3 ms·

> the analytics tools (pretty but pointless) Hah, yeah. I was curious about how these worked a while back and figured out that a tweet's "Impressions" includes

> the analytics tools (pretty but pointless) Hah, yeah. I was curious about how these worked a while back and figured out that a tweet's "Impressions" includes any time someone reloads a timeline that includes your tweet somewhere within. This includes your own page loads, so if you open Twitter a few times throughout the day you'll see your tweets accumulate impressions by default. And if you're followed by a really active Twitter user, those numbers will be inflated quite a bit. I'm not sure whether this is shortsightedness on Twitter's part or "convenient" padding of their numbers.

ktRolster 11y ago
To be fair, that's more or less what an impression is. If you want "unique visits," you can look for unique visitors or unique impressions or just uniques, depending on what company you are working for. (And if you are doing TV, you can get an estimate of how many people saw it on DVR in the next three days)
- mintplant 11y ago
  True, however Twitter doesn't expose uniques at all, only impressions and "engagements", which also include your own activity. Hence the analytics tools they offer are "pretty but pointless".
blakeyrat 11y ago
> I'm not sure whether this is shortsightedness on Twitter's part or "convenient" padding of their numbers. That's exactly the definition of the word "impression" for ad-serving systems.
- mintplant 11y ago
  Sure, but - Twitter highlights impressions when it's a particularly useless metric for their platform. Why not expose uniques? - Why include a user's own activity in their analytics data? Does anyone really want that? - A tweet doesn't even have to appear on the screen to acquire an "impression". It could be far down the timeline and out of view, but as long as it's included in a batch in an API response, the counter goes up. - This number is exposed to everyone, not just ad industry insiders, and is subtitled "number of times people saw this tweet" - but that's not what's counted. The implementation may be technically correct, but it is, I think, misleadingly biased toward higher numbers. I can post a tweet and it can get 10 impressions in less than a minute, just from my clicking around. If I expand the tweet and click "reply" to attach another 140 characters, that's two "engagements". If I click a link to make sure it works, that's three. Honest values for those numbers would probably be zero.

ktRolster 11y ago

To be fair, that's more or less what an impression is. If you want "unique visits," you can look for unique visitors or unique impressions or just uniques, depending on what company you are working for. (And if you are doing TV, you can get an estimate of how many people saw it on DVR in the next three days)

mintplant 11y ago
True, however Twitter doesn't expose uniques at all, only impressions and "engagements", which also include your own activity. Hence the analytics tools they offer are "pretty but pointless".

blakeyrat 11y ago

> I'm not sure whether this is shortsightedness on Twitter's part or "convenient" padding of their numbers. That's exactly the definition of the word "impression" for ad-serving systems.

mintplant 11y ago
Sure, but - Twitter highlights impressions when it's a particularly useless metric for their platform. Why not expose uniques? - Why include a user's own activity in their analytics data? Does anyone really want that? - A tweet doesn't even have to appear on the screen to acquire an "impression". It could be far down the timeline and out of view, but as long as it's included in a batch in an API response, the counter goes up. - This number is exposed to everyone, not just ad industry insiders, and is subtitled "number of times people saw this tweet" - but that's not what's counted. The implementation may be technically correct, but it is, I think, misleadingly biased toward higher numbers. I can post a tweet and it can get 10 impressions in less than a minute, just from my clicking around. If I expand the tweet and click "reply" to attach another 140 characters, that's two "engagements". If I click a link to make sure it works, that's three. Honest values for those numbers would probably be zero.