Statistical Attribution & Optimization in the B2B World.
There has been a lot of activity recently around revenue attribution - marketers want to develop a better understanding of their customer acquisition funnel and be able to measure progress against it. Most of this attention has been focused on the B2C space. However, less work has been done measuring the performance of B2B marketing activities.
Certainly the marketing automation segment is very vibrant with a large number of vendors (both big and small) providing solutions that better enable businesses to manage campaign workflows, nurture prospects through the purchasing funnel, and report metrics around leads and campaigns, However, not as much has been done in marketing attribution and measuring the effectiveness of marketing programs
While Salesforce is an excellent platform for managing leads and campaigns, their business model is founded on developing a sales and marketing ecosystem comprising partnerships with specialist vendors that can provide more focused solutions to specific sales and marketing issues.
As a result, companies such as Full Circle Insights, Bright Funnel and Bizable have emerged to fill the void in B2B marketing attribution by leveraging the Salesforce platform. They effectively bridge the gap between marketing automation and CRM solutions to provide digital marketing campaign measurement solutions. These companies have impressive lists of B2B clients and are well supported by venture capital investors who clearly see the opportunity in this space.
So what do these companies actually do? In short, they take sales and prospect data from Salesforce and marketing data from a marketing automation platform and apply pre-packaged attribution models such as single touch, fractional or weighted attribution to determine how much a specific campaign contributed to revenue (or some other desired outcome).
While we applaud these steps forward in B2B attribution, we can definitely do better. Our insights are driven by years of consulting to the Salesforce.com marketing operations function and other clients on statistical marketing attribution and marketing mix optimization
We think the next logical steps on the path are to move from the default, simple and judgmental attribution methods to statistical attribution methods and then optimization to deliver prescriptive reallocation advice based on the learnings from attribution.
We have developed our SaaS based SpendMetrix (SMX®) system to be used in a complimentary fashion with Full Circle and other B2B campaign attribution systems. Basically Salesforce and Full Circle provide the data to SMX® to estimate and update statistical attribution coefficients. Then these coefficients are loaded back into Full Circle through their custom model interface. SMX® can also provide optimization results to inform better reallocation of marketing campaign resources.
We think the winners in the B2B attribution and marketing measurement space will move through a series of stages:
from judgmental /naïve attribution to statistical attribution,
from statistical attribution to optimization, and
from optimization to simplification and automated test design.
Judgmental and naïve attribution models are a good place to start when moving down the attribution and measurement path. You will quickly learn how stakeholders on the marketing and sales force side want to see attribution results presented to them, and how they will use attribution to improve decision-making. But a few quarters into this process you may also see things begin to fray as the attribution results themselves are seen to lack some credibility.
For example an even credit attribution system will bias results and inflate the contribution and ROI of campaigns with many (possibly minor) touches with possibly low individual purchasing influence. Letting this discontent brew for a little while among stakeholders is not a bad thing, as long as you have your next plank in the system (statistical attribution) ready to move forward.
Statistical attribution draws on variants of machine learning (e.g. CRT) and regression to estimate coefficients that score individual campaign touches and responses with the partial credit they deserve in moving leads to opportunities, and opportunities to sales. A good statistical attribution system will capture a number of important factors:
Some leads are better “at birth” than others, as captured by lead characteristics like industry segment, size, region, and source.
Both the characteristics of the lead company (e.g. Industry, size, and region) and individual contact characteristics (e.g. Title, function) must be quantified.
The time trajectory of a lead’s evolution matters, time lags between responses, and lead age are important signals of influence and the likelihood of imminent conversion.
Some activities and responses (e.g. CXO conference attendance) clearly have more influence on moving leads and opportunities along than others. And their relative importance can vary dramatically by region or segment.
In a statistical attribution model all of the above characteristics will be incorporated into what can be a complex multi-dimensional score for each touch or response on the path to lead conversion and sale.
With statistical attribution methods we distinguish between updating the attribution point system (coefficient updates) and using the coefficients to “score” the credit that each campaign element receives (attribution scoring). Statistical attribution models can and should be updated with newer data over time, but our experience is the shelf life of coefficients is measured in months not days. So real time coefficient updates are not so important but frequent updates of attribution scoring with the current coefficients are very important.
Happily, statistical attribution models can be estimated routinely offline using Salesforce data, and then the coefficients can be uploaded to attribution systems like Full Circle using their very flexible custom model interface.
When statistical attribution models are applied, marketing and sales stakeholders can look at tables and visualizations that capture the impact of individual campaigns and if desired, marketing activities, and individual touches and responses. If campaign costs are known and captured in Salesforce, then campaign ROI can be readily computed.
The review of individual campaign ROI’s by stakeholders will naturally lead them to want to answer more general questions like:
What types of campaigns do better than others, which should we scale up and or down?.
What targeted contact roles and functions produce better ROI results?
What campaign sources do best?.
What regions do better?.
How can we get more bang from our marketing buck?
Ultimately stakeholders want to know how to reallocate marketing and sales resources to improve ROI. When the campaign velocity and variety is low, eyeballing campaign level ROI results can make these generalizations. But marketing organizations with high campaign velocity and variety will quickly exceed the ability of marketing campaign managers to use “eyeball analytics” to get the most out of statistical campaign measurement.
The next logical step in the process is to move to some form of constrained optimization to extract learnings from statistical attribution and automatically suggest budget reallocations over types of campaigns. The marketing organizations that routinely use campaign hierarchies to capture the types and facets of campaigns in a systematic way will be the ones best able to benefit from optimization insights.
The suggested campaign budget re-allocations from SMX®’s optimizer can routinely find ROI gains of 40% plus in the B2B space through informed reasonable re-allocations.
Simplification and Testing
One of the criticisms of marketing mix optimization systems is that they often produce very complex prescriptions on reallocations, which are hard to implement due to their complexity. So very often the ROI gains from such systems tend to be aspirational.
Another complaint is that to realize the results, and realize the gains marketing organizations must be willing to make some big risky bets. What is needed is a way to simplify optimization recommendations and tee up simplified small-scale tests so that stakeholders can wade into the brave new world of attribution and optimization.
Our SMX® attribution and optimization system provides exactly that ability by allowing smaller, subsets of the full optimization recommendation to be implemented as small-scale tests. This gives marketing groups to act sooner on results, and work around complexities and cross group politics, that might come from more complex and sweeping re-allocations that can come out of optimization.
If you are a user of a B2B campaign attribution system like Full Circle, Bright Funnel or Bizable and would like to learn more about how to move along this path from naive attribution to statistical attribution, optimization and testing, please reach out to me.