HubSpot's Latest Marketing Software Sends the Right Message

Poor targeting

HubSpot yesterday launched a “completely re-envisioned and rebuilt” version of its marketing system at its Inbound 2012 user conference. The main thrust of the new release was dynamic customization of emails, landing pages, and forms based on each contact’s profile and behaviors. This is a major expansion beyond HubSpot’s original focus on “inbound marketing” to attract leads.

Specific changes include:

- a new contact database that is much more flexible than the original HubSpot database, allowing access to all types of email and landing page interactions within HubSpot and to social media activities imported to the system. The new database is built with HBase, which accesses Hadoop files. More on that later.

- “smart lists”, which are rule-based definitions of contact groups whose membership is updated automatically as contact data changes. Apologies if that’s a bit jargony; it just means the lists are always current.

- “smart calls to action” which are dynamic content blocks driven by the smart lists. That is, users define which contents go to members of different lists. The blocks are stored in a library and the same block can appear within multiple emails, HubSpot landing pages, or external Web pages.  In practical terms, this means things like: users who have already downloaded one piece of content can automatically be offered something else.

- “smart forms” (do you sense a pattern?) which don't repeat questions a client has already answered. This isn’t quite true progressive profiling, which would replace questions that are answered with ones that are not.  But it removes a major annoyance.

- workflows (hah! Bet you expected “smart flows”) that are triggered by smart list-style rules and can include multiple steps with multiple actions assigned to each step. Available actions include changing contact data, sending a record to CRM, updating a lead score, setting a lifecycle stage, and changing the call to action.

- social media tracking that captures responses to system-generated social media messages within the contact database.  The responses are associated with specific individuals, so they can be used in smart lists and workflow rules.

- iPhone apps to view some reports and individual contact data

This is all good stuff, although far from revolutionary. Dynamic email content, for example, is available in 14 of the 22 systems in our VEST report. HubSpot recognizes that these are not new features but argues they’ve made them easier to use than competitors. I’m not so sure – the rule builder underlying the smart lists and workflows looks pretty much like every other rule builder, and the workflows themselves are also similar to the sequential flows in other systems.

This isn’t a criticism of HubSpot, but just a recognition that these are inherently complicated features which plenty of smart people have already tried to simplify. Radically better approaches may yet be found – I had an interesting chat about some possibilities with HubSpot co-founder Dharmesh Shah – but so far, the state of the art is what it is.

The new release also includes improved email and landing page designers, A/B testing for landing pages (not available in the entry-level version of the system, alas), enhancements to the app and service marketplaces, and expanded training services. The company said the coming year will bring enhancements to existing components including the blogging and search engine optimization applications.

What’s really important about all these changes is not whether they’re unique, but how well HubSpot has pulled them together and how it teaches its clients and resellers to use them. This is where HubSpot has always been strongest, and the vision it set out this week – of highly relevant marketing messages for each individual – is indeed advanced. (It’s also one I agree with – see this post on why the marketing funnel is dead.) If HubSpot can get marketers to focus on that sort of targeting, which is quite different from traditional campaign-oriented promotions, they can indeed have a revolutionary impact on their clients and the marketing industry.

And what about HBase? Although HubSpot didn’t talk about it in its marketing materials, switching from a conventional relational database to the Hadoop-based system is almost certainly the most radical feature of the new release. So far as I know, HubSpot is the only marketing automation system using HBase.

I discussed this a bit with HubSpot Chief Product Officer David Cancel, who joined the company when it acquired Performable, which was itself built on HBase. Cancel said HBase takes more resources than a conventional database engine but provides direct access to all details of each contact’s behavior history. One immediate benefit is that HubSpot now allows custom fields – up to 1,000, in fact – which it didn’t previously.  Ad hoc reports against the HBase data isn't available yet but is due before the end of 2012.

Longer term, I suspect HBase will make it easier to add custom objects and to deal with unstructured and semi-structured data such as Web logs and text comments. This could make HubSpot fundamentally more flexible than most B2B marketing automation systems, whose data structures are tightly linked to CRM data structures. As I mentioned last week, the main exceptions to that rule today are the high-end marketing automation products, which were built for consumer marketing applications and assume a custom data structure. Having that flexibility in product for small-to-mid-size businesses could open up some possibilities that truly do make HubSpot unique.

(Wondering about the alligator man picture?  Well, one of the sessions at the HubSpot conference said that having pictures in your blog posts increases readership, so I thought I'd give it a try.  If you want me to justify that particular image: she's getting a message she doesn't want.)

Kwanzoo Builds Content for Cross-Channel Marketing

I first bumped into Kwanzoo about a year ago at a conference trade show and was frankly puzzled at what they offered. The mechanics were clear: a tool to generate HTML-based forms, surveys, banner ads, and social sharing links that could be used on Web sites or embedded in emails. What puzzled me was the advantage of this over anyone else’s HTML content, including the content that could be generated using standard tools within most marketing automation systems.

