|
 Click Here For Printable Version
Market Research Vendor File Format Specification
1. Purpose of this document
Define the acceptable manner for providing respondent level
data resulting from a survey research project.
2. Acceptable Media
- File transfer via FTP or e-mail, individual project files
(example: data, documentation) should be combined into
a single WinZip® file.
- DVD or CD-ROM.
- 3480 Tape Cartridge or 9-track tape
- Documentation should accompany the media indicating
the meaning or purpose of the enclosed files (example:
RAWDATA.SAV = respondent level SPSS® data file,
DATADOC.XLS = additional documentation).
3. Basic Acceptable File Formats
ASCII flat file, fixed field widths, one record per
respondent.
Maximum record length: 32,768 total characters.
Maximum fields: unlimited within the record length limitation.
Maximum records: unlimited.
SPSS®
Maximum field and record counts subject to SPSS file
structure limitations.
For ongoing (example: monthly tracking studies) projects, data
should be provided CUMULATIVELY to assure that respondent
record revisions, additions and deletions are captured.
4. REQUIRED File Documentation
All questionnaire or script versions associated with the study
must be provided, ideally in electronic form.
File layout and code table documentation described below will
ideally be provided in electronic form. MS-Excel® is typically
used for this purpose.
Specific to ASCII file format – FILE LAYOUT:
- File layout that clearly associates every question on all
versions of the survey with a field (example: column 120
width 2 associates with “Product Purchased”).
- File will contain a unique respondent identifier field that
may be used in subsequent updates or revisions to the data file.
- Any field that is included in the file that is not included in
the survey should contain a complete description. For example:
“Derived Lifestage variable”, “Screener question: respondent’s
next indicated purchase timing”, or “Sales Person Satisfaction Index”.
- Any data that may be provided separately in analysis materials or
presentation decks, for example, the results of cluster of factor
analysis of the data, should be included on the file and indicated
clearly on the file layout.
Specific to ASCII file format – CODE TABLES:
- For each FIELD in the study, separate documentation should be
provided with a clear label and code for each unique code provided
in the field. For example: 100 = Ford Escort, 110 = Ford Taurus,
120 = Ford Explorer, 999 = Unknown, etc.
- For rating (scalar) questions weights (example: Excellent = 100, Very
Good = 90 etc.) should be provided for derivation of means as required
to match separately provided presentation materials.
- For categorical questions that could be presented as a means or
medians(example: household income: under $20,000, $20,000 -
$24,999 etc.), range endpoints should be provided for derivation
of medians as required to match separately provided presentation
materials.
- Verbatim or open-end questions can be provided either within the
respondent record or by separate file. If provided within a separate
ASCII file, the record structure will contain one record per verbatim
and field structure will contain at the minimum three fields: 1) unique
respondent identifier 2) question (field) identifier 3) verbatim or
open-ended response.
Specific to SPSS file format:
- Each field will include a fully documented label (example:
Gender of Respondent as opposed to “REGEN”) that can clearly
associate with the separately provided questionnaire versions.
- File will contain a unique respondent identifier field that may be
used in subsequent updates or revisions to the data file.
- All questions on all questionnaire versions will be provided in the
file as separate fields.
- The coded contents of each field will be fully documented
(example: 1 = “Male”, as opposed to “M”)
- For rating (scalar) questions weights (example: Excellent = 100,
Very Good = 90 etc.) should be provided for derivation of means
as required to match separately provided presentation materials.
- For categorical questions that could be presented as a means
or medians(example: household income: under $20,000, $20,000
- $24,999 etc.), range endpoints should be provided for
derivation of medians as required to match separately provided
presentation materials.
- Any field that is included in the file that is not included
in the survey should contain a complete description. For
example: “Derived Lifestage variable”, “Screener question:
respondent’s next indicated purchase timing”, or “Sales
Person Satisfaction Index”. If required, provide a separate
document explaining the field content.
- Any data that may be provided separately in analysis materials
or presentation decks, for example, the results of cluster of
factor analysis of the data, should be included on the file and
clearly indicated by a separate document.
- Verbatim or open-end questions can provided either within the
SPSS file or by providing a separate ASCII file. If provided
within a separate ASCII file, the record structure will contain
one record per verbatim and field structure will contain at the
minimum three fields, 1) unique respondent identifier
2) question (field) identifier 3) verbatim or open-ended response.
General considerations that are applicable to both acceptable formats:
- Respondent weighting. If the study is weighted or has multiple
weightings, the weights will be provided at the respondent level
(i.e. as separate fields) and will be clearly documented.
- Rotated data: Records should be “derotated”, meaning that the
interpretation of any individual field is not dependent upon
the meaning of any other field. For example: If field version
= 1 then for field “vehicle 1”: 100 = Ford Escort, 110 = Ford
Taurus. If field version = 2 then for field “vehicle 1”, 100 =
Chevrolet Malibu, 110 = Chevrolet Impala).
- It is assumed that verbatim data will be coded. Actual
verbatim responses, in addition to the coded data, can be
provided as indicated above.
- Multiple response data can be handled in either “framed” or
“checkbox” format across fields. For example, question “Top
10 TV Shows and Magazines I Regularly View” with 1,000 possible
responses, but only 10 maximum responses, may be provided in
10 separate fields with an associated 1,000 item code table.
Alternatively, a question such as “Reasons I did not purchase”
with 25 different “check all that apply” responses can be
provided as 25 individual fields with a 0 indicating no check
and a 1 indicating a check (selection).
- Question Basing: Documentation on the basing of questions should be
provided if it is not clearly indicated on the separately provided
questionnaire document. Example: If respondents are presented
question “Items I purchased” only when answering “purchased” or
“leased” to separate question “Method of purchase”, then this needs
to be clearly indicated within the provided documentation.
5. Checking Materials
- At the minimum, an independent document indicating the total number
of respondents in the respondent data file, and an unweighted (and
if applicable) weighted frequency of at least one key question
(example: item purchased) should be provided.
- If possible, a weighted and unweighted frequency for each field should
be provided in electronic (example MS-Excel®) format.
If there are any questions required the specification outlined above, please address them to the attention of the following Productive Access staff members:
Doug Fritz Voice: 800-693-3111 x222 (US/CA) 1-714-693-3110 x222 (International) Fax: 714-693-8747
e-mail: dfritz@paiwhq.com
[ Return To Main Help Page ]
|
|
