Behind the scenes at Labs
cancel
Showing results for 
Search instead for 
Did you mean: 

HPE Haven OnDemand: How to keep your business intelligence game on trend

Curt_Hopkins

renato-team.gifHila Nachlieli, George Kour, Alexander Maydanik, Ron Maurer, Yaniv Sabo, Olga Shain, Renato Keshet, Alina Maor

By Curt Hopkins, Managing Editor, Hewlett Packard Labs

This is the second part of a two-part series on the contributions Hewlett Packard Labs has made to HPE’s newly-launched developer platform, Haven OnDemand. The first part discussed anomaly tracking.

Labs researchers have a unique position in terms of problems. When they identify one, they have the knowledge and opportunity to go about solving it.

That was the case with one of the shortcomings of business intelligence. The status quo today is one of semi-manual manipulation of data. Because it is not fully automated, it exacts a heavy toll in terms of effort and time.

On a trip to Israel, Labs director Martin Fink issued a challenge to the analytics lab: figure out some way to make business intelligence bigger, faster, and easier.

The solution to this challenge, whose development was spearheaded by an analytics team led by project manager and senior scientist Renato Keshet, has become the second API that Labs has contributed to the new Haven OnDemand.

It is called Category Trend Analysis (CTA).

“Traditional business intelligence is about slicing and dicing,” Keshet told “Behind the Scenes.” The user has to choose what categories to aggregate data under and which players belong in which categories. The user has to deal with hundreds or even thousands of options.

“What we are giving the users here,” said Keshet, “is an engine that does all the possible cube dicing, analyzes, and then reduces everything into a single prioritized list of insights. CTA first identifies the categories that exist in your dataset, and identifies the players within those categories.”

After identifying the categories and its players, “CTA then aggregates data about your categories,” Keshet said. “For example, your input data might come from a database that contains raw data describing sales or transactions. Haven OnDemand can aggregate data about each category, for example the number of sales or the value of those sales. The API then runs trend analysis and returns the most significant categorical trends in your dataset.”

The Haven OnDemand documentation of TA offers this example.

You might have a dataset that describes sales for a French supermarket chain in 2014 and 2015. You can use Trend Analysis to find the most significant trends in the data. Maybe sales of cottage cheese in Lyon were significantly lower in 2015 compared to 2014. Maybe shoppers in Paris were responsible for a much greater proportion of all fruit sales in 2015 compared to 2014.

Trend Analysis first identifies the categories that exist in your dataset, and identifies the players within those categories. In the previous example, "cottage cheese" would be a player in the category "Lyon," but "Lyon" would also be a "player" in the category "cottage cheese." Trend Analysis then aggregates data about your categories. For example, your input data might come from a database that contains raw data describing sales or transactions. Haven OnDemand can aggregate data about each category, for example the number of sales or the value of those sales.

The API then runs trend analysis and returns the most significant categorical trends in your dataset.

CTA identifies and prioritizes market changes using several computations.

  • First, it computes the market share change itself.
  • Second, it multiplies the data by what is important to the user, including market size (dividing sheer size by market volatility) and market activity (the significance of a market change, based on how stable the market is).
  • Finally, the user can indicate which elements of the computation have a greater importance, or weight, to customize the results.

This second Labs API contribution to the Haven OnDemand platform is designed not just to make business intelligence easier for businesses. It is also designed to make that intelligence more useful. By accepting Fink’s challenge, Keshet’s team solved a problem that limited the efficacy of the decisions businesses could make.

cat-tool.gifScreenshot of an application employing the CAT API

0 Kudos
About the Author

Curt_Hopkins

Managing Editor, Hewlett Packard Labs