A Better Way: Incremental Pipeline Analysis

Pipeline attribution as we know it is completely broken.??

Legacy approaches like first- or last-touch attribution have not been fit for purpose for years (if ever) and multi-touch attribution models have not meaningfully improved the situation.? The near universal failure of B2B SaaS businesses to accurately measure where and how value is being added in their sales funnels has major consequences for the industry, but also poses a huge opportunity for the businesses that can meaningfully solve this problem.?

Why Measure Attribution in the First Place?

Before we discuss the current state of the problem and some possible solutions to it, it’s important that we step back and cover why it behooves a company to understand pipeline attribution in the first place.

The main purpose of measuring pipeline attribution in sales and marketing is to understand and quantify the impact of various sales and marketing activities on generating and progressing opportunities within the sales pipeline.??

This understanding is critical for several reasons:

Better Sales and Marketing Resource Allocation

By identifying which activities are most effective at originating opportunities and progressing them through the sales pipeline, organizations can allocate their resources and budgets more efficiently.?

This leads to better program accountability and a higher return on investment for sales and marketing efforts.

More Accurate Sales Forecasting and Revenue Projections

Understanding the impact of different activities on sales pipeline aids in more accurate forecasting and predictive analysis.?

Organizations that better predict future revenue are empowered to better allocate capital not just within sales and marketing, but everywhere else in the business.?

Improved Customer Journey Mapping

Accurate attribution helps in understanding how customers interact with different touchpoints on their journey to becoming a customer.?

This knowledge can improve the customer experience by optimizing these touchpoints, help drive down customer acquisition costs (CAC), and provide vital feedback loops for the product team.

The Current State of Attribution

So what are the most commonly used attribution models today and how do they fall short?

Position Based Attribution

This can include first-touch or last touch-attribution where all or the majority of the attribution credit for a lead or opportunity is based on which functional group (e.g. marketing, channel, sales, or SDR) “touched” a contact first or last respectively.

Alternatively, slightly more sophisticated position based models can include “U-shaped” or “W-shaped” attribution where more attribution weight is given to the first and last interactions, but some of the mid-journey touches are incorporated as well.

Problem

All of these models suffer from a kind of “tail wagging the dog” fallacy—in a world where it’s common for buyer journeys to frequently take an entire year or more, attributing a majority or even plurality of credit to one channel based on one or a handful of touches misses the mark.? At best it’s a deeply na?ve model of the real world and at worst it’s frequently subject to “weaponized attribution,” where the whole process is deeply political and produces zero-sum thinking and principal-agent problems that impede your business.

Multi-Touch Attribution

These models are typically composed of either a linear attribution or time-decay attribution model.

In the former, the model assigns equal credit to all touchpoints along the customer journey. The intention of this approach is to supposedly provide a more balanced view of how various interactions influence conversions.

Alternatively, time-decay modeling gives more credit to touchpoints closer to the conversion and less credit to earlier interactions, recognizing the diminishing impact of older touchpoints.

Problem

While these multi-touch approaches reflect a slightly more accurate reflection of the buyer journey than fully na?ve position-based attribution models, they’re still too simplistic to offer any real predictive power.? Furthermore, they frequently suffer from calibration issues that additionally limit their effectiveness.

Algorithmic and Custom Attribution?

Finally, we move into the most sophisticated approach typically employed: algorithmic attribution.

For B2B SaaS businesses, algorithmic attribution is almost always delivered by a vendor (e.g. Demandbase, Bizible, Terminus, 6sense, Circle CI, etc.).? These vendors typically offer a range of customization from out of the box to highly tuned models to reflect your customer journey.??

Problem

Historically, these approaches have suffered from two bottlenecks: data quality and model sophistication.

The increasing use of data lakes to collect information from both sales and marketing efforts throughout the buying journey and customer lifecycle means that for the first time these businesses have the potential ability to measure and predict the power of sales and marketing interactions in a meaningful way.

However, not all businesses can or should set up a data lake or Lakehouse for their GTM data—these can be huge, expensive implementation projects and are not suitable for all companies.

Additionally, the legacy machine-learning models that power most of the current vendors are rudimentary compared to the huge gains in LLMs that we’ve recently seen.??

The weaknesses of algorithmic attribution are not limited to B2B SaaS vendors—P&G and Uber reduced their digital ad spending by $200m and $120m respectively and saw no measurable change in business outcomes!?

