Modeling and analytics provider Luminar is counting on big data technology to help it deliver valuable insights about U.S. Latino consumers. Last year, the company ditched a traditional data warehouse setup in favor of the Hortonworks Data Platform -- a Hadoop distribution -- to help it do the job faster.
The idea, said Luminar President Franklin Rios, is to provide customers with a far more reliable alternative to focus groups and surveys -- and to provide better information about Latino communities than ever before. For example, Luminar can tell its customers how Cubans in Miami spend their money on technology, or how much the typical Puerto Rican male in New York spends on food.
SearchBusinessAnalytics got on the phone with Rios recently to learn more about how his company is using big data technology. Rios talked about how Luminar deals with the complexities of delivering valuable business insights about a specific portion of the population. He also explained how Hortonworks fits into Luminar’s strategy, and how data flows through his organization. He even had some sage advice for anyone considering a big data initiative. Here are some excerpts from that conversation:
What is Luminar and what does it do?
Franklin Rios: In short, Luminar is an analytics and modeling company that specializes in the U.S. Latino consumer. What I mean by that is we drive insights through analytics and modeling by driving into [data] that we analyze and that we ingest from multiple sources. We give consumer packaged goods (CPGs) companies, retailers or what have you, insights into the true behavior of the Latino consumer.
If anybody is starting to look into this technology, into migrating into this technology, don't try to do it alone. You've got to engage with someone to help you through the process.
Why did Luminar see a need for a Latino-focused market analytics provider?
Rios: Before Luminar, a lot of marketers or advertising agents or whoever wanted to start reaching out to the Latino consumer and they relied on highly sampled data from the usual suspects. Companies would use focus groups and self-reported panels and [surveys of under 10,000 Latinos]. They would extrapolate [from surveys and focus groups] and do statistical numbers to suggest how the rest of the 52 million Latinos in the U.S. behave. There is $1.5 trillion worth of purchasing power in the U.S. Hispanic space. With 52 million Latino men, women and children in the U.S., I don't know how a sample of 10,000 self-reported Latinos can give any true indication to any retailer or CPG on their behavior.
What is Luminar doing that is different from that "legacy" approach?
Rios: The traditional way is highly sampled. So we're saying that we are not going to do that. We're going to take transactional data that we're going to license from multiple sources. We have about 2,000 sources of data that we ingest and that we analyze and we clean up, and then we apply what we call cultural filters in order to truly find out who is a Latino.
Some of the data you receive must be ambiguous. How do you tackle the problem of identifying who is who?
Rios: We have access to data that comes from loyalty systems. So, if you belong to a grocery store loyalty system, or if you belong to any kind of [loyalty system], your name and address and all of that is in there. But how we truly start deriving who is a Latino is by starting to look at the purchases and the transactions that the household has made. We also have access to things like magazine subscriptions, and we know if somebody is getting their utility bill in Spanish and all kinds of stuff. We use a scoring mechanism that says if you're doing 55 or 100 of these behaviors, and if in the grocery store the contents of your basket contain products that are very much Latino products for cooking and such, then the scoring keeps going up and up and up to identify a Latino. That is the first step. Then you need to find out what level of acculturation is this Latino.
Luminar has access to names, addresses and other personal information about individuals. How do you deal with privacy concerns?
Rios: The privacy issue would come into play if I were sharing that data with my clients at the personal level -- and I'm not. I'm aggregating it to create personas, and we identify those personas at a group level and we start telling the behavior of the personas to our clients.
Your company launched Hortonworks last year. How were you handling this operation before that?
Rios: We handled it in the traditional way. We had a data warehouse with all of the legacy plusses and minuses that you run into in a data warehouse environment. But we realized that we needed to be significantly more agile in being able to ingest, cleanup, run our cultural filters into it and analyze the data and start deriving insights out of it. We are ingesting 2000 sources of data and the traditional tools for data processing were not cutting it.
For more on big data analytics
Read the Whatis.com definition of big data analytics
Find out why some companies are making big data analysis the norm
Learn about the challenges associated with big data projects
Are you still running a traditional data warehouse as well as Hortonworks?
Rios: Nope. It's off. Thank goodness we turned it off. Not [only] am I happy but my CFO is happy because of the savings. Currently, we do the [extract, transform and load] processing of data using Talend software and the data gets ingested either via Talend or using Sqoop directly into Hortonworks, which is, of course, Hadoop. Once it's in Hortonworks, we then use a combination of Hive and R, where R is being used to load our analytical models. Then, last but not least, the results of that are presented via Tableau.
What advice do you have for other companies considering a big data analytics initiative?
Rios: If anybody is starting to look into this technology, into migrating into this technology, don't try to do it alone. You've got to engage with someone to help you through the process. You've got to have the right technology partner or consulting partner to help you through this project. Also, don't be geographically limited in terms of who you evaluate [in terms of human resources]. We found talent in Latin America that was able to help us within this process and they worked very closely with Hortonworks.