it, pre-determined questionnaires were used, and the analysis was done by the researcher.  The approach described above is briefer, and involves the partipants in more aspects of the process.back ]
  3. Survey feedback uses a written survey as the catalyst for an organisational change program.  In the 70s it was probably the most widely used diagnostic device in organisation development.  Currently it is less frequently used.  For a typical description see French, W.  and Bell, C.H.  (1995) Organization development: behavioural science interventions for organisational improvement, fifth edition.  Englewood Cliffs, NJ: Prentice-Hall.back ]
  4. Convergent interviewing is described in the file "iview" on the archive.  [ back ]



Copyright (c) Bob Dick 1995-2000.  This document may be copied if it is not included in documents sold at a profit, and this and the following notice are included.

This document can be cited as follows:

Dick, B.  (1997-2005) Group feedback analysis [On line].  Available at




   navbar 4 

Maintained by
Bob Dick; this version 2.05w last revised 20050815

A text version is available at URL