Why Hiya use sampled data for analytics
All call data is not created equally. Some call analytics tools use limited data from carriers and applications, which prevents them from accurately calculating key metrics like answer rate or call duration. This can distort one's ability to measure the call outcomes one cares about.
To address this challenge, Hiya leverages a representative sample from data sources that can correctly identify calls answered by humans. This allows Hiya to provide analytics for answer rate, call duration, and other key metrics based on real conversations, not voicemails or automation. This sampling method also allows Hiya to build a foundation for providing additional analytics not available from other providers.
What is the Margin of Error?
All sampling involves a Margin of Error (MoE), a statistical measure that reflects the level of uncertainty—or potential error—of a sample of data compared to the full data set. The MoE helps us understand that the reported results are not an exact representation of reality but rather a range within which the true value is likely to fall.
Consider a scenario where you make 1,000 phone calls. The margin of error would be between 0.3-2.0%, depending on your answer rate. This margin is critical when analyzing trends or comparing different data sets. If the MoE for different data points overlaps, it could suggest that the analytics for these groups are essentially the same. You can learn more about the margin of error here.
Hiya uses MoE calculations to help you better understand your data. Research shows that Hiya’s samples are large enough that the typical MoE is negligible.
Using the Margin of Error in the console
You can view the Margin of Error for each data point by selecting the Margin of Error option and checking “Display margins of error for sample-based data”.
Increasing call volume, more accurate reporting
When the margin of error could be significant, Hiya automatically recommends changing the view to increase the call volume used for the data sample.
Hiya makes this recommendation so you can make the most informed decisions possible and know if the results might be inconclusive. The way you’d increase the volume of calls will depend on which report you’re interested in:
Trend Reports (Answer Rate, Unique Answer Rate)
For trend reports, call volume is calculated by the number of calls in each data point. For example, a weekly report should have enough call volume for each week in the report.
If you don’t have enough call volume for trend reporting, change the view to a higher level of aggregation (monthly instead of weekly) to view more trends with more data.
Histograms (Call Duration, Answer Rate Per Attempt)
For histograms, call volume is calculated for each bar in the chart for the overall timeframe of the report.
If you don’t have enough call volume for a histogram report, lengthen time frame of the report. This will include more data in the report.