Since this particular mystery ranked somewhere between the fate of Amelia Earhardt  and Nacza Lines  in my personal priorities, I didn’t investigate further. But the company reached out to me recently and provided a clearer explanation. Here’s the story.

What makes Kwanzoo special is it creates a sequence of Web pages that can be deployed as a single unit. A typical sequence would be a survey followed by different offers depending on the visitor’s answers. The entire sequence is built in Kwanzoo and deployed as a code snippet which displays the survey and calls the subsequence pages from Kwanzoo when appropriate. The second thing that makes Kwanzoo special is that its pages can be deployed on a client’s own Web site, on external sites and ad networks, embedded in emails, within Facebook, or on a mobile device. Users can also apply Kwanzoo tags to conversion pages to track results.

These capabilities make Kwanzoo substantially more versatile than a conventional marketing automation system, which would rely on campaign flows or manually-embedded links to manage the page sequence and could only deploy on emails or microsites generated by the marketing automation system itself. By contrast, users design the flows by filling out a simple form within Kwanzoo and then receive HTML snippets – simple calls to Kwanzoo-hosted URLS – that can be embedded anywhere. Kwanzoo also provides a powerful editor to build the Web pages and offers.

Data captured by Kwanzoo can be directly posted to Eloqua, Marketo, Constant Contact and Salesforce.com. The Eloqua integration is especially elegant, using an Eloqua Cloud Connector (i.e., a parameterized API call) that makes the Kwanzoo pages available within Eloqua’s own content builder and can read Eloqua cookies in real time to help guide the response selection. Integration is on the way with other marketing automation and CRM systems.

Users can also apply IP-address-based visitor identification to tailor responses to named accounts and different industries.

Kwanzoo says this versatility addresses some critical pain-points for marketers, including needs to create content and capture data across multiple channels and to create more personalized interactions. True enough.  But the system has its limits, most notably that sequences are limited to a couple of steps, a few data inputs, and a handful of actions, and that it doesn't maintain its own marketing database.

In other words, Kwanzoo is more a bridge between different marketing channels than an integrated marketing system. It’s easy to imagine Kwanzoo-captured data making its way back to multiple systems (marketing automation, CRM, Web site, mobile, etc.), which is not the ideal situation. There’s certainly a market today for this approach: Kwanzoo has landed about 25 clients since its launch in 2010 and seems to be growing nicely. Still, you have to wonder whether integrated platforms will eventually add similar capabilities tied directly to their own databases, making Kwanzoo’s external bridge less necessary.

Or, even more directly, will Adobe offer pretty much the same capability?  They already have the dominant tools for content-building (Dreamweaver, etc.) and Web visitor tracking (Omniture), which are the two key pieces of Kwanzoo's offering.  I long ago predicted that Adobe would combine these to create "smart content" that would adjust to customer behavior, although so far Adobe hasn't listened.  But they could.

But that’s Kwanzoo’s problem, not yours or mine, eh? At the moment, it’s worth a look. Pricing is based on number of impressions and starts as low as $499 per month to put Kwanzoo units on Web pages.  It quickly rises to $2,499 to embed units in emails, post data other than basic lead capture, support mobile formats, and use IP-address information for targeting.  

Conversen Simplifies Complex Messages Through Multi-Channel Dynamic Content

Summary: Conversen makes it easy to generate dynamic messages across multiple channels. It's more a supplement than a replacement for conventional campaign management but should save a lot of work for marketers and their agencies.

One of the fundamental challenges in database marketing is that a seriously sophisticated campaign may send different messages to hundreds or even thousands of customer segments. The traditional approach has been to define these segments during the selection process, creating a tree with one end-point for each segment, and then to assign the appropriate message to each end-point. The problem is that this requires creating hundreds of versions of the messages and making sure that each is matched to the correct end-point. This is both labor-intensive and error-prone.

An alternative is to create "dynamic content" the messages that select the appropriate contents for each individual. In essence, this is moving some of the segmentation logic from the selection process to inside the message. Even though this ultimately produces the same number of variations, it lets marketers create fewer messages and segments, reducing manual effort.

Let’s take a concrete example. Suppose you’re sending offers for winter vacation travel. People in New York will be sent offers for Florida and people in Los Angeles will get offers for Mexico. In addition, people in high-income zip codes will be offered a deluxe package while those in middle-income zip codes get an economy offer. A segmentation-based approach would use three segmentation rules (New York or Los Angeles; if New York, high or middle income; if Los Angeles, high or middle income) to create four segments, each tied to a separate message. A dynamic content approach would require just two decisions (New York or Los Angeles, high or middle income) that are each tied to a specific content block.


