Intercom "Smart Campaigns" Replace Decision Trees: Interesting But Not Perfect

I got all excited when I saw this description from messaging vendor Intercom about new "smart campaigns" in marketing automation that automatically send the best message at the best time in the best channel to each person without pre-designed campaign flows.  Their critique of the current process -- essentially that fixed flows are too complicated -- is spot on.

Alas, a deeper look left me a little disappointed.   Here's how Intercom describes the smart campaign process:

1. First, choose the people you want to message and the goal you want to achieve, e.g. send a series of messages to people who start a trial to get them to become paying customers.
2. Then decide how often you would like them to receive messages, e.g. you may want to send, at most, a message every two days.
3. Choose triggers for your messages, based on time, behavior or interaction with other messages.
4. Then simply rank them by priority, with the most important message listed first.
5. When people are eligible to receive a new message, Intercom looks at all the messages in the campaign, identifies the ones the customer matches the rules for and sends them the highest priority message.

 My problem is step 4: messages are ranked by priority.  This means that everyone receives basically the same sequence unless there are triggers that interpose specific messages first.  So, the smart campaigns aren't really figuring out the best message to send; they are applying static rules to pick the messages.

This is still pretty impressive but it puts most of the work back on the user to figure out those triggers.  It doesn't automatically adjust the core priority ranking (which drives the default message sequence) based on user attributes or behaviors.  I'm sure that clever trigger design could achieve pretty much any use case I could imagine, but it means all the thought I previously put into building clever campaign flows now goes into building clever triggers (and to predicting the customer experience resulting from interactions among those triggers).  So the Promised Land of fully automated, optimized campaign design still hasn't been reached. 

Note: I haven't spoken with Intercom.  I'll try to find time for that and to write a real review.  But I did want to put this out because it's a good example of people thinking about alternatives to the current marketing automation campaign flows, even if they haven't found a perfect replacement.

Further note: I did subsequently speak with Intercom to learn more about their system.  It turns out that my initial description, presented above, was accurate.  They do make it pretty easy to connect with Web, mobile app, and other data sources and to send messages by email, text, and in-app.  They also let users assign goals to each campaign as a whole and to individual messages, which helps to report on campaign success and to optimize treatments within a campaign.  So they do have some nice features that make effective messaging much easier than standard marketing automation systems.  But fully automated, self-optimizing campaigns, they are not.    

Iterable Offers Mid-Size B2C Marketers Powerful Campaigns in Outbound Channels

As William Shakespeare never wrote, some systems are born with data, some achieve data, and some have data thrust upon them. What the Bard would have meant is that some systems are designed around a marketing database, some add a database later in their development, and some attach to external data. The difference matters because marketers are increasingly required to pick a collection of components that somehow work together to deliver integrated customer experiences. This means that marketers must first determine whether they're looking for a system to provide their primary marketing database (since you only need one of those), and then figure out which products fall into the right category.

Whether you need a system with its own database ultimately depends on whether you have an adequate database in place. Obviously the key word in that sentence is "adequate".  How that's defined depends on the situation: key variables include the number and types of data you need available, how quickly new data must be processed, whether source data is already coded with a common customer ID, and how you want other systems to access the data.

As I wrote last week, there are a handful of Customer Data Platforms (CDPs) that do nothing but build a database. Many more systems build a database as part of a larger package that also includes an operational function such as predictive modeling or campaign management. This offers an immediate benefit but it complicates the system choice since you have to judge both the database and the operational features. It’s also trickier in a more subtle way because some systems build a great database but don’t make it fully available to other products. That’s spelled s-i-l-o.

These musings are prompted by my attempt to come to assess Iterable, a product I generally like but find as slippery as one of Shakespeare’s cross-dressing heroines. Iterable definitely builds its own database, using the JSON API and Elasticsearch data store to manage pretty much any kind of data you might throw at it. This can happen in real time (yay!) or via batch file imports. The system even provides its own Javascript tag to post directly from Web pages and emails. It organizes the information into customer profiles that can include both static attributes and events such as transactions.  That’s pretty much what you want in your marketing database. Elasticsearch lets the system scale very nicely, returning queries on 100 million+ profiles in seconds. Yay again!

