Wednesday, May 20, 2009

Prediction: Statistical Methods Will Replace Conventional Rules for Marketing Decisions

Summary: basic demand generation features are close to a commodity. Vendors who replace conventional decision rules with automated statistical methods may gain a key competitive advantage because the automated methods produce substantially and measurably better results.

One of the most popular posts ever on this blog is Low Cost Systems for Demand Generation, which listed several options that started at under $500 per month. But it seems that nearly every day brings yet another possibility to my attention. Some really frugal alternatives include Genoo starting at $199 per month; Net-Results starting at $79 per month; and Nurture starting at $495 per month. I haven’t looked closely at any of these but they all seem to promise the core demand generation capabilities of email, landing pages, automated nurturing, lead scoring, and sales system integration.

The question this raises in my mind is where the industry goes from here. Basic demand generation is on the verge of becoming a commodity if it isn’t one already. The more sophisticated vendors will of course continue to add features, but it’s not clear that most marketers will be interested in the additional capabilities or be able to handle the added complexity. Perhaps the key competitive battleground is the ability to add that complexity without making the systems harder to use. But even though there are certainly substantial differences in usability among today’s systems, it’s hard to see why everyone won’t eventually be able to do roughly equal jobs of simplifying their interfaces.

Another possibility is that vendors will compete on their ability to help marketers use their systems – that is, by providing marketing training, usage reviews, and professional services. In other marketing automation segments, including MCIF systems and campaign management for consumer marketers, the ability to provide such services was the single most important difference between winners and losers. The same applies to CRM systems – it was Siebel’s partnerships with big system integrators that ultimately let it pull away from the pack. I do think these services will be a key success factor in the demand generation market, but there’s a big difference: because demand generation systems are offered as on-demand services rather than on-premise software, the actual deployment is much simpler. This means independent consulting firms can more easily learn to work with multiple systems. Because it’s much harder for vendors to build a loyal, locked-in base of resellers, it’s easier for new players to duplicate the service infrastructure of established competitors.

This brings us back to features. Certainly there is a list of hot items right now: Webinar integration, digital asset management, dedicated IP addresses for outbound email, APIs to post data from external forms, integration with Google Adwords, providing contact names from external databases when a visiting company is recognized by its IP address, pulling data from social networks to flesh out a prospect’s profile, and interacting through social media in addition to traditional channels.

The question is which of these features will turn out to be really essential. The only one I personally see as important to a large number of marketers in the immediate future is the Webinar integration, because Webinars are widely popular and integration makes the marketer’s life significantly easier. Everything else on that list strikes me as either of interest to a relatively small fraction of marketers or as simple enough to add that it won’t be a competitive advantage.

So is there something else that could be really important? Well, I wouldn’t ask the question if I weren’t leading up to something.

My particular insight, if it is one, is that consensus has crystallized within the past month that marketing now remains dominant much deeper into the buying cycle, and that sales and marketing must work much more closely together as a result. The idea itself isn’t new, but I suddenly see it referenced everywhere I turn. Part of the reason may be that I’m paying more attention because I wrote a paper on the topic myself (see When Best Practices Go Bad: New Rules for Sales and Marketing Management) although I’m under no illusion that my paper was anything other than one voice among many. It’s simply one of those ideas whose time has come.

As I and others have written, the immediate implication of this change is that marketing systems should provide salespeople with more information about prospect behaviors – what Steve Woods of Eloqua elegantly calls “digital body language”. This gives the salespeople insights into customer interests, replacing to some extent the information that they previously gathered for themselves when dealing with prospects directly.

But those direct interactions also built a relationship between the salesperson and the prospect. Watching their behaviors doesn’t do that. To the extent that anything does build the early relationship today, it’s the automated nurturing programs and behavior-driven responses executed by marketing systems. I don’t really believe that even the cleverest marketing systems can really replace the trust built by a good salesperson, but at least the automated programs can educate prospects and leave a positive impression about the company’s responsiveness to their needs.

I haven’t seen much written about the burden that this change places on the marketing systems. We’re not talking about some simple drip marketing to keep leads warm and educate them a bit until they move closer to their purchase. Rather, marketing must come as close as possible to simulating the interactions between a prospect and a good salesperson to build an essential relationship. This means that the marketing system has to be really smart. And I think providing this sort of intelligence might be a major competitive battleground for the vendors.

