Correlation of GIF views and rankings in GIPHY’s search. A study with currently +100k lines of GIF ranking and view data
You ever wondered how important a ranking in GIPHY, Tenor and their search service in social and messenger apps is to get views for your GIFs? Here are some findings from scraping rankings and views on Giphy.
Giphy is showing some hints in the Dashboard, but in general the reporting is super basic.
Tenor is even worse. They show just shares:
My idea was to get an idea of what is possible in views if I rank 100 gifs or stickers on positions 1 to 20 in Giphy’s search. (Not saying that a ranking on position X causes Y views. There are other factors, too)
Study setup
First I needed useful keywords to test. GIF searches are different from normal search terms. No one is looking for “iPhone 11 case” in a GIF search engine. “tired” is much more popular.
Reasonable GIF search terms
I first scraped a list of GIF search terms, which are popular in Tenor. You can test any keyword if it’s a Top 10.000 Tenor search term here: https://tenor.com/insights/sun?range=30days This list is used in random order for searches in GIPHY’s search
What kind of data is collected
For every keyword, I try to collect Top 20 ranked GIFs and Top 20 stickers. Besides I collect the views, date published and hashtags of the single images
E.g. like this:
Study results
The easiest graph shows rankings and the average views of GIFs and Stickers per position without any filter.
As expected the first positions are much higher than e.g. position 20. But it’s not like in Google where page 2 / position 11 is almost zero clicks. It’s about 50% of the views compared to position 1.
I’m watching this now for a while with more and more data. And the numbers are not really changing. +40 million views on average on position 1.
The next two graphs are related to the publishing date of the GIFs. 2019 vs. 2016. My idea is that the two outliers on positions 9 and 11 are maybe caused by better rankings of those two in the past? What do you think? To me it makes sense that those older GIFs published in 2016 have about twice as many views compared to the newer ones from 2019
GIFs vs. Stickers
There are GIF and Sticker search results. In Giphy they are separated e.g. like this:
https://giphy.com/search/Zrce-stickers
Stickers are transparent background GIFs and they are newer as a service on Giphy. The date for GIFs looks like this
Stickers like that
But there is a problem with the availability of data here. Which is in general a problem with the data in this study. Many GIFs and Stickers don’t show view numbers public because the accounts switched off to show them. Besides even with popular search terms in GIFs there are often less than 20 stickers ranked.
The method of populating the results seems to be an exact match search for multi word keywords in Stickers, but an OR search for GIF search results.
Hashtags
In terms of Hashtags used per position there no so much difference.
I plan to do more with hashtag data. What do you want to know?
Additional thoughts
In all the graphs above views were counted on average per position, but if a GIF was ranking on positions 1 and 5 the views were counted for both positions. GIFs which rank for multiple Tenor Top 100 Keywords are probably the really good ones, but just for those of you who are interested in the first graph without those
Another option is to share the views of an image equally with all achieved rankings (So 100 views for position 1, 3, 5, 7 is 25 views for each position):
Results
To me personally, the biggest new learning from the data is that success (counted in views) in Giphy / GIF search engines isn’t connected that strong to the first positions. Position 1 has more views, but it does not drop as fast as in normal search engines (I wished I had gone for Top 50 rankings, but I don’t want Giphy to block me or wait forever).
Hack
Create nice stickers. A lot of high search volume keywords do not even have 20 stickers listed. So you can be instantly top 20
That’s it for now. But there may be part 2 and an update with more data.
I really looking forward to feedback, questions and problems with my reasoning…
These articles might be interesting for you too:
