Twitter integration

[pab...@gmail.com]

Hi guys - I'm really impressed with what you're doing with the attention score. I assume you have already looked at integrating Twitter data into the score. I can understand why you would stick with FB data to start with, but you could potentially use Twitter as a proxy score to validate against the FB attention score. 

I would suggest something along the lines of:

1) pull article URLs tweeted from authorised publisher account & store URL and time of tweet

2) return to tweet URL after a set period of time (e.g. 30 / 60 mins, 4 / 8 hours) to measure increase in twitter engagement score. I would score engagement with tweets as:

number of re-tweets, likes, total number of comments. Probably in that priority order. 

4) calculate the velocity increase in engagements, i.e. how quickly did the tweet gain engagements over the period of time measured, how did this number change during the next period, and so forth

This gives you a twitter engagement score for each article URL, and a velocity engagement score for each tweeted URL (which should be an indicator of how much attention that URL gained over a set period of time). It may also be a way to get you access to Twitter data via the free API rather than having to immediately go to Gnip for the firehose (which would of course be great, but probably quite expensive). You can then go on to compare this Twitter score against other tweeted articles, FB scores and front page exposure within the overall attention score. 

Happy to discuss this further if it's of interest - I've spent a fair bit of time working with the twitter API on this type of scoring mechanism recently. 

All the best from your former colleague, 

Paul R.

[Ian Kennedy]

Perhaps this is implied but it's also helpful to measure a url's acceleration against a domain's average. Some publications have better and broader distribution than others so spike up more rapidly. I think twitter's api does this for you but you also need to strip away all the UTM parameters and roll up all the shortlink variables so that you are scoring a single, canonical url and not just one variant.

Ian

[Matt]

Hi Paul! Glad you're following along as we go here.

Twitter is out of scope at the moment, but your recipe here is a *lot* like what we use to collect URLs and track FB engagements over time. We vary how often we check the API based on how long ago we discovered it and how the velocity is changing.

When we're ready to do Twitter we'll definitely give you a shout.

[Matt]

Hi Ian - We get around the parameter issue by following links to their canonicals and then start tracking from there. The point about relative acceleration is a good one. Some publishers are never going to get into the hundreds of engagements per minute rate and yet knowing that they are having a spikey moment is crucial. 

At the moment we do track peak acceleration which is where scaling would be useful, but that figure isn't used in the calculation for the Attention Score. Might need to revisit that. 

Point taken. 

Thanks!