►
Description
In this session we set up an experiment for a redesign on the free trial page.
A
A
B
B
A
A
B
B
A
B
B
A
What
if
and
there's
obviously
trade-offs
but
like
what?
If,
instead
of
treating
us
as
like,
okay,
we're
replacing
the
old
feature,
so
we're
rewriting
we're
writing
on
top
of
these
files
and
is
that
how
I've
interpreted
this
diff
like
or
I
guess
this
is
a
yay
so
like
if
I
ran
this
on
the
GDK
locally,
this
would
be
replacing
the
sign
up
page
or
like
the
sign
up
trial
page
just
like
the
old
one's,
not
there
at
all
right.
Yes,
okay!
A
So
even
though
it's
not
great
but
like
I
see
us
when
we
do
like
feature,
Flags
will
sometimes
for
the
sake
of
like
just
having
One
path,
branching
entry,
point
of
a
duplicate,
the
file
like
so
we'll
have
the
old
version
that
we
don't
touch
and
then
we'll
have
like
the
new
version,
which
is
duplicated.
But
my
favorite
thing
to
do
is
then
like
add
comments
to
these
files,
pointing
to
each
other
saying,
like
hey,
there's
duplications
in
this
file.
A
B
Was
thinking
about
doing
that,
but
then
wrapping
it
like
just
doing
it
in
the
same
file
and
it
looks
like
it
can
wrap.
Oh,
this
is
this
is
wrong.
It
looks
like
I
can
wrap
this
and
I
think
a
class,
or
maybe
this
is
just
for
a
different
yeah.
It
looks
like
I
can
wrap
like
my
code
and,
like
maybe
the
experiment,
class.
A
A
A
B
B
B
A
That
could
that
could
help
out
too,
because
yeah
that's
gonna,
get
really
gnarly.
If
you
have
the
switch
here,
but
I
can
see
how
a
class
is
interesting,
it's
like,
oh,
maybe
you
could
just
have
the
class
encapsulate
the
two
diff
well.
The
class
then,
would
be
super
large
I,
yeah,
I,
I'm
I'm,
not
I'm,
still
not
entirely
sure
what
the
experiment
class
does
so
I
don't
know,
I
might
be
saying
things
that
are
so
off.
So
please
you're
the
expert
here.
Let
me
tell
me,
if
am.
A
A
Yeah
I
would
I
would
suggest
creating
yeah
I
would
suggest
creating
a
duplicate
file
and
then
like
here.
If
we
can
reset
it
back
to
the
original
content,
then
we
could
do
somewhat
like
a
feature
flag
check.
We
could
do
an
experiment,
check
okay
and
render
our
new
file.
If
the
experiment
is
under
the
candidate
variant,
I
don't
know
yeah.
B
A
A
A
B
A
So
here's
the
idea,
if
we
are
changing
the
CSS
such
that
it
is
okay
on
both
scenarios.
Then
it's
like.
We
don't
really
need
to
feature
flag
anything
here.
Do
you
know
what
I
mean
like
we
can
also
possibly
like
put
all
the
new
rules
or
something
under
some
other
class.
That's
like
we
can
have
at
the
top
left
the
top
of
our
sign
box.
Experiment
Hamill
have
like
a
class
selector
of
sign
box
experiment
and
then
everything
new
kind
of
just
goes
there.
A
A
A
Well,
I:
the
challenge
is
that
we
can't
necessarily
like
reference
experiments
and
stuff
in
here,
and
we
could
split
it
up
into
a
separate
bundle,
but
I'm
almost
wondering
if,
like
for
CSS,
what
do
you
think
of
if
we
just
take
all
of
our
changes
that
we
made
and
put
them
under
a
under
a
class
like
like
here,
there's
like
the
class?
Oh.
B
A
Right
so
I'm
saying
maybe
we
could
just
leave
it,
but
if
we
really
wanted
to
like,
what's
the
word
gate,
these
changes,
we
can
set
a
like
top
level
content
class
from
this
partial
of
the
Hamel.
We
could
do
something.
