B2B advertising faces an ongoing challenge: reach “rare birds.” They don’t need a mass audience – they need the right audience. Our client, a large, US-based mutual fund firm we’ll call Fund Firm – faces an ongoing challenge: reach a niche financial advisor audience.

With the proliferation and commoditization of third party data sellers, the goal of reaching rare bird audiences seems increasingly actionable. However, source data can be unclear, and the programmatic activation nefarious. Like many high value B2B brands, Fund Firm was uncertain about leveraging those techniques.

Bloomberg B:Match, a data-powered insights engine, provides B2B advertisers unique, authoritative insights. Both a service and a product, B:Match connects advertisers with valuable audiences across Bloomberg digital properties with three core steps:

1. Match data: identifying the core customer and web behavior

2. Match for Insights: to provide valuable knowledge to advertisers

3. Match targeting: to activate insights through strategy and technology

Going beyond standard audience targeting or data re-targeting campaigns, B:Match mines data to deliver meaningful insights. Bloomberg’s digital visitors come in a professional context, thereby revealing behaviors that cannot be extracted from syndicated or survey-based research. These insights infuse strategies, solving the “rare bird” problem.

The four B:Match custom audience segments ensured that Fund Firm’s communications effectively reach a high-majority financial advisor audience – at times when this audience is focused on finding and absorbing information that is vital to their business success. Each segment was set live, in addition to a control group (users who belonged to neither segment A,B,C or D), via Doubleclick and the DMP.

Fund Firm had aggressive KPIs and ambitious goals based around three measures – which weren’t always aligned in terms of optimization. These key measures were:

rate of visitors landing on its website; total visitors landing on its website; and cost per visitor landing on its website. They were not looking purely at click through rate.

The data-powered campaign out-performed the advertiser’s previous campaigns with Bloomberg, and the data segments out-performed the control group. We saw 60x greater CTR lift and 12 times higher landing rate from the B:Match segments and 500% more landings than the control group. Most importantly the cost per landing for this campaign was at the peak among the media properties and the performance of the B:Match segments ensured that the client KPIs were met and exceeded.

Gather | Data Collection

Every user session on a website generates data points. Bloomberg.com visitors generate over 1.3 billion data points per day that are stored in our Amazon EC2 cloud-based server. In many organizations, this data is left in the warehouse to eventually expire and die. We used 2 months of web log data period for Fund Firm’s B:Match program. For the initial stage, Bloomberg spent 6 weeks analyzing data and building customer models, in partnership with the agency and client. Insights were then delivered back to the client and used to drive the final stage of a six-month media activation.

Harness | Bloomberg connects with the FA Customer

To isolate Fund Firm’s desired customers, we used data held in the Bloomberg Terminal to generate a list of companies with Investment Advice industry SIC codes. We analyzed that list against the web log data, matching IP addresses of visitors to a lookup of the SiC-coded companies to produce a core data set of Financial Advisor web behavior – without using any PII.

We analyzed only those who consumed at least two pages in common with each other. We call these our “content authorities.”

Interpret | Finding meaning in the data

Once the content authorities were identified, statistical analysis of their behavioral patterns was applied, including Naïve Bayes probability analysis. Comparing the behaviors of the content authorities to the behaviors of regular Bloomerg.com users brought to light the key differentiators in their behavior. Differentiators included what types of content they consumed, what operating systems were used, what browser types were preferred, what times of day were most likely, and more – allowing us to build a comprehensive, uniquely relevant picture of behavior among Fund Firm’s rare birds.

Data informing media strategy

The B:Match results were presented to the client with resounding success. The results revealed key behavioral insights about the customer segment that informed the media strategy and execution. Insights included media consumption habits of financial advisors – illuminating how they work in real time during an actual business day. This is not how they respond on a panel or to a survey; this is going deep below the covers to understand the content that drives their professional activity day to day. The insights emerging from the data are just as significant as the data targeting itself.

The results showed that financial advisors were most interested in consuming content around Bonds, in particular government and muni bonds quotes pages (data listing pages on Bloomberg.com). Financial advisors seldom read long form articles, instead preferring to consume vast quantities of financial and economic data.

Specific site areas where visitors were significantly more likely to be prospect’s Content Authorities included:

• Portfolio-focused videos

• Benchmark bond indexes

• Emerging markets news

• High-end real estate news

• Corporate bonds

• Financial advisors news

Data informing media execution

The insights derived from the data analysis were used to create audience targeting models that were executed via Bloomberg’s DMP. The analysis involved running 4 separate naïve Bayes models, depending on the criteria included in each model. Each model in turn was converted into a distinct audience targeting model in the DMP. For example, the reach model (segment A) looked exclusively at the types of pages that a Financial Advisor visits, controlling for all other variables. In contrast, the relevant model (segment B), analyzed types of pages combined with other features such as operating system, browser, country and time visited. Adding in these additional characteristics resulted in some interesting discoveries.