I spent much of last year writing about Customer Data Platform systems and have reviews on tap for a half dozen more. But I thought I’d start out 2014 with something different, just to show I’m not totally obsessed. Although, as you’ll see shortly, there’s a CDP angle to this story as well.
Today’s topic is BrightInfo, which uses semantic technology to automatically recommend the most relevant content to Web site and blog visitors. Specifically, the system crawls the client’s Web site and blog to find and classify existing content, and then tracks visitor behavior to offer new content relevant to what the visitor has already selected. The recommendations can be presented on a fixed area of a page or as a pop-up. They can appear continuously or only when a visitor indicates they are about to leave by moving their mouse towards the browser’s URL bar. BrightInfo uses Javascript tags and cookies to track visitors over time, allowing it to base recommendations on individual behavior as well as content similarity and popularity. Behaviors across Web sites, landing pages, and blogs are all tracked by the same cookie, so each recommendation reflects a consolidated history. However, the system does not incorporate other information sources – meaning it can’t support mobile platforms (which don’t allow cookies) or use the central customer database a CDP would provide.
It would probably be fairly easy for BrightInfo to support mobile and access external data sources, since the necessary changes are unrelated to its core technologies of semantic analysis and recommendations. Most companies today could probably use the system as it stands, since they lack a centralized customer database or policies to coordinate customer treatments across channels. BrightInfo already lets users override the purely algorithmic recommendations by specifying that some content will be shown in all circumstances, that other content will never be shown, and that recommendations will appear only on specified pages. Sophisticated marketers might want more refined controls, such as limits on how often the same content is offered or recommendations based on expected response value rather than the simple click rate. But BrightInfo is targeted at small and mid-size businesses, which are less concerned with such refinements.
What those businesses do care about are easy deployment and low cost. BrightInfo provides those by automating the content discovery and classification, running as a service rather than installed software, and pricing based on visitor volume. Javascript tags, cookies, and isolation from other data sources also simplify deployment, whatever their other drawbacks. Measurement is similarly simplified by providing reports that compare how many clicks were made on native content and system-recommended content. Clicks on system-recommended content are a rough measure of system-added activity, although presumably some visitors would have chosen other content had the recommendations not been available. BrightInfo considered setting up formal a/b tests to measure true incremental value, but found that most small and mid-size businesses have too little volume to support this.
BrightInfo officially released its product last September, after about a year of development. The underlying semantic and recommendation technologies came from sister company Softlib Software, which uses them for automated service and knowledge management and was founded in 2004. Pricing is published on the BrightInfo Web site and is free up to 1,000 visitors per month, $89 per month up to 5,000 visitors, and $224 per month up to 15,000 visitors. The system has several dozen clients.
To summarize, then: BrightInfo provides a very simple, very low cost way to increase engagement with Web visitors by making targeted content recommendations. It’s worth knowing about because traditional recommendation engines are often harder to deploy and more expensive.
But what’s the CDP angle? It’s not simply that BrightInfo is an example of an application that could use the customer data in a CDP to make more accurate recommendations. It’s actually a somewhat deeper question of where the recommendation functions belong in a CDP-based architecture. I’d argue that recommendations should be part of the central platform, so they can be used to coordinate treatments across all touchpoints. In other words, it’s probably wrong to imagine BrightInfo as an application that attaches to a CDP and uses its data to improve Web and blog results. Rather, in an ideal world, BrightInfo’s technology would be used within the CDP to generate recommendation that the CDP itself feeds to all applications. This is pretty theoretical and largely irrelevant to BrightInfo itself, which is targeted at companies that don’t have a CDP in the first place. But as marketing technology continues to evolve and more companies have CDPs, or centralized customer databases by any other name, it’s important to understand how the pieces should fit together.