I
forgot
what
it's
called,
but
you
could
do
something
like
I
have
to
find
an
example,
but
you
do
something
like
content
class
sign
box
experiment
and
then
we
can
Nest
everything
under
signed
box
experiment
so
that
only
these
CSS
only
ever
applies
under
sign
box.
A
Experiment
partial
has
been
loaded
which
would
assume
assume
that
the
experiment
has
been
loaded.
That's
one
idea,
but
that's
only
if
we
only
really
need
to
do
that
if
things
have
are
affecting
each
other,
which
maybe
they're
not
like.
Maybe
the
CSS
Works
in
both
the
candidate
and
control
group.
So
then
you're
fine.
A
A
A
B
A
We
could
do
like
an
Elsa,
so
we
can
have
just
like.
There
are
three
conditions
there
that
sounds
cool,
but
I.
I
say
this
naively,
because
I
know
that
our
experiment
checks
aren't
like
our
feature:
checks
and
I'm
I'm.
Still
in
the
back
of
my
head,
I'm
like
what
is
what's
that
class,
the
experiment,
class
I
hadn't,
seen
that
before
I
gotta
go
find
any
I
gotta
find
the
experiment.
Example:
okay,
I
gotta
resolve
this
for
my
head,
so
I.
B
B
A
B
A
A
We'll
fix
that
later,
so
it
kind
of
looks
like
a
feature
flag
check.
So
there's
like
there's
the
name
for
the
experiment
and
then
you
pass
in
the
contacts
and
stuff.
But
then
we
pass
in
this
block
of
like
so
rather
than
doing
like
if
it
doesn't
just
return
a
Boolean
it
like
returns,
a
block
where
then
we
do
this
like
e
dot
control
and
then
there's
another
block
that
will
return
and
then
e
doc,
candidate
and
there's
another
block
that
gets
returned
so.
A
Yeah
and
so
I,
don't
I,
don't
think
that
there's
a
any
kind
of
so
for
this
example,
I
found
on
the
code
base.
I
am
pretty
sure
we
don't
have
any
like
any
there's,
not
a
class
backing
this.
It's
just
like
it's,
it's
more
akin
to
like
how
we
do
these
feature
flag
checks,
but
for
some
reason
experiments
are
kind
of
a
different
API
where
there's
a
we
use
these
blocks
instead
of
instead
of
just
doing
like
if
checks
but
I,
think
this
I
think
assuming
this
isn't
like
deprecated
way
of
doing
it.
A
B
A
B
A
A
B
A
A
A
A
Should
yes,
nice,
okay,
I,
think
I
think
that's
kind
of
how
it's
gonna
look
and
so
similar
to
the
other
page
that
we
had.
So
we
could
do
like
a
the
same
kind
of
thing,
but
we
can
render
our
new
partial,
where
we
moved
everything
over
and
maybe
possibly
add
comments
of,
like
hey
make
sure
you
check
things
across
files
and
so
there
we
would
do
something
similar.
A
A
So
if
it
doesn't
I
guess,
we'd
have
to
add
some
sort
of
container
class
to
the
sign
box
experiment
one
and
then
we
can
wrap
all
of
our
new
rules.
Inside
of
that,
but
yeah
yeah,
good
work,
I
know
it's
I
know
it's
I
know
that
it's
a
little
frustrating
when
you've
like
done
something
and
then
like
oh
I've,
gotta,
like
unwrap
all
of
it
to
like.
B
Yeah,
that's
what
I
was
hoping
I
wouldn't
have
to
do
when
I
I
just
got
back
from
vacation
yesterday
and
I
was
like
oh
okay.
I
want
to
get
into
this
I
was
like
great
I.
Don't
want
to
have
to
undo
all
of
my
wet
spaghetti
on
the
wall
and
there's
got
to
be
a
way
where
I
don't
see
my
wet
spaghetti.
There's
not
I.
A
Know
I
know
yeah
yeah,
but
no
good,
good
work.
It
seems,
like
it's
gonna,
be
hopefully
fairly
straightforward
to
do
the
do
the
experiment
so
yeah.
Let
me
know
if
you
learn
anything
about
the
experiment
classes
and
if
you
find
yourself
using
that,
because
I
haven't
ever
seen
that
I've
only
ever
seen
it
used
like
this.
So.