Aggregate Traces and Sankey Analysis

If your familiar with funnel analysis from web/mobile product analytics such as in Mixpanel and Amplitude, you notice some similarities between funnel analysis and Moesif’s traces and sankey views. Think of these Moesif views as reverse funnel analysis of your API errors to trace across multiple isolated business transactions.

The last API call in the sequence is the error, while the others are API calls that lead up to the error. This is then aggregated so you can identify trends driving the error across multiple traces.

Traces Views

The below aggregate trace view is for a 500 error on POST /likes.

  • The red boxes is the actual 5xx error
  • The green boxes are API calls with 20x OK that lead up to the error.
  • The grey circle specifies the number of sampled traces that took the particular path

aggregate traces chart of triaged API errors

You can click on a row to bring up individual HTTP traces for a particular session or API token.

Sankey Views

Sankey is similar to the traces view. Sankey charts plot multiple traces on the same graph. The wider the path, the higher number of errors that took that path. The below sankey chart explores an error on GET /widgets/:id, and works backwards to see most taken paths.

Sankey chart of triaged API errors

You can change the depth of the Sankey analysis chart by using the slider below it.

If you click on a node in the Sankey Chart, the events for that Node will be displayed below the chart.

Filters

You can apply any of the filters to narrow down the events as well. In addition to standard fields such City or Region, we also supported auto generated fields based on your API data, such as any custom headers you use. Note, the filters apply to the Error node, i.e. the end of the Sankey Chart.

Filter of triaged API errors

Segmentation/Aggregations

You can generate charts and metrics by segmenting and aggregating on the node of a Sankey chart. See documentation on aggregations..

Updated: