►
From YouTube: Clustering issues for Product Management Analysis
Description
I needed to find a solution for working through organizing a large number of issues. I found a no-code way to experiment with clustering.
A
Hi,
I'm
john
mcguire,
I'm
the
global
search
product
manager,
and
today
I
wanted
to
kind
of
discuss
and
kind
of
make
a
video
about
something
that
I
think
a
lot
of
product
managers
can
kind
of
run
into
less
than
a
year
ago
I
became
the
product
manager
for
global
search
and
global
searcher
get
lab
had
a
series
of
multitude
of
years,
probably
overall
about
four
years
where
issues
had
been
collected
from
users
from
people
that
had
worked
on
it.
A
It
was
kind
of
contributed
to
by
lots
of
different
groups
across
kit
lab
and
as
coming
in
as
a
product
manager.
I
need
to
start
figuring
out
how
I
want
to
analyze
those
things
and
kind
of
group
them
together
build
out.
What's
what
is
it?
Our
team
is
going
to
work
on
understanding
new
prioritization
group
together
some
epics
and
as
I'm
looking
through
it,
there
is
quite
a
hefty
number
of
issues
that
have
been
collected
over
those
four
years.
A
Try
to
understand
what
is
needed
from
those
and
figure
out
how
to
really
kind
of
organize
them,
then
reorganize
them
make
sure
they're
labeled
correctly
and
of
course,
gitlab
has
lots
of
tools
for
doing
batches,
but
you
need
to
create
groups
in
order
to
understand
what
those
batches
are,
and
that
is
the
problem.
How
do
I
create
these
groups?
How
do
I
group
together
this
kind
of
clump
of
issues
to
understand
what
is
a
common
path
with
them,
and
I
have
come
up
with
a
process
that
may
work.
I
wanted
to
share
it
out.
A
A
A
A
A
So
I
don't
really
need
those
items.
I
can
go
ahead
and
take
those
out
because
I've
already
reviewed
those
and
there's
you
know
this.
This
is
probably
the
list.
I
need
there's
431
items
and
and
889
overall.
So
that's
quite
an
extensive
list.
A
Now.
What
I
came
up
with
was
a
tool
that
does
text
analytics
and
analysis
and
it
does
some
kind
of
machine,
some
really
kind
of
lightweight
machine
learning
concepts.
It
can
do
some
things
in
more
advanced
too,
but
why?
I
chose
meeting
cloud.
A
A
Now
I
could
see
where
I
would
easily
be
able
to
go
over
twenty
thousand,
but
I
don't
think
I
would
do
that
every
month
and
I
think
I
could
probably
balance
that
out
from
a
month-to-month
cycle
and
that's
important
because
there,
the
the
plan
they
have
that's
not
free
the
least
expensive
plan
they
have.
The
software
is
a
hundred
dollars
a
month,
which
I'll
be
honest,
it's
probably
worth
it.
A
If
you're
going
to
need
to
do
more
than
20
000
api
queries
a
month,
I
I
think
you're
probably
doing
something
that
probably
is
going
to
be
worth
that
effort.
So
it's
probably
worth
to
support
them
with
that.
However,
in
this
case,
I'm
just
kind
of
experimenting:
I'm
not
ready
to
invest
100
a
month.
I
just
want
to
kind
of
see
you
know
if
this
is
going
to
work
for
what
I
need.
A
So
with
that.
I
can
just
export
this
to
a
csv
file,
which
is,
of
course
very
easy
to
open
up
into
google
sheets,
which
I
have
right
here,
and
the
next
piece
is
to
set
up
your
account
with
meaning
cloud
in
which
you
will
do
the
normal
type
of
account
setup.
You
will
be
given
an
api
key
and
then
you
would
need
to
configure
that
in
the
add-ons.
Oh
actually
you'll
want
to
install
the
add-on
for
meeting
cloud.
A
Then
you'll
need
to
configure
api
key
and
there
are
good
instructions
on
their
website
if
you're
not
familiar
how
to
do
that,
but
it's
actually
really
simple.
There's
only
the
api
key
that
it
needs.
Everything
else
is
already
linked
into
your
account.
I
want
to
mention
again
that
I
had
this
is
only
the
the
issues
that
are
public.
