Step right up and get your latest API performance highlights! And you’ll never believe – the U.S. Government’s APIs are still broken!
Up first we have a company we don’t normally see in our API performance roundup – Bitfinex. They’re in our in the Exchanges collection and last week, they dropped 7 places. This was due to a large increase in outliers and median latency affecting their Get Tickers History endpoint. No other endpoints were affected; however this endpoint saw its median processing time jump by 1342ms to 2701ms.
Next up is a first-timer. Lob in the GGV Capital API-First collection dropped 5 places. This was due to four clusters of outliers affecting a couple of their endpoints this week. On the Get Postcards endpoint, 9.4% of calls were classed as outliers. Whoops!
And now, the ongoing saga of U.S. Government APIs.
Bureau of Labor Statistics in the US Government collection dropped 13 places. This was due to a very high number of failures, leaving an effective pass rate of only 81.5%. These failures were a mix of errors occurring in the SSL handshake after the agent received a 302 redirect (middle and bottom bands), connection timeouts (top band). These API performance failures are simply all too common.
Dept of Energy in the US Government collection was in the Red Zone with a CASC score of 3.56. This is just a continuation of the continuous poor performance we have seen from the Department of Energy over the past month and a half. 80.0% of calls were classed as outliers this week, giving an effective pass rate of only 20%. This incident began on the 24th of March and still has not improved.
An Interesting Error
Nationwide Building Society in the Open Banking UK Production collection was in the Red Zone with a CASC score of 5.6. This was due to a large increase in processing time on the 11th of the month. Nationwide acknowledged this in OBIE-11340 and it appears that the issue was resolved.
Note, though, that there were rumblings for a couple of days before the spike. If they’d been constantly monitoring, they would have caught this before it became an issue. Guess they should be using APImetrics!