Critical bug on Twitter´s Report Panel?

ads
bugreporting

#1

Hello, we develop a product that persists report data coming from, among other sources, the Twitter Ads API.
We spent the last two weeks trying to understand what seems to be Daylight saving time (DST) bugs from the Twitter Ads API side, as described by my posts on that thread:

But now I am facing a worst problem, because the Twitter Ads Report panel is giving different results for the same days depending of chosen time range.
The links below are for the same campaigns, same everything except by the time range. The first is 10/25/2017 to 10/31/2017 and the second is 10/14/2017 to 11/02/2017

https://ads.twitter.com/ads_manager/18ce548xann/campaigns/?breakdown=daily&endDate=2017-10-31&filters={"account"%3A[{"key"%3A"fundingInstruments"%2C"operands"%3A["43230669"]}%2C{"key"%3A"objectives"%2C"operands"%3A[]}%2C{"key"%3A"displayStates"%2C"operands"%3A["8"%2C"9"%2C"5"%2C"6"%2C"1"%2C"7"]}%2C{"key"%3A"hadDelivered"%2C"operands"%3A[]}]}&startDate=2017-10-25

https://ads.twitter.com/ads_manager/18ce548xann/campaigns/?breakdown=daily&endDate=2017-11-02&filters={"account"%3A[{"key"%3A"fundingInstruments"%2C"operands"%3A["43230669"]}%2C{"key"%3A"objectives"%2C"operands"%3A[]}%2C{"key"%3A"displayStates"%2C"operands"%3A["8"%2C"9"%2C"5"%2C"6"%2C"1"%2C"7"]}%2C{"key"%3A"hadDelivered"%2C"operands"%3A[]}]}&startDate=2017-10-14

Take a look on the day of October 26 on the second campaign of both links. The results have a difference of 5% or more. The same happens with 27,28,29 and so on. This is a huge margin of difference for clients highly investing on the platform.

I really believe that this is still about the Daylight saving time (DST) bugs. Take a look on this other report link:
https://ads.twitter.com/ads_manager/18ce548xann/fundingsources/41299808/campaigns/14795929/ads/?breakdown=daily&endDate=2017-10-15&filterSet=1&filters={"account"%3A[{"key"%3A"fundingInstruments"%2C"operands"%3A["41299808"]}%2C{"key"%3A"objectives"%2C"operands"%3A[]}%2C{"key"%3A"displayStates"%2C"operands"%3A["8"%2C"9"%2C"5"%2C"6"%2C"1"%2C"7"]}%2C{"key"%3A"hadDelivered"%2C"operands"%3A[]}%2C{"key"%3A"campaignName"%2C"operator"%3A"1"%2C"operands"%3A["ios"]}]}&startDate=2017-10-15

One Tweet, segmented by day, on the October 15th. Total is different from the day result. :confused:
And 15th October is exactly the day when DST starts here in Brazil.

We need to explain to the client what exactly is going on, because this can cause serious divergences on spending.


#2

Thanks for the question, @controledash.

Those links you provided don’t work, but we’ve looked at a specific campaign: 8t4m1. On ads.twitter.com, when the date range is [2017-10-25, 2017-10-31], we see 44,830 impressions. This matches what we get using the Ads API:

$ twurl -H ads-api.twitter.com "/2/stats/accounts/18ce548xann?entity=CAMPAIGN&entity_ids=8t4m1&start_time=2017-10-26&end_time=2017-10-27&metric_groups=ENGAGEMENT&placement=ALL_ON_TWITTER&granularity=TOTAL"
{
  "data_type": "stats",
  "time_series_length": 1,
  "data": [
    {
      "id": "8t4m1",
      "id_data": [
        {
          "segment": null,
          "metrics": {
            "impressions": [
              44830
            ],
            "tweets_send": [
              1
            ],
            "qualified_impressions": null,
            "follows": null,
            "app_clicks": null,
            "retweets": null,
            "likes": [
              8
            ],
            "engagements": [
              389
            ],
            "clicks": [
              253
            ],
            "card_engagements": null,
            "poll_card_vote": null,
            "replies": [
              5
            ],
            "url_clicks": [
              3
            ],
            "carousel_swipes": null
          }
        }
      ]
    }
  ],
  "request": {
    "params": {
      "start_time": "2017-10-26T02:00:00Z",
      "segmentation_type": null,
      "entity_ids": [
        "8t4m1"
      ],
      "end_time": "2017-10-27T02:00:00Z",
      "country": null,
      "placement": "ALL_ON_TWITTER",
      "granularity": "TOTAL",
      "entity": "CAMPAIGN",
      "platform": null,
      "metric_groups": [
        "ENGAGEMENT"
      ]
    }
  }
}

However, when the specified date range is [2017-10-14, 2017-11-02]—on ads.twitter.com—we see 45,340 impressions.

For completeness, we used the asynchronous analytics endpoints to check the the impressions on the 26th, given the two date ranges specified. In both cases, we see 44,830 impressions.

Given that this is an ads.twitter.com issue, your best bet is to submit a request through that interface (the “Help?” link on the top right).


#3

Tks @juanshishido

Ok, I will do that, but I believe that the API also has DST related bugs.
Please take a look at my answers on this thread:

Best,
Pablo


#4

Thanks, @controledash. Will follow up on that thread.