WeRateDogs — A look inside its unique and delightful tweets

Analyzing this delightful twitter account in these challenging times.

We are going through a very difficult time. The year 2020 has brought us many surprises one after another and there no relief in sight. The news channels, newspapers are filled with all negative news going over each problem every single day. Moreover, almost every country is also going through is own set of unique problems. Don’t know what I am talking about? Look over the local news and you’ll know.

Now, all this takes a toll on the human mind. We feel depressed, sad, angry, anxious, and probably many other feelings when we take a look at the news or just go outside and look at what is happening in our neighborhood, our country, our world.

We all look for comfort in small things like reading a book, watching a movie or a series, talking to our loved ones, playing with our pets, or sometimes even looking at cat or dog videos and photos and boy-oh-boy, the internet is full of these. I came across this twitter account WeRateDogs and almost immediately I was smiling looking at their tweets. The account is full of incredibly cute dog pictures 😃.

WeRateDogs twitter account

The account gathered millions of followers over the years partly because of its unique rating system. If you look closely at the tweets, almost every single dog is rated like 13/10, 15/10, 14/10, and like so. I can understand why anyone would rate dogs like that… well because they all are very cute dogs! 😄

But I was still very intrigued. So, I decided to take a look into it and find out what makes this account and its tweets so special (dogs, of course, what else…?). I also looked into the pattern of tweets, the most popular dog breeds that they come across, and more.

I used data from the account’s archive which was provided by WeRateDogs to Udacity for educational purposes, scraping live tweets, and an image prediction file that would classify each dog’s breed.

Let’s dive right into it.

1. Based on the predictions, most liked dog breed: Which breed gets retweeted and favorited the most overall?

Most tweets that go through the account are of golden retriever and Labrador retriever. The graph shows that these two breeds dominate the others:

Fig 1. Count of classified breeds in the dataset

The golden retriever has the highest sum of retweets and favorites followed by Labrador retriever.

Fig 2. Sum of favorites and retweets per breed

The golden retriever has the highest sum of retweets and favorites followed by Labrador retriever.

However, when looking at ratings of a single tweet and its retweets and favorites, the labrador retriever comes out on top. What’s interesting is, golden retriever doesn’t even make the list. The soft-coated wheaten terrier got a remarkably high mean rating. The labrador retriever got overall the highest rating with 165, but since there are a lot more tweets with labrador retriever than for the soft-coated wheaten terrier, one of the ratings of the soft-coated wheaten terrier has a higher weight than the one of the labrador retriever (14 tweets of soft-coated wheaten terrier and 105 of labrador retriever).

Not taking the outlier into account, leads us to an average rating of 11. Based on the number of posts, retweets, favorites, and mean ratings, we will give the title of Most overall liked dog of this twitter account and its community to the labrador retriever.

2. How did the account develop (speaking about the number of tweets, retweets, favorites, image number, and length of the tweets)?

The account is highly active and sends out many tweets during the week. We can see that most of the tweets are posted on Monday. From Tuesday to Friday, it is the same number of posts. On the weekend the twitter profile tweets a little bit less. The count of tweets sent out each month does vary a bit.

Fig 3. Count of tweets for each day of the week

The following graph shows the number of tweets sent out each month for the year 2016

Fig 4. Count of tweets for each month in 2016

But when looking on month by month basis for each year, we get the following graph. The tweet has decreased quite a bit which may tell that the account performance has gone down, but if we look on the tweet favorites and retweet number for each tweet, the account has been performing well.

Fig 5. Count of tweets over the years

Fig 6. Sum of favorites per month per year

Fig 7. Sum of retweets per month per year

So well, that the mean favorites and retweets have been increasing on a linear scale. The sum of favorites and retweets are quite high.

Fig 8. Mean favorites per month per year

Fig 9. Mean retweets per month per year

We see that mean favorites and mean retweets have been increasing consistently. Which tells that the quality of the tweets has increased quite a bit. When looking at the images posted, the numbers are stable, and we see no upward or downward trend. Same for the upper text (tweet length).

Fig 10. Mean images posted per month per year

Fig 11. Mean tweet-length per month per year

3. Is there a pattern visible in the timing of the tweets?

Fig 12. Number of tweets posted for each hour in a day in 2015

Fig 13. Number of tweets posted for each hour in a day in 2016

As we can see in these graphs of 2015 and 2016, most posts are during the night between 0 and 500hrs. Between 400 and 1500hrs there are very small amounts of tweets. There is no single tweet for the hours 700–1200hrs and 1400hrs. There are some tweets after 1400hrs, but not as much as between 0 and 500hrs.

Conclusions

In this article, we took a look at the tweets of the WeRateDogs Twitter account. We take a deep dive to understand how the account developed and rose to popularity.

1. We first took a look at the most like dog breed which turned out to be a labrador retriever.
2. We then looked at how the account developed. We saw the number of tweets posted for each week, each month, and tweets posted going forward year over year. We also looked at the total number of tweets in a month in a year, mean favorites, mean retweets, mean images posted, and mean tweet length. I know this is a lot of analysis in a single area of interest but this actually shows how the account has developed over the years and rose to popularity.
3. In this one, we try to see if there was a pattern in the timing of the tweet and we do see a pattern. We see that most of the tweets went out during the evening and at midnight which tells that tweets were posted probably after regular working hours of people. Usually when people unwind after a hard day at work.

The findings here are observational, not the result of a formal study.

To see more about this analysis and how I got these result, you can look at the source code here