Lytics Adds Marketing Recommendations to a Customer Data Platform

It’s just over one year since I first spoke with Lytics*, which at that time was (accurately) calling itself a Customer Data Platform but had not yet released a beta version of its product. The company has been busy since then, raising $7 million to supplement its initial $2.2 million funding, enrolling about 30 beta clients, releasing its initial system and a new self-service option, developing an automated process to recommend marketing programs to its clients, and abandoning the CDP label to call itself a “marketing activation platform”. CEO James McDermot said the label was changed because big companies thought a CDP sounded like an IT project, not something run by marketers. Fair enough, but Lytics still perfectly fits my definition of a CDP: a marketer-controlled system that supports external marketing execution based on persistent, cross-channel customer data.

In fact, Lytics could pretty much the poster child for the CDP concept. While many CDPs also provide some execution services, Lytics draws a sharp distinction between its core data layer, supporting analytics, and message delivery.  Data and analytics are included in the system; execution is not.  Also in the CDP spirit, Lytics makes extensive use of external products within its data and analytics layers, relying on third party systems to connect social media, email and postal identities; to import social and Web site data; for reporting; and to do natural language processing. All told, the company has prebuilt connectors with more than 80 software-as-a-service products. Execution systems on the list include Salesforce.com, Marketo, Eloqua, Act-On, Facebook, Twitter, Youtube, Demandware, Optimizely, Adobe Target, and most major email providers.

But perhaps I’m getting ahead of myself.  I should really start with what Lytics does.  Basically, it imports data from multiple sources, builds a consolidated profile for each customer, tracks individual behavior over time, builds segments of customers with similar behaviors, and makes those segments available to external systems for marketing messaging. It uses several data storage technologies, including Cassandra, Elasticsearch, and Titan Graph DB, to handle large amounts of structured and unstructured data. It combines its own identity matching techniques with third party resources to consolidate the profiles across channels, add more data, and extract meaning from text. It lets users define and extract audience segments and can push alerts to execution systems as customers change audience segments in real time.

Lytics would be a perfectly fine CDP if it did nothing beyond what I’ve just listed. But the system actually takes two additional steps – and is tip-toeing towards a third – that make it quite exceptional.

The first step is to summarize customer behavior with scores for interaction momentum, quantity, frequency, responsiveness, and intensity. These are combined to create about thirty segments, such as “burn out” customers, defined as people with high intensity and low momentum. The segments can be further qualified based on what addresses are available (email, postal, phone, Facebook, etc.) and on other profile data specified by the user. The resulting audiences give marketers a structured way to manage customer treatments.

The second step is to actually recommend those customer treatments. Lytics has a database of marketing tactics, such as reengagement programs for dormant users or upsell programs for active users. It looks at existing audience segments and the execution tools the client has in place, and calculates which tactics to which audiences in which channels would yield the highest results. It then recommends the most promising options to the client, who can activate the suggested program with the push of a button. This isn't actual program execution: Lytics only sends the audience to the selected tool, where the client must still set up the program and its message. But it's still a big stride towards helping marketers make choices that otherwise depend entirely on their own expertise. This is important because shortage of marketers with adequate skills has been a major stumbling block for many advanced marketing technologies.

The third step, which Lytics hasn’t yet taken, is to select the content itself.  McDermot was quite adamant that the company is not in the content recommendation business, leaving that marketers’ creativity. But he did say Lytics is experimenting with a “content graph” that classifies content and shows how it is related to individuals, which suggests the system will eventually be able to make some suggestions. There are other capabilities Lytics would need to make optimal content recommendations, notably decision rules to address business goals such as selling excess inventory or satisfying unhappy customers. These don’t seem to be on the company’s radar. But they could appear as it moves ahead.

So, about that self-service option. This might Lytics’ most impressive news of all. The initial release of the system was targeted at large enterprises and relied on traditional programing to connect with external systems using APIs. McDermot and I didn't discuss pricing but you can be sure it was in the five or six figures.  The self-service version enables automatic connections to the 80+ partners already in place. Pricing is based on the number of customer profiles and channels managed and includes all the existing connectors. It starts at a shockingly affordable $1,000 per month, making Lytics an option for just about any business. Combined with the product’s predictive scoring and tactic recommendations, this could empower a huge number of marketers whose firms couldn't previously afford a powerful marketing database and the integrations needed to make it useful.  We'll see how this plays out, but Lytics could be revolutionary indeed.


_________________________________________________________________________
* not to be confused with Lityx, which offers LityxIQ cloud-based predictive modeling and data management and is worth a look in its own right.