Performance recommandations

 

Pre-calculation helps accelerate the API calls processing time

API calls can be queried on-the-fly. However, when the data set gets more and more complex, it may take some time to get the expected results.
AT Internet has developed a smart calculation engine that allows the pre-calculation of certain API calls.

There are 2 ways to set up the pre-calculation of an API query:

  • Save a data export template in Data Query. By doing so, templates will be computed on a daily basis all templates that have been used at least once in the past 15 days, offering fast response times.
  • If you do not want to use Data Query, you can call the Reporting API directly. If you wish to schedule queries, we recommend executing at least 3 identical calls on 3 different days over the last 7-day period. In doing so, your call will be computed automatically every night, guaranteeing very short response times.

Please note that some dimensions and metrics prevent the usage of the pre-calculation.

 

What is the best time to run scheduled Reporting API calls?

Should you wish to automate queries whose period relates to the previous day (D-1), we recommend that you schedule the queries in the green time ranges indicated below, guaranteeing that your data will be available:

We recommend that you schedule your queries after 6am CET/CEST to take advantage of the shorter response time. If you wish to have even better response times (using computed data), we recommend using the API after 8am CET/CEST.

The performance indicators shown in the above table are based on average data availability and response times. During busy periods (beginning of the week, beginning of the month) the indicated time slots may fluctuate.

 

Filters

For performance reasons, the use of certain filters is recommended over others:

Filter typePerformance level
Is equal to (“$eq”)****
Starts with (“$start”)***
Ends with (“$end”)***
Contains (“$lk”)**

Performance may also vary depending on the number of items contained in the application.

As far as filters on metrics are concerned, the performances are similar irrespective of the type of filter used.

Last update: 04/03/2019