The one-two punch that results from the use of garbage-in garbage-out data and simplistic algorithms, means that in practice, most of these attribution vendors are not fit for purpose as they exist today (though I fully expect this to change in the very near term with the rise of the data intelligence platform and other related technologies).?

Summary So Far

As a reminder, effective pipeline attribution enables us to:

1. Better allocate resources across sales and marketing efforts
2. Improve the accuracy of sales forecasting and revenue projections
3. Improve the understanding of the customer journey

The most frequently employed pipeline attribution approaches fail on some or all of these fronts.? They’re plagued by:

• Internal politics that seek to back rationalize budgets and headcounts, instead of seeking the truth.
• Inaccurate or missing customer journey data.
• Simplistic or na?ve models that aren’t fit for purpose.
• Complicated integrations that aren’t tuned effectively.

So while I fully expect that enterprise LLMs will eventually solve this problem, what are we to do in the meantime?

A Better Way

Let’s pose a thought experiment:?

Say you have an established sales and marketing team, with no SDR support.? You have several years of performance baselines, an understanding of seasonal fluctuations, and a growth trend line.??

Then one day you add several SDRs to your go-to-market motion.

Fast forward 6 or 9 months—taking into account seasonality, the growth trend line, and assuming no major changes in product releases or marketing efforts, you should be able to na?vely attribute any additional qualified pipeline and closed won business to the addition of these new SDRs and be fairly confident in calculating their return-on-investment.

This simple counterfactual analysis and its slightly more sophisticated cousin, differences-in-differences, can be extended to analyze all areas of your pipeline.? In fact, you can use this kind of analysis on problems as hairy as trying to calculate just how many people were killed by COVID-19 or to understand the effect of federal reserve policy on the Great Depression.

Applying modestly sophisticated analytic techniques to better understand how programs contribute to your sales pipeline can help to disarm internal politics, avoid expensive and time consuming vendor implementations, and empower you to FINALLY answer the question, “how much value is being added by this program.”

Incremental Pipeline Contribution

You may or may not have straightforward natural experiments in your go-to-market motion like this SDR example, that you can take advantage of to gauge the performance of a given program.

If you have a data scientist on staff and a large enough dataset, consider working with them to see if there are differences-in-differences you can analyze.

For everyone else, I strongly encourage you to regularly test the incrementality of your major pipeline sources and programs by following these steps:

1. Define the ObjectiveStart by clearly defining what you want to measure.?This could be the impact of a specific marketing campaign, the effect of a given SDR:AE coverage ratio, or an ABM program.
2. Identify Key MetricsDetermine which metrics you'll use to measure the impact. This could include the number of leads generated, conversion rates, average deal size, or sales velocity.
3. Segment Your DataDivide your target audience into two groups—a control group and a test group.The control group does not receive the marketing or sales intervention, while the test group does. This segmentation is crucial for comparing results.
4. Implement the TestExecute your sales or marketing program with the test group while keeping the control group isolated from these activities.
5. Measure the OutcomesAfter a predetermined period, measure the outcomes in both groups. Look at your key metrics and compare the results between the control and test groups.The difference in performance between the test and control groups indicates the incremental impact of your sales or marketing program.?For example, if the test group shows a 20% higher conversion rate or pipeline build than the control group, this increment can be attributed to your intervention.
6. Control for External FactorsEnsure that any external factors (like market trends, seasonal variations, rep turnover, etc.) are accounted for in your analysis, as they can skew the results.

Running this kind of experiment isn’t as easy as setting up a Salesforce dashboard and hitting refresh before a weekly pipeline meeting.? It is, however, unmatched by almost all other existing pipeline contribution approaches or vendors in its ability to tell you what kind of value your various go-to-market strategies and functional teams are adding to your business.

Be sure to run this kind of testing frequently enough that you can make sure your operating assumptions are up to date and so that you can measure changes in time.

In Closing

I once had a CEO tell me, “I assure you, there is no other way to measure pipeline attribution than either first-touch or last-touch.”

Let me tell you: there are other ways and not all of them are equal.

Whether you’re investing $500,000 or $500,000,000 annually in your sales and marketing programs, you owe it to yourself and your company to better understand what you’re actually getting for that investment.