Using Excel for Qualitative Data Analysis

Ive been using an Excel spreadsheet to organize focus group and interview data for
analysis for several years. Its a nice compromise between the manual cut-and-paste
method and commercial software lie !tlas or "#ivo.
$o be clear% no system&!tlas% "#ivo% or Excel&can analyze the data for you% no matter
how expensive or sophisticated it is. It taes a human brain to do that. 'ut% unless youre
conducting a large% multiple investigator research study that produces a ton of data% Excel
does a fine (ob of organizing textual data for analysis. In this article I describe a step-by-
step process for doing that.
Assumptions
)irst% the following assumptions are made about any *ualitative study for which you plan
to use this methodology+
,ou used a rigorous data collection process.
$he *uestion guide was structured and intentional.
-ata collection resulted in high *uality data.
.omplete% accurate transcripts were produced.
Worksheet Template
'efore you start entering data into Excel% your spreadsheet must be formatted in a way
that eases data manipulation. /etting up the template properly will reduce frustrations
often associated with trying to manage word data with Excel. I start by creating a single
worsheet template for the first *uestion% which I then duplicate for each additional
interview *uestion. $his saves me from having to recreate the spreadsheet each time.
0y guidelines for developing the template are as follows+
1
1. /et margins to 2 and page orientation to landscape to maximize space for data
entry.
3. Enter the title of the study at the top of the page and leave a blan line for the
study *uestion 4remember% each *uestion has its own worsheet5.
6. .reate columns with headings for each of the following+
7I- 47articipant identification number5
.ode
8esponses 4as wide a column as possible without flowing over to the next sheet5
,ou can also create separate columns for face codes such as gender% age% occupation%
location% etc. !ttaching face codes to each response allows you to sort the data by those
parameters. )or example% you may want to contrast responses made by women versus
those made by men. 9r you might want to loo at how responses vary by age. 9f course%
this presupposes that you have collected face code data on each participant 4through a
short pre-interview survey for example5 and labeled it with the corresponding 7I-.
9nce youve created the template% mae a copy of it 4within the same worboo5 for each
*uestion in the study. -ont forget to insert one *uestion at the top of each worsheet and
label each tab at the bottom accordingly.
Data Entry
$he goal is to transfer all of the study data youve collected into the spreadsheets. /ince
my transcripts are in :ord% I convert the document to table format 4$able ; .onvert ;
.onvert $ext to $able5 before I transfer the data to the Excel spreadsheet. !ssuming there
is a carriage return after each response on the transcript% each response will transfer over
to a separate cell on the spreadsheet when you copy and paste. !lternatively% Ive also
used voice recognition software to spea the responses into cells.
!s I transfer responses to the Excel spreadsheet% I simultaneously add the uni*ue 7I- and
face codes that go with each entry. <enerally each response occupies one cell% but if the
response is multivalent 4has more than one meaning unit per response5 I split it into
separate cells and copy the 7I- and face codes for each split response.
'ecause I lie to see entire response entries on one screen% I use :rap $ext to eep
long entries from running across the next several sheets of the spreadsheet. :rapping
responses eeps them neat and tidy for printing too. :hen Ive completed the data entry I
lie to print out the data set and sit in a comfy spot with a nice cup of tea to begin
reviewing it.
3
Coding and Categorizing
I start the categorization by (otting down category labels that come to mind as I read and
re-read responses one *uestion at a time. $hese category labels can be a word or% more
often% a short phrase 4often verbatim from the transcript5. I use a large 4= > ?5 sticy
pad that I attach to the corresponding printed spreadsheets to denote the category labels
as they come to mind.
:hen I thin I have a fairly comprehensive list of mutually exclusive category titles% I
assign each a letter code 4!% '% .% -% etc.5 to mae the categories easier to wor with.
$his first pass through the data is only my first best guess of categories emerging from
the data. I then use a pencil to assign one of these codes to each entry on the spreadsheets.
9n this first swipe at the data I always find entries that+
-ont fit in any category
)it in more than one category
!re a ma(or subcategory of one of my categories
<enerate a whole new category
!re the only one of their ind
I mae ad(ustments in coding categories to accommodate my new insights@collapsing
categories with others that may only have one response% teasing out sub-categories or
smaller categories from larger ones. It still might not be perfect the second or even third
time through% but each time I move through the data% categories becomes less vague and a
more precise fit with the data.
9nce I am satisfied that I have found an appropriate code for each response% I enter the
codes into the .9-E column on my electronic spreadsheet. $hen I highlight the entire
data set 4mae sure you include !AA columns and rows5% clic /ort 4under -ata in
the top toolbar5% choose the .9-E column% and% voila% all of the data appears in
separate little groups according to the codes (ust assigned.
"ow I can loo at each group of responses and add% delete% change% subdivide% or collapse
categories if necessary. I sort again and repeat the process as many times as necessary
until I am satisfied that my categorization reflects the true essence of the data. I as
myself if this is what respondents were actually trying to collectively convey. -o their
voices come through or have I imposed my ownB
An Example
'elow is a simple analysis spreadsheet for a focus group study I conducted for the
Cniversity 9f .olorado /chool 9f -entistry regarding dental hygiene. I interviewed
flossers and non-flossers in separate groups. 7articipants were ased who influenced their
dental habits% about flossing follow through% and how they felt about themselves when
they flossed. $his example includes a partial list of responses to *uestion = 4:ho
6
influenced youB5. "otice the coding ey that corresponds to the letter codes in the .ode
column. In this example I have not yet sorted the data by code.
aking Comparisons
If you faithfully entered face sheet and identification codes with each response% then
Excels /ort function will enable you to mae comparisons among subsets of your data.
)or example% if you entered the gender of each participant% then you will be able to
compare responses of men to those of women for each *uestion if you sort all of your
data by the gender column. ,ou will be able to do this for each of the face sheet codes
entered. Aie a aleidoscope% the picture changes each time you re-sort.
'elow is a sample spreadsheet from a worplace inclusivity study I conducted for the city
of 7ortland several years ago. -uring analysis I sorted data by the eight separate face
codes 4role% location% wor group% etc.5 you see in the example. 48emember% you must
grab all columns and rows before hitting the /ort button5.
=
!tep"#y"!tep $uide
,ou can find a more detailed step-by-step guide on using Excel to organize and code data
for analysis by clicing here.
I developed these guidelines to help others learn how to use Excel to mae sense of the
piles of *ualitative data they generate in research pro(ects% evaluation studies% and open-
ended survey *uestions. I thin youll find it useful for your pro(ect too.
D