By Cameron Afzal - July 13, 2021
At Sisu, we want to enable everyone to leverage their data to quickly understand what happened, why it happened, and how to take action. To help you operationalize your organization’s data and make the best possible decisions, we offer multiple metric types to give you the flexibility to analyze cloud-scale data.
If you’re an analyst helping marketing, finance, or customer success teams quickly understand what’s changing in their data and why, the core metrics you monitor are probably not as straightforward as those where each data point is treated equally. Instead, you need to take into account the varying degrees of importance of the numbers in your dataset using numerical rate and weighted average calculations.
Sisu calculates weighted average and numerical rate at cloud-scale, allowing you to quickly and comprehensively pinpoint the key drivers of your core metrics using all your data. To better understand how analysts calculate these metrics in Sisu, let’s look at weighted average and numerical rate in more detail.
In Sisu, using a weighted average metric allows you to analyze your metrics by weighting specific columns in your data table as more important than others. This metric type gives you the flexibility to quickly pinpoint the key factors impacting your key metrics while allowing you to incorporate context beyond the raw data into your analysis by using the metric’s calculation logic.
To see this in action, let’s look at an example. A growth marketing team at a fintech company uses weighted average to optimize their acquisition funnel. When this data team came to Sisu, they had tried to run analysis in traditional BI tools but found themselves spending hours analyzing changes in their data only to get the high-level results of their campaign. As they explained, “You only have so many hours in a day and trying to uncover what is contributing the most to why something is performing really well or doing poorly. It can sometimes be hard to get to it all.” They came to Sisu hoping to better understand how combinations of factors like ad copy, ad creative, and channel impacted complex metrics like weighted return on ad spend. The user noted, “Almost any marketing metric will be an aggregate metric, calculated with a numerator and denominator and often weighted.”
Analysts of the growth marketing team set up this analysis in Sisu by first identifying the metric column as `return_on_ad_spend` and then choosing the metric calculation type as `numerical rate`. To analyze their ROAS, they calculate their `Return On Ad Spend` metric as `return_on_ad_spend / ad_costs`. To confirm the analysis is set up how they need it, the analysts team can hover over the “i” button next to the calculation type menu to see how the metric is calculated given the columns they selected.
Once this metric is set up in Sisu, the growth marketing analyst or data team is able to continually iterate on their analysis to drill down into how factors like keywords, channels, and creative impact their target metric. The weighting will automatically adapt to the analysis type selected, whether it is general performance or a comparison analysis between time periods or groups, without recalculating or building a new project.
Being able to quickly uncover what’s driving their ROAS metric has allowed this data team to better optimize and improve their acquisition funnel. They commented saying, “With weighted average in Sisu, it’s easy to see what is positively or negatively influencing a campaign and dive deeper in seconds or minutes. With that extra time, I’m able to spend more time drilling down to find the impact of a creative or a channel. Then I can point our team in the right direction to make changes.”
When you run a numerical rate analysis in Sisu, you can represent one value as a share of another which results in a percentage. This type of analysis is a good fit whenever you’re calculating funnel-related metrics, like those in growth marketing or in finance, because it allows you to find the key drivers impacting your core metrics in proportion to the total.
To see this in action, let’s look at how analysts at a global financial institution use numerical rate in Sisu to identify the key drivers of fraud. Minimizing fraud is a critical goal for financial institutions, but the high volume of transactions result in massive datasets that analysts have to dig through to find the drivers. Often, the granular transaction data is aggregated to capture the relative importance of certain columns or samples.
Analysts of the financial institution wanted to use this aggregated data to identify factors that contribute to decreasing the rate of fraudulent transactions. To calculate fraud rate, they use the calculation `sum of the fraudulent transaction count` / ` sum of the total transaction count across all cohorts`.
Setting this metric up in Sisu, the analyst first identifies the metric column as `fraud_transaction_count` and then chooses the metric calculation type as `numerical rate`. Like numerical rate, analysts can review the set up of their analysis at any time by hovering over the “i” button next to the calculation type menu to see how the metric is calculated given the columns they selected.
After setting up the numerical rate calculation in Sisu, the financial analysts can choose to analyze the factors impacting their fraud rate as an overall rate for their whole dataset or comparison over time to determine what factors impact the change in rate over the duration of their dataset.
Interested in seeing how Sisu can help you quickly find the driving factors in your weighted average and numerical rate at cloud-scale? Get in touch with our team or view our knowledge base articles of numerical rate and weighted average to learn more.