These
did
not
include
any
confidential
issues,
because
this
is
going
to
use
some
cloud
api,
which
I
don't
know
the
details
of
where
that
is.
A
I
can
classify
text
analyze
sentiment,
identify
language,
extract
topics,
cluster
top
text,
categorize
sex,
I'm
not
going
to
go
into
everything
that
it
does
and
there's
good
documentation.
That
explains
all
this,
but
it's
also
pretty
easy
and
inexpensive
to
actually
just
try
it
and
see
if
you
get
like
the
output
that
it
has,
which
I
definitely
would
encourage
trying
to
do
that
in
this
case,
I
want
to
do
cluster
text
now.
A
I
have
two
pieces
of
text
and
the
thing
that
I've
kind
of
found
is
that
the
way
the
descriptions
kind
of
show
up
with
the
the
markdown
not
rendered
does
tend
to
confuse
the
ml
a
little
bit
so
it'll
look
at
something
like
overview
and
summary
and
say:
oh
that's,
a
pretty
important
group,
because
I
see
that
happen
a
lot.
It's
like
it's
not
right.
It's
actually
should
be
ignored.
A
So
when
that
does
happen,
there
are
abilities
to
use,
stop
words
so
that
it
does
not
include
those
items,
and
that
is
a
very
important
feature,
but
you're
gonna
probably
need
to
do
a
few
iterations
of
that.
If
you're
using
the
description
field,
lots
of
good
information
in
the
description
field,
probably
a
better
cycle
to
do
some
of
the
other
types
of
analysis
that
they
have
for
like
deep
learning.
So
deep
learning
would
be
the
categorize
of
text.
A
That
is
a
whole
another
piece
that
requires
that
you
create
a
a
method
of
actually
understanding
what
that
text
is
about
in
the
context
that
you
mean,
which
has
several
more
steps,
but
the
one
that
is
easy
is
just
to
cluster
it
now
what
that
means
and
what
clustering
is
actually
going
to
do.
It's
just
going
to
say
I
looked
at
these
things.
These
seem
similar
enough
boom.
There
you
go
and
I'm
going
to
just
grab
a
word
from
that
apply
this
label
and
we're
going
to
call
that
group.
This.
A
That's
that's
alternative,
so
we'll
group,
these
things
together,
take
a
word,
apply
that
to
it
and
say
these
are
that
and
it's
going
to
also
output
some
degree
of
certainty
for
it.
So
all
those
things
are
very
nice
they're
very
helpful,
and
I
want
to
try
them
out
so
with
that.
I
have
put
in
a
few
things
that
I
I
don't
need
the
problem
to
solve,
because
that
was
in
the
description
field.
I
don't
need
summary,
but
I
do
not
need
get
lab.
I
do
not
need
elastic
search.
A
I
don't
need
elastic.
Those
are
going
to
happen
quite
a
bit
because
a
lot
of
things
we
do
have
these
words
es
that
stands
for
elasticstack.
That's
something
I
don't
really
need
that.
I
expect
will
come
up
quite
a
bit.
There
are
quite
a
few
that
are
just
support.
Type
related,
that'll
end
up
in
the
titles
I'm
going
to
block
those,
and
now
this
doesn't
block
the
whole
item.
A
It
just
blocks
that
word
from
being
analyzed
with
the
clustering
right,
because
it
just
it's
just
gonna
appear
too
often
it's
not
a
meaningful
word
for
the
thing
that
I'm
looking
for,
and
hopefully
that
works.
So
with
that.
I
have
set
this
to
the
active
range
which
I
have
selected
the
title.
I
have
said
that
it
starts
with
a
header
and
I
have
that's
it.
I
just
have
the
english
set
and
I
want
to
do
document
grouping.
A
I
also
the
option
of
topic
modeling.
I
say
experiment
with
both
those,
but
let's
go
ahead
and
kick
off
one
for
document
grouping
and
the
only
other
step
I
need
to
do
is
click
analyze,
and
here
it
goes
so
what
I
will
see
here
in
a
minute
or
two
just
a
few
seconds.
It
should
pop
up
with
a
new
sheet
there.
It
goes,
and
it
is
now
doing
the
text
clustering
and
we
can
watch
it
happen
real
time
as
it
clusters
these.
