I led a session at the National Center for Database Marketing (NCDM) conference in Miami last month on how to “Target Customers Effectively Through Advanced Analytics” with Arup Ray from Intuit. Full disclosure: my employer, Netezza, has a business relationship with Intuit and I am therefore limiting what I share in this post to information already in the public domain.
Intuit is No. 1 in every market it serves. Twenty million consumers use its TurboTax software and another 15 million use Quicken. 50 million people use Intuit’s products every year. The 27 year-old company has an impressive history of producing high-quality products.
In recent years, many of Intuit’s products have been transitioning away from shrink-wrapped desktop software distribution toward SaaS web-based delivery. As a result, Intuit now invests heavily in digital advertising for customer acquisition, buying 20 BILLION display impressions in the fourth quarter of 2010 alone. Critical to their success is deep in-house expertise in utilizing large-scale, high-performance data analysis to maximize the effectiveness of their digital advertising campaigns.
Intuit uses DoubleClick to monitor all of their online acquisition channels and Omniture to track visitor movement across their web properties. But those tools don’t enable Intuit to track and analyze individual web behaviors. Nor do they facilitate behavioral targeting, time series analysis, or predictive modeling and scoring. Therefore, Intuit unifies nearly 40 terabytes of DoubleClick and Omniture data in their “customer experience data warehouse”. Intuit’s data warehouse is a strategic asset for their business. It supports a broad variety of custom solutions focused on customer acquisition and retention, including attribution analysis, visitor-level behavior analysis, channel analysis, website optimization (and A/B testing), and more. Intuit discussed these solutions and highlighted four key digital advertising insights at NCDM:
Insight #1: Attribution analysis creates value for Intuit by enabling better decisions that maximize channel effectiveness and minimize channel interference.
I’ve written about how attribution analysis solutions create value in the past (see “Attribution Analysis and Campaign Efficiency – Getting More Bang for your Buck“), and Intuit’s experience reinforces many of those ideas. Intuit’s key goals for attribution analysis are as follows:
Maximize channel effectiveness:
Intuit utilizes display advertising, paid search, organic search, affiliates, and email for their digital campaigns. Optimizing and coordinating spend across all of these channels is critical to Intuit’s success. One specific example cited by Arup:
We were wasting a lot of money by continuing to advertise to customers that had already converted, and we were annoying those customers by showing them irrelevant ads. But now we are able to filter converted customers out of our campaigns so that we no longer waste those impressions, and even better, we often now use those impressions to cross-sell with more relevant offers.
Minimize channel interference:
This topic led to one of the session’s more interesting exchanges, when Arup revealed that – contrary to conventional wisdom – the “last click” in Intuit’s campaigns are typically NOT coming from search advertising:
Our data shows that if users became aware of our products through a display banner ad, typically they would follow a display banner ad to come back, use our product, and convert. So the data is saying that it is the same channel through which they became aware, that they convert. This was a new learning for us – the fact that on our campaigns, click-thru behavior is very consistent across the purchase funnel and we see little cross-channel overlap. We don’t know exactly why this is the case, but we spend a lot more on display than we do on search, and it could be that this is part of what drives this channel-centric behavior.
Insight #2: Visitor-level web behavioral analysis for ad targeting creates value for Intuit – increasing conversions and subscription revenue.
Intuit analyzes every click from every user of its products, creates behavioral models from that data, and matches current users against those models in order to take action in real-time – targeting each individual user with relevant messages and experiences – both within Intuit’s products and across the Internet. Note that achieving this person-centric view of website visitor behaviors is essential for efficient marketing – an important concept I wrote about in the “Seven Reasons Your Website Analysis Belongs in a Data Warehouse” – but it’s also critical within Intuit’s desktop software.
One specific example we discussed involves Intuit QuickBooks, the popular desktop accounting software product (there is an online version, too). Intuit collects in-product discovery (IPD) data from QuickBooks users with custom instrumentation built into its software (data is only collected from users that explicitly opt-in). Analysis of this IPD data enables in-product behavioral targeting and turns QuickBooks into an important sales channel for Intuit Payroll and Payment Solutions. For example, if Intuit sees that a small business has hired employees, it will offer Intuit Payroll—immediately. The final outcome of this in-product behavioral targeting: a “non-trivial” lift in revenue.
Insight #3: Funnel-aware messaging creates value for Intuit – both onsite and across their customer acquisition campaigns.
The QuickBooks scenario cited above is a good example of this, and Intuit take this even further, incorporating some of the ideas I discussed previously in “Using Dynamic Message Optimization To Increase Campaign Response Rates“. Arup also stated:
We have created a predictive model that detects within a few clicks whether a user’s navigational behavior is likely to lead to a conversion, and if not, we present a “click-to-talk” window offering assistance or possibly even make a one-time offer.
I’ve written about this idea a lot so I won’t go into detail on this point, other than to say that Intuit’s experience illustrates the significant value that can be captured with an analytical approach to delivering the right messages to the right people at the right times based on their positions in a conversion funnel.
Insight #4: It’s nearly impossible to execute on the first three insights without a data analysis platform that enables extremely rapid execution of complex analytical queries against extremely large data sets.
Rapid analysis of granular, record-level detailed data is required in order to make all of the above strategies work. The volume of data involved and the complex nature of the analytical queries that are required create fundamental technical challenges that can hinder even organizations full of great engineers. As Arup said, “Before Netezza we simply did not have the ability to track individual users at the level of a cookie.” Good engineers can always find workarounds for their big data challenges, but solving these types of problems in a way that simplifies development and delivers high performance is particularly hard – and particularly valuable.
Arup highlighted the keys to Intuit’s success in his last slide – note his emphasis on reducing data latency (both for data loads and query processing) – a tactic that is proven to increase targeting precision and campaign lift:
Here is the complete slide deck for your reference:
If these insights pique your interest, you might also appreciate this webinar recording that my colleague, Krishnan Parasuraman, led with Intuit’s Vineet Singh and Predictive Analytics World‘s Eric Siegel just a few months ago discussing Intuit’s data analysis platform in greater detail: “The New Age of Analytical Marketing“.