We are building a platform in which our clients are able to see their twitter account stats/metrics. The clients will be comparing the info available on ads.twitter.com for their accounts (https://ads.twitter.com/user/CLIENT_NAME/tweets) with our dashboard. These should match up very closely.
I want to compare impressions and engagement data on a per day basis for a given account. I have retrieved the time series engagement data (by hour) for the whole history of an account. Basically, it boiled down to sending requests such as
https://ads-api.twitter.com/1/stats/accounts/18ce54gzgbo?entity=ORGANIC_TWEET&entity_ids=8u94t&start_time=2017-05-19&end_time=2017-05-26&granularity=HOUR&placement=ALL_ON_TWITTER&metric_groups=ENGAGEMENT
for all the account tweets, changing start_time and end_time so that the tweet’s whole history is included (I have done it in batches of 20 tweets). Important note: I have retrieved data only for organic tweets. This way I got to a point in which the values that we have for Impressions, and those shown on the ads.twitter.com/user/CLIENT_NAME/tweets main graph, are the same/close.
The Problem
I get totally different results when I try to compare values for engagements that we got via the API with those on the Engagement Rate, Likes, Retweets, etc. graphs.
Note:
Perhaps not totally relevant, but good to know: when I compare impressions and engagements values for individual tweets, they are ok (same/close). The problem only appears when I try to sum them up at the account level.
Another approach
I’ve also tried retrieving account level data using requests such as:
twurl -H "https://ads-api.twitter.com" "/1/stats/accounts/18ce54gzgbo?entity_ids=18ce54gzgbo&entity=ACCOUNT&granularity=DAY&metric_groups=ENGAGEMENT&placement=ALL_ON_TWITTER&start_time=2017-10-01T00:00:00-07:00&end_time=2017-10-08T00:00:00-07:00" | jq
{
"data_type": "stats",
"time_series_length": 7,
"data": [
{
"id": "18ce54gzgbo",
"id_data": [
{
"segment": null,
"metrics": {
"impressions": [
0,
0,
0,
1949,
2957,
2986,
6965
],
"follows": [
0,
0,
0,
0,
5,
0,
1
],
"retweets": [
0,
0,
0,
0,
2,
0,
7
],
"likes": [
0,
0,
0,
0,
2,
4,
9
],
"engagements": [
0,
0,
0,
26,
46,
75,
345
],
"replies": [
0,
0,
0,
0,
0,
0,
1
]
}
}
]
}
],
"request": {
"params": {
"start_time": "2017-10-01T07:00:00Z",
"segmentation_type": null,
"entity_ids": null,
"end_time": "2017-10-08T07:00:00Z",
"country": null,
"placement": "ALL_ON_TWITTER",
"granularity": "DAY",
"entity": "ACCOUNT",
"platform": null,
"metric_groups": [
"ENGAGEMENT"
]
}
}
}
However, these number (all of them) differ widely from what I see on the ads.twitter.com page (look at ‘0,0,0,…’ impressions).
Changing ‘placement’ value to ‘PUBLISHER_NETWORK’ (or adding this up with ‘ALL_ON_TWITTER’) didn’t produce correct values.
Questions
I don’t know what/if I’m doing something wrong or whether my assumptions about the data retrieved are wrong.
I guess there are a few questions here:
- How are these metrics (available on the graph) calculated? What fields are included in calculating them?
- Is it possible to get the same engagement numbers by using data from the ‘organic_tweet’ request I have mentioned above? Or do I need to call a different endpoint to get that?
Help greatly appreciated.
