What is a funnel-based attribution model

Funnel Driven Allocation for SaaS B2B Companies - How We Viewed the Value of All Marketing Efforts

I want to share how we solved the marketing effectiveness assessment problem and as a result developed our own attribution model.

The mapping indicates how conversions and the value of those conversions are distributed across different traffic sources. For example, by default, Google Analytics uses the last indirect click assignment model. That is, 100% of the value is assigned to the last channel in the interaction chain that was before the direct transition. The downside to this model is that we don't know how important the role of other channels is. If we refuse, how likely is the purchase?

Our company is a typical SaaS representative for B2B. That is, trial version, monthly subscription, multiple plans. In such companies, the funnel the user walks through is non-linear, with many steps that cannot be followed on the site. For example, holding a demo, personal letters, calls, meetings, conferences. Some of the tasks on this list are closed by the sales department, others by marketing or customer support, and it is important to evaluate the contribution of each task. We have identified five main areas of responsibility:

  • marketing
  • activity
  • sales
  • Customer care (customer success and support)
  • Product development

It was necessary to understand how much profit each department brings, what it is worth investing more time and money for, and what to give up. To do this, you need to evaluate the effectiveness of the efforts made. Efforts that we call absolutely any income generating activity: placing contextual advertising, conducting webinars, meeting with clients, etc.

What is attributable?

In B2C projects, the answer to this question is obvious. If the deal is a transaction, the revenue from each transaction is allocated. If not transactional, the number of conversions.
When deciding what to map, we looked at several options:

First payment

This option has been abandoned as it is not suitable for a subscription service for which it is important to consider not only the first but also all subsequent payments. In addition, efforts to attract customers (initial payments) and retain customers (repeat payments) are the responsibilities of different teams that we would like to evaluate separately.

LTV or Projected LTV

These options are better, but the LTV is only reliably known when the customer no longer uses the service. And the forecast LTV is an unstable and not entirely accurate indicator. It depends heavily on the size of the company and the tariff plan chosen by the customer. So we chose the third option.

Projected LTV minus all payments received

We take the projected LTV and subtract the values ​​of the payments already received. For example, if a customer's projected LTV is $ 1000, he pays us $ 100 a month and has already paid three times, then the value of the first payment is $ 700, which is $ 700. H. $ 1000 - (3x $ 100). After 6 months, the first payment is already allocated $ 400 for the same customer, i.e. H. $ 1000 - (6x $ 100).

To whom and how should this be attributed?

An important test for any attribution model: That total allocated income must equal the income the company actually received.

For example, assigned conversions or the simultaneous use of attribution models for the first and last click have the following disadvantage: The same income is assigned multiple times. Therefore, before the valuation, it was important for us to pool all of the company's efforts and assign them real income.

It would be wrong to create separate models that distribute income between sales, then marketing, and so on. Because the head of the sales department can distribute the income from the attracted businesses to his employees and the head of marketing to his own. As a result, all good companions are only credited back to being more than deserved.

I had to classify efforts within the areas of responsibility and for each of the areas to highlight all events in the client's life. As a result, there were around forty events in five categories - here are a few examples:

marketing: SMM, content marketing, webinars, paid advertising, email marketing, PR
activities: Buy tickets, visit the conference website
sale: Demo, personal letters, calls, meetings, chats
product: Trial, Freemium, the rest of our products
Customer support: personal letters, responses to customer inquiries, chats, meetings

This is what the funnel fragment looks like in our attribution model:

From a technical point of view, the data fusion is the most difficult phase, as the events we highlight are recorded not only in Google Analytics, but in a variety of systems: SalesForce, Intercom, Gmail, Calendar, Conference Pages.

We solved this question as follows:

  1. We have already collected data from all of these systems in a single Google BigQuery repository.
  2. We have a cross-location cookie that can be used to track target audience overlap between different websites - product pages, marketing pages, conference pages - even if there has not been a direct transition from one website to another.
  3. It is possible to analyze events at project level (account-based or user-based). In addition to the user ID, we track the project ID and record successive chains of events within the project. Because in B2B business it often happens that one user moves along the funnel, another connects and a third comes in and pays.

For the logic of value distribution, we used our own funnel-based attribution model. It goes well with event chains. In short, we collect all possible options for the passage of the funnel. Then we consider the probabilities of all possible transitions between events (steps of the funnel) and distribute the value according to the principle: the harder it is to pass a step, the more valuable the effort that the user has made.

What is the bottom line

As a result, we have answers to the question "How does the effort affect the current result?" Formulated. in the form of several tables and a dashboard.

In order to calculate the ROI of all directions except value, we needed costs. We have included salaries and additional expenses in expenses. For example, marketing paid advertising while a product had a technical infrastructure.

We also realized the need to take into account the forecast income and expenses. The fact is that the sales cycle in the B2B area is very long - in many cases it takes months from the first contact to the changeover. So when you look at the value of the past few months' efforts, you might get the impression that things are bad. This impression is misleading as a large part of this effort contributed to conversions that have not yet happened.

Here are the reports we got in the end. All the numbers in the screenshots are for example purposes only and are not real.

1. Income and expenses according to areas of responsibility and months:

This pivot table shows how expenses, planned expenses, income and projected income are distributed by area of ​​responsibility and month. It is important for us to consider projected earnings as the sales cycle is long. Without this revenue, the contribution of trying to get users to the top of the funnel would always be grossly underestimated.

This table will help you compare projected income and actual income. The higher the effort along the funnel, the more predictable the income.

2. ROI according to areas of responsibility:

In another pivot table based on the same data, in addition to total income and expenses, we see the ROI of each department. The ROI shows the profitability of each direction by their tab across the entire sales funnel. In our example, the sales force's ROI was negative. With this in mind, it is possible to rethink investing in this department.

These reports help us see how much we will make with some effort in marketing, sales, etc. This way you can simulate results depending on your budget. However, there are some limitations, for example we cannot spend less on a product.

3. Actual and Predicted Value of Efforts:

This dashboard gathers information from the previous tables. This shows all the important key figures: income, forecast income, expenses, planned expenses and ROI. The graph above shows the same metrics (excluding ROI) by month. The pie charts show the distribution of income by area of ​​responsibility. The bar chart gives you an idea of ​​how the ROI is distributed across the areas of responsibility.

Please note that in each subsequent month the proportion of the forecast income (light green in the graphic above) is greater. However, over time the data will be updated and the actual income will have a larger share.


  1. We have an attribution model that measures the effectiveness of all business efforts: marketing, sales, product, events, and customer success. This is important because if you evaluate each direction individually (just marketing or just sales), more revenue will be allocated than we actually get. At the same time, one model allows you to see both the general picture and the details for each department.
  2. We have made sure that only what can be measured can be evaluated. For example, early versions of our model underestimated some unrecognized sales efforts. As a result, we have digitized most of the activities of this department: all letters to clients, meetings and calls.
  3. We have found that due to the long sales cycle and monthly billing, it is important for us to consider projected earnings. Because the current actions take a while to produce results. We go to great lengths to convert to the first payment, but then customers stay with us for a long time.
  4. After evaluating the departments' performance, it is important to consider channel capacity to understand where to and where not to continue investing money and time. If any of the directions turns out to be profitable, then one has to ask the question, how much does it scale, that is, how much the capacity of that channel is now depleted. In our case, just like in the classic evaluation of marketing efforts, there is a point after which an increase in consumption per channel no longer generates additional income and only leads to an increase in the costs of this channel.