Patrick Moran knows marketing. He's made a career of using emerging online channels to acquire customers.
"I've never been able to be as data-driven as I am today," says Moran, who was formerly head of SMB Marketing at WebEx.
Moran's new company, New Relic (in which I am an indirect investor via Trinity Ventures), helps make web sites faster. The company's product gives software developers insight into what's consuming the most time so they can focus their efforts.
What's changed? Today, Moran can track and correlate nearly all marketing activity, product usage, and customer behavior. He no longer needs to guess. He can run marketing experiments and make decisions based on what works.
The company hired a data scientist to analyze its marketing and product usage data, using technologies like Hadoop and the analytics package R.
Here's how Moran uses Big Data to do marketing:
1. Get all the data in the cloud. A SaaS company, New Relic uses a range of other SaaS applications to run its business, including products from marketing automation company Marketo, customer service ticketing company Zendesk, Salesforce.com, and Facebook, Google, and Twitter for online ads and promotions.
2. Run marketing campaigns. The company spends about $150,000 per month on Twitter and $10,000 per month on paid search.
3. Capture the data. The challenge for Moran was putting all the marketing data together with product usage and sales data in a form that would enable asking key marketing questions. For example, which of the following drives more revenue?
- Signup form (which A/B test)
- Time of day (of tweet)
- Did click come from retweet?
- Blog post or content promoted
- Path within New Relic's site-
- Retargeting ads seen (if the potential customer didn't sign up the first time)
- Programming language (Ruby/Python, Java, etc.)
- Geography
- Deal size
- # of Helpdesk tickets received
- # of servers on which customer has deployed New Relic's monitoring software
The company needed to combine this data with customer product usage data:
- Real user monitoring?
- Set up server monitoring?
- App throughput?
- Slow performance?
- Login frequency
And with sales rep data:
- # of contacts
- Days in pipeline
4. Get the data in one place. To be able to run the analysis, Moran's team pulled all the data into Hadoop, an open source software package for storing large amounts of unstructured data. New Relic does this "about once every two weeks," says Moran.
5. Analyze. New Relic hired a data scientist. The data scientist analyzes the data using R, an open source statistics and analysis package. She looks at the correlation of data from disparate systems (Marketo, homegrown, product, and Salesforce.com) with New Relic users.
One example of this is product usage tied back to an original tweet. She also evaluates which campaigns drove active usage as opposed to users who signed up but did not convert.
6. Create a profile. "The end result is a 'profile' that we then create cohorts upon. From there, we can recreate success."
Companies like GoodData, which recently announced additional funding of $25 million, provide business intelligence capabilities on top of Salesforce.com and other cloud offerings. According to Moran, New Relic has not "invested the cycles" to evaluate whether GoodData would meet the company's needs. Moran has been testing Mixpanel and Kissmetrics.
But Moran has learned that when it comes to data analysis, "structure is not great. We need to pivot on multiple axes to discover real value. Any one tool is consciously in the 'structure your data business,' so sometimes you need to pull it out and visualize it in new ways."
