I've had an analytics reporting API running for a while now and unfiltered view results from the API match the web reporting. The issue I'm seeing is when adding a segment to the API report request. The web reporting is frequently returning different values than the API for a handful of the segment/view_id combinations. I'm looking for a recommended settings to review here to understand what is causing the discrepancy, as I'm not sure if this is an program code/API issue, web reporting issue or a configuration for segment/view_id issue.
Notes:
- When incorrect, it appears that the web reporting numbers for sessions is averaging 10% higher than what the API returns
- A single segment is applied to many view_ids we manage and a high percentage (~80%) are showing the discrepancy, the remainder match.
- the modified and created dates for this segment are 5 months old per the web interface, meaning there is not a configuration change within the segment causing the discrepancy
- we've compared 2018 YTD to eliminate a time lag data update as an issue.
- segments appear to be link to our master account level and applied to the accounts we manage.
- currently using v4 of the analytics API for .Net (C#)
Current Questions:
- Could this be a setting in how a particular segment was created?
- Why would some segment/view_ids match and others not?
- Is there a account, property or view_id permission/configuration setting to review as it relates to applying segments?
Any help or insights on what to review here would be helpful.
Forgot the code snippet:
var segmentDimension = new Dimension { Name = "ga:segment" };
var DefaultReportRequest = new ReportRequest
{
DateRanges = new List<DateRange> { dateRange },
Dimensions = new List<Dimension> { date, SourceMedium, Campaign, AdContent, Keyword },
Metrics = new List<Metric> { sessions, Users, NewUsers, Bounces, pageViews, SessionDuration, Goal01Completion, Goal02Completion, Goal03Completion, Goal04Completion },
ViewId = v_id,
PageSize = 10000
};
if (!(segmentId == ""))
{
DefaultReportRequest.Dimensions.Add(segmentDimension);
Google.Apis.AnalyticsReporting.v4.Data.Segment segment = new Google.Apis.AnalyticsReporting.v4.Data.Segment() { SegmentId = segmentId };
DefaultReportRequest.Segments = new List<Google.Apis.AnalyticsReporting.v4.Data.Segment> { segment };
};
var getReportsRequest5 = new GetReportsRequest
{
ReportRequests = new List<ReportRequest> { DefaultReportRequest }
};
var batchRequest5 = reportingService.Reports.BatchGet(getReportsRequest5);
var response5 = batchRequest5.Execute();
Thanks in advance for your help, Mike
Update 2:
After reviewing this further the API call is always pulling a single day of data "Yesterday". The web reporting when pulling that single specific day of data matches. If the web reporting pulls a time range of data around those specific dates (ex: +/- 3 days) the numbers no longer match. It seems like sampling could be in play here, but the web reports we are running indicate 100% of sessions in both pulls. I think the question is how to determine which is more accurate a single day or a time range of data. Has anyone investigate this, I've reproduced it on several of our view_ids.
Thanks, Mike
Update 3 (rseolution):
Turns out the issue was with how the segment was created and being applied to web reporting. The segment was focused at the User level, meaning aggregated values would change based on the time frame selected. The desired state was having the filters apply to a single day, making session focus a better then user as it contained the segment to the session.
Thanks all, Mike
