To gain crucial insights, use BigQuery ML's contribution analysis feature.
As data volumes increase, organizations find it more difficult to understand changes in the data. Making decisions is hampered by organizations' inability to pinpoint the cause of significant trends and swings. "What factors drove revenue growth between Q1 and Q2?" or "Why did an advertisement click-through rate decrease by 5% over the last week?" are some questions that a firm might have.In order to identify statistically significant critical drivers, one must use tools to evaluate data segments at a time. Google Cloud is launching BigQuery ML's public trial of contribution analysis to assist businesses in interactively and massively finding patterns and insights in their data.
Analysis of contributions
Contribution analysis, also known as key driver analysis, can assist you in understanding the changes that have occurred in significant indicators within your multi-dimensional data. Contribution analysis can be used, for example, to compare two training data sets in order to see changes in sales figures over a two-quarter period or to comprehend changes in an ML model's performance. Use the setup MODEL statement in BigQuery to create a contribution analysis model.
Contribution analysis is a part of augmented analytics, which is the use of artificial intelligence (AI) to enhance and automate data analysis and interpretation. One of the primary goals of augmented analytics is to identify trends in user data, and it facilitates the achievement of this objective.
A contribution analysis model locates data points that show statistically significant changes in a metric over time by comparing a test set of data with a control set of data. This enables you to monitor changes in data according to variables such as location, time, and client segment. A table snapshot from 2023 and 2022 can be compared to observe the changes in data over a two-year period.
A metric, or numerical value employed in contribution analysis models, is used to measure and compare the difference between the test and control data. You can offer a summable metric or a summable ratio metric using contribution analysis models.
A segment is a subset of the data that has a certain set of dimension values that set it apart. For example, in a contribution analysis model, any conceivable combination of the store_number, customer_id, and day dimensions makes up a segment.
Contribution analysis allows you to look at pertinent metrics from your dataset across designated test and control subgroups. It works by figuring out which combinations of "contributors" produce unanticipated changes. It also expands efficiently by reducing the size of the search space through pruning optimization. An extensive range of industries and applications can profit from this type of analysis. Among the examples are:
Telemetry monitoring: Look for differences in events that the software has recorded.
Sales and advertising: Look into user activity to modify campaigns and advertisements based on click-through rates.
Retail: Evaluate the implications of price changes and inventory management strategies to optimize stock levels.
Healthcare: Examine significant factors that impact patients' health in order to enhance treatment strategies and prognoses.
BigQuery ML's contribution analysis model
How does it function?
To develop a contribution analysis model in BigQuery ML, all you need is a single table containing rows of a control set of baseline data and a test set to compare against the control, a metric to analyze (like revenue), and a list of contributors (like product_sku, category, etc.). Next, the model identifies meaningful data slices, or what we refer to as segments, based on a specific mix of contributor values.You can look at summable metrics and summable ratio metrics, which are two different types of metrics, with contribution analysis. Summable metrics condense each data segment by aggregating a single measure of interest, like revenue. Summable ratios are metrics that look at the relationship between two significant metrics, such as earnings per share.
Furthermore, contribution analysis models come with pruning optimizations by default, which lets you apply the Apriori pruning method to get insights faster. Based on a minimum support value, the model can cut down the search field and rapidly identify significant portions. The support number indicates the size of a segment relative to the entire population. By eliminating segments with low support values, you can reduce the query execution time and concentrate on the larger segments.
Contribution Analysis is now available in preview on BigQuery.
0 Comments