RunninginSkiBoots.jpgImagine learning how to run, but in ski boots. You can work hard and get better at it, but it may get so exhausting that you skip the marathon and go watch TV.

That’s what happens frequently in modern day organizations -- we decide on building a strategy around real time marketing, but most of the time it just takes too much effort to manage proper adoption and the results don’t seem as revolutionary as we anticipated.

A big part of the solution is to realize that we don’t need a protein bar -- we need to ditch the ski boots.

Companies are feeling the pressure to constantly evolve as their customers’ problems and expectations change. In the data- and process-driven world we work in, real time conversations with our customers through sophisticated software solutions set the golden standard.

But there's a problem. There are some real flaws in how we abstract and design this golden standard, and this creates gaps in the tools and processes we use to execute. The good news? These problems are fundamental, which means that fixing them will also have a very fundamental effect on results. 

To bring things into context, while the model for real time marketing is much better and more sophisticated than it was a year ago, or five years ago, it still has flaws. A lot of the decisions we make are still based on assumptions and opinions -- and we all know they are mighty difficult to argue about. Former CEO of Netscape, Jim Barksdale said it best: "If we have data, let’s look at data. If all we have are opinions, let’s go with mine."

Here are few of the key problems with some of these assumptions:

Assumption #1: The Funnel is Linear

You acquire customers by guiding their journey through a funnel, where you lose a few folks along the way, but some convert and move forward in a linear fashion.

linear funnel.png

The Problem: A linear system can never work in a finite space. Let’s assume you have a really high conversion rate -- say 10 percent -- of anyone who ever visits your website. That means you lose 90 percent. Let’s say you get 1000 visitors a day, and you lose 900. Now unless your consumer base is growing at rate of more than 10 times, you are eating off your market segment fast.

The good part of this picture is that it’s not what happens in the real world of business -- folks do come back. There’s churn, there’s retention, there is word-of mouth, visitors use different devices, which makes them harder to track.

So if not a funnel, what exactly do we need as a figure? A cycle, a bow-tie perhaps? Only data science has a meaningful way of extracting structure from data points, but we may have to accept that the funnel is fundamentally not the best way to describe our customer’s journey.

The solution is to build a new customer journey model that takes all these considerations into account and is impacted by the questions you want answers to, without necessarily carrying over old-school concepts.

Assumption #2: Lead Scoring is the Best Way to Improve Velocity

This one is particularly fun to challenge. The theory behind lead scoring is that one way to increase the velocity of the funnel is to find that sweet spot, the moment when your prospects are likely to convert to the next step and before they go cold. Then you need to provide them with something shiny, something relevant or start a conversation. But in order to decide what this moment is and what will be relevant to them, we establish a framework of numbers -- some calculation of that lead score.

lead score funnel.pngThere's a lot of statistical theory behind lead scoring, but the current established technique for it has major flaws. Chances are a bunch of people in the company argue for a few months about what gets scored and how, and then we assign those numbers to activities and visitors start accumulating some points, or gold stars or special dots (the measurement unit doesn't seem to matter). Then you decide on a number of points, stars or dots that deem the lead qualified, send some more emails and then it’s sales’ problem. This has the benefit of being both extremely inaccurate and subjective, as well as very demanding to maintain, because it doesn’t account for the dynamic nature of the dependencies that it is trying to measure.

The Problem: Can you imagine Google employees deciding scores for all web pages and all users on the web to determine what ad will likely be clicked? There are much better solutions to this and they actually stem from being a little bit better at the math behind all this -- machine learning has come up with elegant and fast solutions to build these models and dynamically drive them based on data rather than gut feeling.

Assumption #3: Conversions Can be Optimized in Smaller Chunks

People have different ways to optimize conversions. But the industry has taken a bit of a divide-and-conquer approach. We look at little windows of interaction. Multichannel outreach to landing page, to lead, then to sales opportunity, then to customer, then renewal. We typically not only look at this as linear, but we calculate some base conversion rates in each window of time, try to optimize them and leave it at that.

optimal funnel.png

You can do an A/B test to see what banner will drive more clicks, but what ultimately matters is what will drive the most purchases, or better yet, create the most happy customers.

The Problem: We assume that all of these outcomes are independent and to find the best possible connection between A and Z means finding the best one from A to B, then from B to C and so on, but that’s often not true. Imagine that the losing page in an A/B test results in twice as many purchases. Would that still make it a losing page in this test because it got fewer clicks?

A good solution exists. Only long-sighted experiments can derive insights that can drive process in the most effective way. Similarly if you are building personalized experiences to improve conversions and providing your users with serendipitous content, then this serendipitous content should be micro-targeted and should also look at the customer experience from A to Z. Google gets paid per click, but for most part none of us do, so we need to design engines smart enough to strategize, experiment and personalize based on a goals that go further than a click.

There are also many more that come to mind, and undeniably even many more that you have encountered yourselves. These are all questions that can be solved, but are yet to be solved well. But the best thing that we can do is start the conversation, better define and explain our problems and start asking these questions. Trends and technology will catch up with meaningful and specific answers to them.

Title image by Galyna Andrushko (Shutterstock)