Notes about the test results:
Test A was used as the control. On this test the verbiage and copy that
would be used on all of the subsequent tests was established. Both the
headlines and the opt-in box were above the fold of the page giving the
greatest chance of opt-in.
Conclusion Test A:
The plain text format was a second to the last finisher in
performance. Possible reason for this is that this test and page design
did not clearly display the theme of the page and cause any excitement for
the offer. It lacked the ability to draw the visitor into reading the sales copy.
Test B was designed to test what would happen when an appropriate graphical
element was added to the offer. The graphic illustration was custom
designed to portray the essence of the sales offer. It clearly
demonstrated a person doing dog agility with their pet.
Conclusion Test B:
I believe this test outperformed the others for a couple of reasons. First,
it was the winner probably because a picture really is worth a thousand
words. The illustration helped clearly explain the offer. Since there were
no interactive elements it probably had the second fastest load time other
than the control test. Page load time can be a significant factor in
conversion.
In Test C and D we were adding an audio interactive element to our test of
sales copy and a graphical picture element. Test C was designed to give
the visitor the ability to decide if they wanted to hear the audio clip or
not. Test D automatically played the audio element once the page was
loaded. Over time we have seen that whether we allowed the visitor to start
the audio them self or we automatically played it for them, the conversion
came out the same in the end.
Conclusion C & D:
There may be a few reasons for these tests dropping in conversion as
opposed to the test with only the graphical picture element. First, the
conversion could have dropped possibly because of the slightly increased
time to load the interactive audio element. Later as we load even bigger
interactive elements the conversion drops even more significantly.
I am suspicious that part of the cause was because of the nature of how
audio delivers a message. In test B with just the picture, the whole
message is delivered instantly with a picture and some words. In the case
of audio a visitor must take in the story in sequence. They have to listen
to the whole message to understand the whole story and offer verses a quick
visual scan.
We would also have to consider that maybe the visitor picked up a prejudice
to the voice that was associated with the message. To rule out this factor
we would have to run the test more times with different voices and cadence
of speech and accents. In the graphic only test the visitor gets to hear
the voice of their imagination the way they most want to hear it. Once we
add reality we may have shattered their image of what they though Girl
Friday should sound like.
For Test E we now drop the graphical image and switch to a small video
image and video produced by Armand Morin's Video Generator. The test here
of course is to see if a real human can out convert a graphical picture or
audio. Once again the load time and delay goes up slightly as we add a
more complicated interactive element. We only tested one version where the
video clip was set to automatically start playing.
Conclusion Test E:
Once again as the load time goes up we see a diminishing conversion
rate. I cannot hold the load time totally responsible as it is not that
much longer than that of the audio. We are seeing though an increased
possible bias now against our spokes model as our actress may not be their
image of a Girl named Friday. Our story line might be losing credibility
at this point.
We have now also lost our clear illustration as to the theme of the
website. It would be an interesting test to add back in the graphic
illustration with the addition of the video to see how much of a factor the
loss of the graphic illustration has cost us here.
Test F and G we are now continuing down the path of looking at other types
of video. Worrying that a real person might have lost credibility with our
Girl Friday story we have switched to an animated character. In Test F we
are using our spokes model's voice and in Test G we are using a computer
generated voice. The animated video by far has our largest load time
before starting to play out its message and to actually display the first
glimpse of the animated character.
Conclusion Test F & G:
The load time is probably one of the more significant factors for the
significant drop in conversion on these two tests. The character I chose
may appear to be too young, too hip, too something and is now degrading the
reputation of our offer. We only saw a slight drop with the use of the
computer generated voice, but it is hard to tell how this factor may have
played out over a longer expanse of time. Once again we also did not have
a graphical representation of the service/product that was being offered.
Overall Conclusion:
As an overall conclusion and what I learned from these set of tests is the
importance of testing in the first place. These tests more than anything
just showed me how many more things there were to test before I can really
come to a concrete conclusion about the effectiveness of using audio and
video to increase opt-in conversion. If I had to pick one or two things
that I believe I learned would be that there is a big chance that load time
is a major factor on opt-in pages, and that the visitor must see a clear
graphical representation of what they are opting in for. Even these two
things I would have to test a little more to know whether they were true or
not.
Tools used for these tests.
- Textpad editor.
- Ask Database
- AWeber Autoresponder
- Ken Giddens personal Split-testing software
- AudioGenerator
- VideoGenerator
- Oddcast Video
 |
 |
 |
|
Did you find the information you were looking for?
|
|
|
|
|
|
|
|
|
|
|
 |
|
 |
|
|
|
|
|
|
|
|
|
|
|
|
