New Insight: Most faved/liked post
adampash opened this issue · comments
One-liner
Personal best: Most faved/liked post.
Full explainer
This insight keeps track of the user's most faved tweet/liked status update and gives the user a heads up when she's broken her previous record. Should make the user feel good, like a personal best should.
Audience for the insight
This insight works best for Twitter and Facebook.
How often this insight runs
- One-time only, if it's never appeared before
- First crawl only
- Daily
- Weekly
- Monthly
- Yearly
- Triggered by a data event
This insight should run as soon as we've scraped all of the user's initial data to find the baseline. Then run daily to check if the user has surpassed the baseline.
Headline
- New Personal Best: Most %like_type-d %post_type!
- %usernames has a new most-like_type-d %post_type!
Body
- Liked it? Looks like they LOVED it!
- Whatever %username is doing, it's working.
Then:
- With %total %like_type-s, %username's new most-%like_type-d %post-type is:
E.g.:
New Personal Best: Most faved tweet!
Whatever @adampash is doing, it's working. With 82 faves, @adampash's new most faved tweet is:
If this is the first run for this insight, need copy to explain the limitations of the accuracy of this first run. E.g.,
- In its early days, Twitter didn't keep close records of faves and retweets, so while the tweet above is the most-faved according to the data ThinkUp has, we can't be 100% certain that it's your best ever. But here’s the good news: Now that you’re using ThinkUp, your data is archived daily, meaning that when you hit big milestones from this point forward, we’ll get it right, every time.
Maybe could live in a footnote?
Criteria and logic
The baseline for this insight requires all of the user's existing posts. Then we need to find the post with the most faves/likes.
After the baseline is set, the insight needs to pull the user's posts from the previous week (or whatever API calls will allow) to check for anything that surpassed the previous personal best. It's probably also worth continuously updating the fave/like count of the original personal-best on each run, in case it's changing.
Actually, to avoid overusing API calls, we could also possibly just check the 5 most faved/liked posts in the previous 30 days or something like that and continue updating their faves/likes in the ThinkUp DB.
The minimum threshold for posting this insight is, of course, > baseline.
Included elements
- Headline
- Text
- Header image (image off to the left in side-by-side style insights)
- Hero image (giant image on top)
- List of user(s)
- List of post(s)
- List of link(s)
- Action button
- Line chart
- Bar chart
- Other viz
- Other graphic treatment
Hero image for Twitter:
http://www.flickr.com/photos/meddygarnet/3423734717
http://www.flickr.com/photos/franzifisch/2090951156
So, currently we have a 365-day high favorite count insight. (It's inside the activityspike insight.)
One thing about that, in relation to this, is that we specifically changed it so it would only twll you about a 365 day high, if we had 365 days of crawling data. Because of twitters limits to looking at back-tweets, I believe that we were getting enough false positives that we disabled it, rather than give people bad data. Gina may have more insight there. (cc @ginatrapani )
My perspective is that it's better to craft a way to appropriately address this issue than to say we can't do all-time personal bests. 365-day best is still a good insight, but it's a different insight. Best ever is a really powerful thing.
That said, we do need to figure out what the well-crafted path to addressing the issue looks like. I'll work on that and update when I've got something.