It’s still possible to make a mistake: you could accidentally link the Mexico offer to New York. But each assignment is made only once so it’s easier to be sure it’s correct.

Note that the advantage of dynamic content increases as you add complexity: a three city-pair, three level program would require four segmentation rules (one for city, three for city/level combination) and nine unique messages, while dynamic content still needs only two rules (one for city, one for level) and six message blocks (three destination cities, three luxury levels).


So where’s the catch? Well, dynamic content requires the marketing automation vendor to work inside the message itself, using different technologies for each medium. This is significantly trickier than just pointing each segment to a message created elsewhere.

One way to avoid this complexity is to generate a file containing the customer records and segmentation variables and let channel-specific output systems generate the customized messages. But this adds its own costs and risks, since the external systems must be configured separately for each project. As a practical matter, most high-end marketing automation vendors have compromised by providing dynamic customization for email and Web pages, and letting external systems handle the other channels.

Conversen has taken a different approach, building a specialized system to support dynamic content across as many channels as possible. This puts it in a somewhat confusing business position, since it can sometimes replace a traditional campaign management system but more often receives output from one. Resolving this confusion is largely Conversen's own problem, however, since it sells to marketing service providers rather than end-users.

Conversen is organized primarily around campaigns. These include filters to select an audience, processing steps and content. The key here is consistency: the rules used in filters, steps and dynamic content are exactly the same. It's not just that they're built with the same interface and run against the same data structures: the same rule can actually be used for any purpose. Rules can also be shared across multiple campaigns and referenced within other rules. This reuse substantially reduces the number of rules needed, and thus both the effort and opportunity for error.

The rules themselves are quite powerful, extending beyond the usual selections on field values to include advanced features such as if/then/else loops. One gap is missing support for a/b testing, which Conversen decided to omit because it added too much complexity. The system doesn’t maintain an audit trail of changes to each rule, but does provide reports listing everywhere each rule is used. This helps to avoid unintentional consequences when a rule is changed.

Rules connect with data gathered from source systems through batch processes or a real-time API. The resulting database is stored in Microsoft SQL Server and hosted by Conversen. This is important point, since it means that Conversen doesn’t simply attach to an existing marketing database. Although moving data into a separate database does add some cost, it also provides options to maintain persistent customer histories, combine data from multiple sources, and directly capture events such as campaign responses.

The system includes basic features to define data structures and map data from external sources into those structures. Load maps can include basic rules for whether to update or append matching records, but more advanced processes such as name/address matching have to be done externally.

Users who don’t need any of these functions could simply send Conversen the output files from a conventional campaign manager. This costs no more than loading files into any other message delivery system.

The heart of Conversen are the marketing messages. Conversen defines each message as an XML template. This holds any static elements plus the rules used to select content blocks.

The blocks themselves are created outside of Conversen and stored in a content library. This is another example of Conversen drawing the line between its core functionality and supporting functions to be handled elsewhere. It also probably reflects the reality that content will be created by external vendors, such as ad agencies, who will want to use their own tools in any event. Lack of an integrated content-builder does mean that personalization tokens such as [First Name] must be manually embedded within the content block. This can be done in the original content creation system, requiring a relatively inconvenient cut-and-paste from a list provided by Conversen, or be added after the content is loaded into Conversen.

Each Conversen content block currently supports a single medium. Thus, there would be separate content blocks for 10% discount in email, Web, direct mail, mobile and other types of messages. Conversen is working on multi-media content blocks that could be inserted into any medium. This would further simplify marketers’ lives.

One Conversen campaign can deliver multiple messages over time, based on dates such as a contract expiration or recent activity, or on events such as promotion responses. The system can react to qualifying events at regular intervals or in near-real-time as they are posted.

Clients can also build custom interfaces by direct access to the Conversen API. This lets them create branded systems and offer specialized portals with limited functionality. These might give designers access to the content-management features of the system, or make predefined campaigns to available to field offices.

Conversen supports email, mobile (SMS), RSS feeds such as blog posts, print, call center and Web. The system provides specialized services for each channel, such as rendering to preview emails and postal sorting for direct mail. Print output is integrated with Bitstream PageFlex, which supports direct output to high-speed printers. Conversen sends the digital messages itself and ships print and call center files to third parties for execution.

The system also provides operational reporting on campaign volume and responses. The reports are designed to provide activity information rather than detailed marketing analysis.

Conversen was introduced in 2007. The company now has about 25 marketing agencies as customers, serving more than 125 end clients. The system is offered only as a Conversen-hosted service. Pricing includes a $15,000 setup fee plus $1 to $20 per thousand messages based on volume and type.