Adform API: totals mismatch when querying with dimension and without dimensions - facebook

Question

I'm working on PHP application that retrieves data from Adform API every day and saves them to Database.
There is a problem that data queried using dimensions, for example date and banner do not match totals, retrieved without dimensions.
For example, I make a POST to https://api.adform.com/v1/reportingstats/agency/reportdata
{
"metrics": ["ctr"],
"dimensions": ["date", "banner"],
"filter": {
"date":{
"from":"2016-08-01",
"to":"2016-08-30"
},
"campaign":{
"id":campaign_id
}
},
"paging":{
"page":1,
"pageSize":10000
}
}
Response is:
{
"reportData": {
"columnHeaders": [
"date",
"banner",
"ctr"
],
"columns": [
{
"key": "date"
},
{
"key": "banner"
},
{
"key": "ctr"
}
],
"rows": [
[
"2016-08-02T00:00:00",
"banner_1",
0.00816326530612245
],
[
"2016-08-03T00:00:00",
"banner_1",
0.0024213075060532689
],
[
"2016-08-03T00:00:00",
"banner_2",
0.001207653432082372
],
[
"2016-08-04T00:00:00",
"banner_1",
0.003472222222222222
],
[
"2016-08-04T00:00:00",
"banner_2",
0.000802886393241096
],
[
"2016-08-05T00:00:00",
"banner_1",
0
],
[
"2016-08-05T00:00:00",
"banner_2",
0.000676782107058102
],
[
"2016-08-06T00:00:00",
"banner_1",
0
],
[
"2016-08-06T00:00:00",
"banner_2",
0.000926995987708068
],
[
"2016-08-07T00:00:00",
"banner_1",
0.00904977375565611
],
[
"2016-08-07T00:00:00",
"banner_2",
0.0010050565441998231
],
[
"2016-08-08T00:00:00",
"banner_1",
0.0022935779816513758
],
[
"2016-08-08T00:00:00",
"banner_2",
0.000736000744868224
],
[
"2016-08-09T00:00:00",
"banner_1",
0.0052219321148825066
],
[
"2016-08-09T00:00:00",
"banner_2",
0.000636109173796044
],
[
"2016-08-10T00:00:00",
"banner_1",
0.0057971014492753624
],
[
"2016-08-10T00:00:00",
"banner_2",
0.000724849063441972
],
[
"2016-08-11T00:00:00",
"banner_1",
0
],
[
"2016-08-11T00:00:00",
"banner_2",
0.000581557986484298
],
[
"2016-08-12T00:00:00",
"banner_1",
0.0043103448275862068
],
[
"2016-08-12T00:00:00",
"banner_2",
0.000671168239505369
],
[
"2016-08-13T00:00:00",
"banner_1",
0
],
[
"2016-08-13T00:00:00",
"banner_2",
0.000803754549989838
],
[
"2016-08-14T00:00:00",
"banner_1",
0
],
[
"2016-08-14T00:00:00",
"banner_2",
0.000989294421086104
],
[
"2016-08-15T00:00:00",
"banner_1",
0.0064516129032258056
],
[
"2016-08-15T00:00:00",
"banner_2",
0.000638244734480941
],
[
"2016-08-16T00:00:00",
"banner_2",
0.000549180805298763
],
[
"2016-08-17T00:00:00",
"banner_2",
0.000551568224222697
],
[
"2016-08-18T00:00:00",
"banner_1",
0
],
[
"2016-08-18T00:00:00",
"banner_2",
0.000678091480705215
],
[
"2016-08-19T00:00:00",
"banner_2",
0
],
[
"2016-08-22T00:00:00",
"banner_2",
0.000360310246085577
],
[
"2016-08-23T00:00:00",
"banner_2",
0.000498299661680756
],
[
"2016-08-24T00:00:00",
"banner_2",
0.000561990345005873
],
[
"2016-08-25T00:00:00",
"banner_2",
0.000364882901395197
],
[
"2016-08-26T00:00:00",
"banner_2",
0.000372206184069575
],
[
"2016-08-27T00:00:00",
"banner_2",
0.000696784292763943
],
[
"2016-08-28T00:00:00",
"banner_2",
0.00084914217341799
],
[
"2016-08-29T00:00:00",
"banner_2",
0.000441720501352769
],
[
"2016-08-30T00:00:00",
"banner_2",
0.000556096204643403
]
]
}
}
So, total average CTR = 0.0015624820601283 = 0.15%
When I query without any dimensions:
{
"metrics": ["ctr"],
"filter": {
"date":{
"from":"2016-08-01",
"to":"2016-08-30"
},
"campaign":{
"id":campaign_id
}
},
"paging":{
"page":1,
"pageSize":10000
}
}
I get:
{
"reportData": {
"columnHeaders": [
"ctr"
],
"columns": [
{
"key": "ctr"
}
],
"rows": [
[
0.000729710843077063
]
]
}
}
CTR = 0.000729710843077063 = 0.07%
What is wrong?
Why such a difference: 0.15% vs 0.07%?
May be I should count CTR not by simple AVG but another way?
I also spotted the same issue with other APIs, such as Google Analytics API and Facebook Marketing API. Totals and averages counted locally using data sampled by dimensions are not always the same as totals provided by APIs themselves.