On the other hand, Iterable doesn’t let other systems query the data directly. Users can do analytics and build segments using Iterable’s own tools or export selected elements to other systems in a file.  They can also push data to other systems through integration with the Segment data hub.  So while Segment might be the core database supporting other marketing systems, Iterable will not.  Nor does Iterable do much in the way of identity association: new data must be coded with a customer ID to add it to a profile. This is a pretty common approach so it's not something to hold against Iterable in particular.  Just be aware that if you need to solve the association problem, you’ll have to look outside of Iterable for the answer.  Fortunately, there are plenty of other specialized systems to do this.

Perhaps Iterable provides so many operational functions that there's no need for other systems to access its data?  The answer depends on exactly what functions you need.  Iterable provides a flexible segmentation tool that can build static lists and can update dynamic lists in real time as new data is posted. This can be combined with exceptionally powerful multi-step workflows, including rarely-seen features such as converging paths (two nodes can point to the same destination) and parallel streams (the same customer can follow two paths out of the same node). It also supports more common, but still important, functions including filters, splits, a/b tests, waiting periods, API calls to external systems, and sending email, SMS, and push messages. One notably missing feature is predictive modeling to drive personalized messages, but Iterable recently set up an integration with BoomTrain to do this. Iterable still doesn’t offer Web site personalization although it might be able to support that indirectly through BoomTrain, Web hooks, or Segment.

Iterable includes content creation tools for its messaging channels – again, that's email, SMS, and push.  This means users must rely on third party software to create forms and landing pages.  Nearly all B2B marketing automation systems do have form and page builders, but Iterable is targeted primarily at mid-tier B2C marketers, who are less likely to expect them.  Iterable’s B2C focus is further clarified by its prebuilt integration with Magento for ecommerce and with Mixpanel and Google Analytics for mobile and Web analytics. The system also provides a preference center to capture customer permissions to receive messages in different channels – a feature that is essential in B2C, although certainly helpful in B2B as well.

So where does this leave us? Iterable is more powerful than a basic email system but not quite as rich as full-blown marketing automation, let alone an integrated marketing suite or cloud. Page tags, JSON feeds, and Webhooks make it especially good at collecting information, although it will need help with identity association to make full use of this data.  It builds powerful outbound campaigns in email, SMS, and mobile apps.  Ultimately, this makes it a good choice for mid-size B2C marketers who want to orchestrate outbound messages  but are less concerned about Web pages or other inbound channels. Marketers could also use Iterable as the outbound component of a more comprehensive solution with Segment or something similar at the core.

Iterable was founded in 2013 and first released its product about a year ago. It currently has more than 30 clients paying an average around $3,000 per month. List prices start much lower and some clients are much larger.

Campaign Management Is Dead. Here's What Next-Generation Marketing Automation Looks Like.

Scientists tell us that the attention span of the average human is now shorter than the attention span of a goldfish.(1)  In such a world, the chances of anyone reading this 2,000 word blog post are pretty much nil. But I think the topic is extraordinarily important, so here is a summary in sushi-sized bites:

- the conventional flow-chart model of marketing campaigns can’t capture the complexity of today’s  disjointed customer journeys.

- a new approach is emerging that identifies stages in the customer journey and picks “plays” (small, highly targeted sets of treatments) to execute in specific situations within each stage

- this approach is easier for marketers to manage because it lets them think in smaller, more comprehensible units

- it will eventually lend itself to greater automation as machines take over more of the marketer's job in a “madtech” world

You can stop here if you need to watch an important cat video.  But if you want to understand my thinking in more detail, please read on.

The first modern campaign manager, Third Wave Network’s MIND, was released in 1991. What made it modern was a standard relational database(2) and multi-step campaigns that that sent users down different paths depending on their actions during the campaign. The system displayed each campaign as a set of boxes connected with lines to represent movement of customers from one step to the next. This flow chart interface has been fundamentally unchanged ever since and remains the gold standard of marketing automation(3).

The significance of the flow chart is what it replaced. The previous standard was a list of segments, each described on one row of a (paper) spreadsheet, with characteristics including selection criteria, description, key code, promotion materials, and quantity. Marketers filled out these sheets and handed them to programmers to execute. This was how direct mail marketers worked for decades, and often still do: it’s an efficient way to manage dozens or hundreds of cells within a large outbound campaign. It supports multiple contacts, such as mailing a second catalog to high-performing segments several weeks after the initial drop. But the list for the second mailing is pulled at the same time as the first, so it doesn’t allow changes in treatment based on subsequent customer behavior. This adjustment is what branching flow charts provide.

