Tuning And Sampling Improves Cost Savings
Depending on what’s being monitored, measured or reported on, capturing all data can become expensive. Trying to understand the results of an AB test, or trying to monitor Quality of Experience in a very granular way, or link datasets together for Observability, it might not be necessary to capture data from every session, and can still get the insights we need from data sampling.
All companies need to standardize and normalize data when receiving from multiple sources but building that capability, and providing ongoing maintenance, can be an unnecessary drain on resources.
Tunable Data Collection
As your data and analytics strategy evolves, your data collection may need to evolve along with it. It can be tempting to “collect everything” but ultimately if an event or metadata value is not being used actively for analysis, you can save costs through slimming data collection down.
Further, sometimes we might need to collect more data for certain occasions. When a new app or version is launched, when AB testing a new product feature, or when there’s a critical live event, more data than usual might need to be collected.
Sessionization: A Key Consideration For Sampling
Sampling can be a great way to reduce data ingress and egress, but sampling can be tricky business if not deployed correctly.
Let’s say that you want to only collect 10% of data. Simply collecting one event and skipping the next 9 would be problematic. It would result in having a little bit of data from many sessions, but not having 100% of data from any session.
Session-persistent sampling, or ensuring that 100% of data is collected around a single experience and enforcing sampling by session, requires special capabilities and signaling between the client-side and server-side.
How Datazoom Samples Data
Sampling on Ingest
Using Datazoom, we enable data to be sampled upon entering our platform. This means that we will only ingest a certain percentage of whole-session data. This can be helpful when you are looking at high-frequency data sets, like those collected for Quality of Experience Analytics or Observability. We offer a slide-bar for you to set your sample rate between 1 and 100%.
Sampling on Egress
Alternatively we can also enable data to be sampled upon leaving our platform. This means that we will only egress a certain percentage of whole-session data. This can be helpful to control costs at a Connector destination. We offer a slide-bar for you to set your sample rate between 1 and 100%.
How Datazoom Enables Data Tuning
Datapoints can be turned on and off, at any time, from within our UI. Custom metadata can be turned on and off also in the UI. The new configuration will activate from the moment the next app session is initiated.
Turning On/Off Data Pipes
Data pipes can be turned on and off, at any time, from within our UI. This does not impact data collection if other data pipes are activated. Collection will start or stop from the moment the next app session is initiated.