How OWOX BI Created And Tested An Attribution Model For SaaS Businesses

Moreover, businesses use a lot of different ways to acquire and retain customers, including offline channels like phone calls, conferences, and meetings. The challenge here is that you need to know which of your efforts worked out, attracting customers and making them become your regular ones.

OWOX BI faced exactly the same challenges that prompted us to create our own funnel-based attribution model for B2B SaaS companies.

In short, we collect data on all the user touchpoints on their way through the funnel. Next, we evaluate the probability of users passing on to the next funnel step. If the step is more difficult to pass, the efforts that lead users through it get more value.

But it's vital to record every phone call and a meeting in a CRM system, track and evaluate every step of the user journey so that you could get measurable results.

Similar to many other B2B Saas companies, our customers interact with us outside the website as well. We actually developed quite a complicated system for that purpose:

Due to so many teams involved in the user journey, it's quite challenging to find out whose input was the most valuable at every step of the funnel. It's also difficult to discover what stage of the customer journey requires more time and effort, and which one can be easily skipped out. If you try to evaluate every team's contribution separately, you'll surely attribute more revenue than you really get, because the same deal can be considered for each of the departments.

If your business is subscription-based, not only the first but also the following payments matter to you. So the value of the acquired customer should be measured by the first payment along with the customer LTV. Moreover, businesses normally use multiple tools to attract and retain customers, with different teams involved, and it's important to know their contribution.

That's why we decided to combine two approaches: first payment and the predictive LTV after considering possible ways to distribute value among the funnel steps. We suggested that the value of the first payment should decrease with the following months. Due to this, the customer value gets measured and then redistributed after getting every next payment, depending on how many efforts every team spends on each of the use steps.

To calculate the marketing attribution model, we use the value of the predictive LTV and deduct the value of the payments we already got. For example, if the money obtained from a customer's predictive LTV is $1,500, and he or she pays $100 each month 3 times in a row, the value of the first payment will be $1,200, which is $1500 - (3x$100). In case we talk about 6 months instead of 3, the value of the first payment will be $900, which is $1,500 - (6 x $100).

We defined all of the customer events and distributed them based on the area of engagement within the company. This is how we got 40 types of events within 5 categories:

All of the aforementioned data is collected in Google Analytics and other systems: SalesForce, Intercom, Gmail, Calendar, Zendesk. That's why we brought all data together in Google BigQuery, using OWOX BI Pipeline. To track how our audience overlaps across domains (owox.com/ru/ua and conference websites), we used OWOX User ID. We also analyzed project-level events and tracked User ID along with Project ID. It is necessary to do that, as in B2B business a product can get tested by an analyst during the trial period, though purchased by a CMO from the very same company.

Moreover, the sessions that lead a customer through more difficult steps should get more value, and the value should increase with the step complexity. The attributed revenue should match the real revenue obtained by a business.

To look at the data from a different angle, we built 3 reports.

The first one below demonstrates how actual and predictive revenue, as well as the actual and planned costs, are attributed to each of the departments. Due to the long user journey, we have at OWOX BI, it's important to take note of the predictive revenue. This allows us to measure the contribution of all the efforts that lead the customers along the funnel.

For instance, you can see that the most costs and predictive costs are attributed to the Product department, though it also drives the most of revenue both, actual and predictive.

The second report is based on the same data, but it demonstrates ROI for each of the departments.

The report above demonstrates that not all of the Product teams have positive ROI yet, like, for instance, Smart Data. This means that the team's processes should be optimized to boost the overall Product department performance.

The third one visualizes the actual and predictive value of our efforts and contains metrics from previous reports. You can see the metrics of the first report by months on the top graph, while the circle diagrams show how revenue is distributed among departments. The bar chart allows you to see the ROI by departments.

This report shows, for instance, that you should pay more attention to the Product and Sales departments and help them streamline their work.

Having evaluated the performance of every department, you can realize how efficient is each of the marketing channels and understand how to make more of the channels that pay off.