A quarter century after its introduction, the flow chart is now ripe for replacement. Flow charts assume that customers will follow a small number of predefined paths. This was realistic when interactions were limited to a few company-controlled touchpoints.  But it doesn’t describe today’s self-directed, random-walk journeys through an ever-shifting media landscape. In this environment, the best a company can do is react intelligently wherever a customer appears, taking into account both the current situation and whatever it knows about the customer’s past. Even company-initiated messages, while not wholly reactive, must be consistent with other treatments.

The basic features of a better approach have been obvious for years. I’ve long described it as “do the right thing, wait, and do the right thing again.” What’s changed is visibility: marketers now can see vastly more data about their customers and can interact with them through vastly more channels. In the “madtech” vision I’ve been articulating recently, this translates to assuming that all data about each customer’s demographics, interests, behaviors, locations, intentions, and other attributes is available to everyone. This includes both data a company gathers through direct interactions and data aggregated by third parties and offered for sale. Indeed, third party data is essential for building a complete picture of each customer’s experience.(4)

The vision similarly assumes that messages can be delivered through channels the company does not own directly. These extend beyond conventional advertising to private channels that other companies have opened to external messages. A concrete example would be third party offers delivered through a company’s Web site. My shorthand for this is “everything is biddable”: meaning that marketers can pay to embed a message within every interaction the customer has with anyone. Since all marketers have the same opportunity to bid on all impressions, a corollary is that buying the right messages at the right price is the key to success – or, in another catchphrase, “the smartest bidder wins”.

Sadly, neither you nor I can make a living by repeating clever phrases. Someone has to do the hard work of figuring out what treatments to deliver in each situation. A flow chart can’t come close because the situations are far too varied for any one chart to capture all the alternatives. What’s needed is a system that will assess each situation as it arises and come up with a custom-tailored solution.

If the only goal were maximizing immediate response, this would be pretty simple. Existing recommendation engines and predictive models can easily tell you which content a person is most likely to pick or which product they are most likely to buy. Today's products often do this with limited information about the individual being targeted, but that’s just a reflection of what data is currently available: at least some current systems could incorporate individual details and history without a major revision. There are practical details of speed, scope, and accuracy but the broad design of such a system is straightforward.  It connects with every touchpoint (including exchanges that deliver external opportunities up for bidding), receives information about each interaction, and returns an optimal message along with the value to bid for the right to deliver it.

But immediate response isn’t the only goal. Each interaction is embedded in the context of a customer’s relationship with the company. The best message is the one that maximizes the long-term value of that relationship. This won’t necessarily be the message with the highest response rate or greatest immediate financial value. Automated systems can incorporate long-term value in their recommendations but only if they are tracking long-term outcomes and analyzing what changes them. I believe – and this is the main point of this entire post – that automated analysis can only optimize for long-term outcomes if customer data is classified by stages in the customer journey. In other words, you can’t just randomly test all possible treatments in all situations and let the best approaches bubble to the top. There are simply too many variables for that to work. Rather, marketers must assign customers to journey stages and use those stages as inputs when selecting treatments and evaluating results. The stages themselves can be adjusted over time in the light of experience. But without a journey map as a starting point, marketers and their machines will flounder endlessly in a sea of big data.

Maybe you're unimpressed.  Maybe that cat video beckons.  Maybe you're thinking, "All you've done is change the labels.  A map of journey stages looks a lot like a campaign flow, and for that matter an old-style campaign funnel."  I understand your doubts. But there are significant differences:

- journey stages are not associated with specific messages, while campaign steps are.  In fact, the whole purpose of campaign steps is to define which messages are delivered when. So a journey stage is a significantly higher level of abstraction. Put another way, journey stages are just one factor in deciding how to treat a customer, while campaign steps are the only factor.

- journey stages are inherently random, while campaign flows and funnel stages are relentlessly linear. The goal of a campaign or funnel is push customers from one stage to the next as quickly as possible. Journey stages do track a funnel-type motion but they’re more intended to capture a set of customer needs and interests. In conformance with the general notion that customers will control their own movement, journey stages are more descriptive than directional. It’s no problem for a journey map if customers move back to an earlier stage or if they stay in one stage indefinitely.  For stages such as “satisfied customer”, that’s actually a good thing.

I do recognize that “journey stage” implies motion, which means it's probably not the best term for what I have in mind.  A more neutral term like "state" would be better.  But I’ll stick with journey for now because it will intuitively make more sense to most people.