A
It
has
a
percentage
complete,
which
is
pretty
nifty
little
feature
as
it's
adding
these
in,
but
you
can
see
that
I
already
have
some
cluster
kind
of
labels
showing
up
here.
Project
file
commit
top
data,
show
here's
kit
lab,
which
I
I
think
I
need
to
actually
specify
the
capital
l
so
that
it
gets
omitted
correctly.
A
Es
still
got
picked
up,
which
I
don't
understand
why
that
might
not
be
intended,
but
it
could
also
just
be
that
the
label
applying
is
not
about
that
being
a
stopper,
it's
just
applying
that
label,
so
it
is
almost
done.
A
27
28
and
the
output,
which
I
already
have
is
gonna,
look
something
like
this,
and
so
I
have
now
gone
through
and
sorted
these
by
these
categories,
and
you
can
see
that,
like
admin
and
administration,
that's
probably
going
to
be
the
same.
I
would
want
to
go
through
a
mailing.
Your
manual
review.
These
still,
I
understand
they
all-
are
still
related
to
admin
and
maybe
add
some
more
detail
like
in
this
case.
Maybe
it's
related
to
administrative
console
and
that's
what
those
issues
are
about.
A
The
other
thing
that
I
can
do
with
this
is
a
new
feature.
That's
it's
fairly
new
with
google
sheets,
I
believe,
is
the
explorer,
so
explore
will
analyze
your
sheet
and
give
you
just
drag
them
on
and
give
you
some
quick
graphs
to
kind
of
look
at
and
see
what
your
data
looks
like.
A
This
is
incredible
because
I
didn't
really
do
anything.
I
just
clicked
the
button
that
said
analyze
after
setting
it
up
and
if
I
wanted
to
do
it
again,
the
only
thing
I
have
to
do
is
add
some
stoppers
and
analyze
it.
In
fact,
that
will
be
my
next
step
from
the
analysis
is
there
are
things
I
need
to
emit
so
index
had
several
things
that
were
counted
as
index.
This
count
is
the
number
of
items,
and
you
can
see
at
the
bottom.
You
have
a
list
of
some
of
those
like
git
lab.
A
So
honestly,
I
would
like
to
get
those
things
were
that
were
grouped
into
gitlab
to
be
re-analyzed
into
some
other
cluster,
and
so
the
way
to
do
that
is
to
then
add
these
to
your
stop
words
right
so
I'll
take
the
ones
that
I
think,
like
global,
doesn't
help
me
either.
So
I'm
going
to
re-analyze
global
I'm
going
to
put
that
in
the
stoppard
list.
I'm
going
to
I
might
keep
index
index
is
actually
a
very
valuable,
valuable
thing
for
search.
So
I
may
just
keep
index
but
I'll
probably
add
to
the
stop.
A
Word
list
global
and
get
lab,
and
then
I
will
run
this
process
again
and
since
those
are
the
two
kind
of
long
pulls
that
happen,
I
expect
that
those
issues
will
get
re-clustered
into
other
groupings
and
as
they
do,
I
should
be
able
to
kind
of
refine
this
down
to
at
least
the
number
of
things
that
are
topical,
that
I
can
then
apply
like
some
epics
to
that
group.
These
things
together
now
it
doesn't
give
me
the
structure.
I
need
to
go
and
do
prioritization
and
all
that.
A
But
at
least
I
have
a
way
to
look
at
what
is
our
massive
backlog
without
having
to
spend
a
tons
amount
of
time
to
read
through
hundreds
and
hundreds
and
hundreds
of
issues,
although
eventually
I
do
read
through
all
those,
but
this
will
give
me
more
of
a
structure
to
do
that.
Okay,
that
is
it.
I
hope
you
enjoyed
this
video
and
if
you
do
find
anything,
that's
useful
to
do
with
this
tool.
I'd
appreciate
the
feedback.
A
You
know
there's
three
different
kind
of
technologies
merging
here,
but
I
was
able
to
drag
and
drop
and
basically
create
some
histogram
analytics
of
a
massive
amount
of
information.
That's
in
the
index,
that's
exciting
to
me
as
a
product
manager
and
thank
you.