# Filters

The shorter the interval, the more granular the aggregation results will be. Longer intervals create a broader perspective and better identification of overall trends. If you choose an interval too short for what you want to measure, you may wind up with noisy or lacking data. If you choose an interval that is too long, you may smooth out important variations.

For example, if your desired aggregation is a count of daily active users, but you collect data every hour, your aggregation interval would be daily. If you selected an interval of minutely, you would lack results in data for several intervals.

Consider the robustness of your data producers. In a perfect world, your producers may emit data once a minute, but in the real world, this may not be practical. For example, client-side events are imperfect, and electrical sensors may suffer from interference. In these situations, we may want to use a slightly larger interval, such as every two minutes, to mitigate that variability.

Higher-level aggregate forecasts are typically easier to create and more accurate than more granular forecasts. So for forecasts, use the longest interval that still satisfies your business case.

For example in retail supply chains, forecasts often focus on weekly sales because supply trucks deliver once a week. In this case, we don't care about sales on any particular day, so we choose weekly. Choosing daily and aggregating those intervals to weekly after would mean biases from the dailies would get compounded.

To detect irregularities, consider an interval that adheres to your data volatility and fluctuations. For example, if you want to analyze daily patterns in website traffic, your aggregation interval would be daily. Choosing an interval that is too long may not accurately capture those changes because it smooths out important variations.

Consider how you want the aggregation results to be presented in reports or visualizations, and who they are for.

Visualization: If you're interested in seeing monthly sales performance in the form of a line graph, daily metrics would be more readable than minutely.

Reporting: Align your interval with your reporting frequency. Teams may want daily reports for operational monitoring purposes and monthly or quarterly reports for strategic planning. In these scenarios, you would set your interval to daily for ops-related aggregations and monthly or quarterly for strategic aggregations.

In real-time monitoring situations, shorter aggregation levels are more typical as they allow quicker detection. For example, if you desire to monitor server performance for rapid response to issues, your desired metrics may require shorter intervals such as 30s or minutely. For quality control, where compliance standards are critical, shorter intervals allow quicker detection and correction of any deviations from quality standards. To analyze the effectiveness of a marketing campaign, especially digital marketing, shorter intervals allow marketers to make timely adjustments based on real-time performance data. For businesses that require a real-time understanding of customer behavior, such as e-commerce or online platforms, shorter intervals provide better insights into user interactions and preferences.