That last sentence was a bit of a leap, so let me fill in the blanks. Today’s demand generation systems are largely rule-driven when it comes to selecting prospect treatments. Whether those rules are embedded in list definitions, campaign flows or dynamic content doesn’t matter. The problem is that rules are hard to build and remain unchanged until somebody writes a new one. They’re generally based on somebody’s best guess about how the world works and they tend to be fairly simple. As a result, rule-driven systems just can’t be very smart, in the sense of reacting appropriately to subtle clues or changes in behaviors.

The limits of rule-driven systems don’t matter when there isn’t much data to work with and there aren’t many choices to make. That was arguably the case in the past when lead management systems worked with only a small amount of data from a postal reply card or brief telephone survey. But today’s demand generation systems are dealing a flood of behavioral data related to emails and Web visits. Rules can’t deal optimally with that much information. In addition, the demand generation systems have many more decisions to make, since every personalized email and Web page involves many choices for information to display. No one can create enough rules to handle all the possibilities.

Nor is the challenge limited to rules for selecting messages. Demand generation systems also use rules to decide when to alert salespeople about prospect behaviors. Lead scoring formulas are essentially rules as well. In addition to the fact that these rules are all defined manually and pretty much arbitrarily (that is, based on users’ best judgments), there is little feedback to check whether they are effective.

All of this absolutely guarantees that demand generation systems will produce suboptimal results. That would be annoying under any circumstances, but if the demand generation system takes on the primary responsibility for early relationship building, it’s more than merely annoying. It could destroy your company.

There is an alternative. Marketing systems can deploy automated statistical techniques to select messages, issue alerts and send leads to sales. Consumer marketers have used such methods for years with proven success. In addition to dealing with many more options than rules can handle, such systems can automatically learn from past results to improve their accuracy and adjust to changes in behaviors. Nicer still, marketers and salespeople can actually observe the success or failure of the decisions by watching objective criteria such as return visits and close rates. This last point is critical because it means marketers have a way to actually compare the value of decisions made by different systems. This means that vendors can meaningfully compete to offer the best decision-making capabilities, and marketers can choose the system that does a better job. And, unlike a feature that appeals to just a small fraction of marketers, better decisions are important to everyone. A system that could show it made better decisions would therefore have a very major competitive advantage.

So far, everything I’ve written here is just my private little theory. I haven’t heard any vendor, pundit or client suggest anything similar. This could well mean that I’m wrong; after all, I do like fancy automated systems with their cool bells and whistles. But I think maybe I’m right. Demand generation systems are getting more and more complicated, and something is needed to radically simply them before they collapse into chaos. Given that the stakes are nothing less than the sales process itself, allowing this to happen is unthinkable.

3 comments:

mkamrk said...

David:

This is a great post. I think you have touch of a lot of important aspects that were the foundational elements of our thought process behind Nurture, which is to build the simplest lead nurturing application for marketers. Email is the killer application on the web, and we want to aim toward that level of simplicity with Nurture. Another important focus for us has also been delivering automated and realtime actionable lead data to sales teams without requiring any manual effort on marketing's behalf which can lead to instant feedback and rapid improvement in campaign design.

Mike Pilcher said...

David it is almost like you have seen our design docs. We could not be in more vehement agreement.

Our concept is as you move items around on our campaign flow builder each new element would have a predictive number of results and responses. It would look to the segment size, look to how many people would be likely to respond at that stage. It looks back to prior response rates in the flow and what you get is something showing the perfect bell curve where one email was too few and one hundred emails and a dozen phone calls was too many as response rates drop off.

Initially you set it against best practices results. Over time it becomes more compelling as you build in and compare to actuals for the campaigns and then incorporate these into the model for the next one so a marketer delivered their own best practices.

So you see where over-nurturing delivers diminishing returns - as the application learns from prior campaigns it delivers self-learning.

David Raab said...

Mike, that sounds very cool. However, I think we're talking about slightly different things. If I understand correctly, you are planning automated analysis to find the optimal structure for each campaign. I had in mind selecting the optimal treatment for each individual during each interaction. This is similar to behavioral targeting for Web ads.

Still, we are generally looking in the same direction and I wouldn't claim to know precisely what approach will ultimately win out. The point, on which we both agree, is that companies have to use objective analysis to design and optimize their campaings and tailor those campaigns to the prospect's situation.