So let’s assume, at least for sake of argument, that you’re convinced marketing systems should consider  journey stage when they’re picking the best message for a specific customer in a specific situation. Does that mean campaign flows can be replaced by any recommendation engine that adds journey stage to its list of customer attributes?

I think not. The system needs an intermediate level between broad journey stages such as “interested prospect”, “active buyer”, “satisfied customer”, and “advocate”, and actual treatments during single interactions. That level is needed because marketers have to create the content that the machines will choose from during those treatments. Marketers will decide what content to create by envisioning experiences that span multiple interactions and then creating content for a complete experience. The best analogy I’ve found is “plays” in a sport like football: tightly choreographed sets of actions that serve a narrow purpose. Teams prepare plays in advance and pick the right play for the moment but they don’t try to set the sequence of plays before the game begins. They know that games, like customers, follow their own unpredictable course and all the team can do is react appropriately in each situation. It’s true that each play is directed towards a long-term goal, but execution of the play is really a self-contained project. What’s critical is that the marketing play can incorporate multiple interactions over time; this allows a more coherent, richer customer experience than treating each interaction as wholly independent.

Good marketers and sales people already think this way, although they usually describe it in terms of tactics to handle different types of buyers, personas, or situations. I’ve also recently heard several innovative marketing system vendors describe approaches that I believe are fundamentally similar to this concept, again in different terminology. I’ll tentatively adopt the term “plays” precisely because it’s vendor-neutral and because I think most people are familiar enough with sports plays for the analogy to be helpful.

To clarify, then, I believe that marketers of the future will think in terms of broad customer journey stages (i.e., states), such as “active buyer” or “satisfied customer”. Each stage has strategic objectives, which might be to move an active buyer closer to purchase or to strengthen the relationship with a satisfied customer. Marketers will pursue those objectives through “plays” that are appropriate in specific situations, such as “active buyer requests a demonstration” or “satisfied customer has a service problem” and take into account other context such as location, device, and recent behaviors. They will create content and process flows to execute those plays.  These plays will resemble current multi-step campaigns but on a smaller scale and with narrower goals. This limited scale is exactly what makes plays so useful, because marketers can easily visualize each play as a whole. This lets them construct coherent sets of messages, rules and metrics to execute each play from start to finish, measure its impact, and make changes to optimize its effectiveness. Today’s massive nurture campaigns are too large and complicated to do the same.

Marketing systems of the future will also be designed to support this model. It may even happen that systems designed on this model come first and marketers adopt the model as they come to appreciate the systems. Or, because the stage/play model is especially well suited to automated campaign design and the general “madtech” world, perhaps it will be embedded in automated systems and grow as such systems are adopted.

Whatever the mechanism, the traditional campaign flow model has reached the end of its useful life.  I believe the stage/play model will be its successor.

______________________________________________________________________________

(1) The study actually measured the attention span of average Canadians.  It didn’t specify the nationality of the goldfish.

(2) The preceding generation of campaign managers, called MCIF systems, used proprietary columnar databases to wring adequate performance from the PC hardware available at the time. The first of these, MPI Max$ell, was introduced in 1984; other major products included OKRA Marketing (1986), Customer Insight (1987), and Harte-Hanks P/CIS (1988). Interestingly, the height of sophistication among these products was a feature called “matrix marketing”, which identified the best offer to make each customer after each (monthly) update. Sound familiar?

(3) click here for my 1994 review of MIND and a couple significant contemporaries. If you add a reference to digital media, I could be describing any of today’s cutting-edge marketing automation products:

“At the core is a very powerful campaign management function that allows a marketer to define sequences of marketing events–each including a mix of direct mail, telephone contacts and personal sales efforts–to be followed in different circumstances, and then to automatically execute these sequences.

“The system uses an efficient graphical interface to lay out the alternate sequences that can be followed within each campaign, the tasks associated with each step in each sequence, and even the specific promotional materials used with each task. As a result, the marketer gains extremely precise control over the marketing approach used with each customer–including the ability to switch the customer to a different sequence depending on actions during the campaign.”

(4) I'm perfectly aware that reality will be messier than the vision implies.  Coverage will be incomplete, people won’t always be recognized across devices, and some predictions will be wrong. But perfection isn’t necessary for success: systems using this data only have to be more effective on average than systems that don’t.