Building a predictive attribution model for an 18-month B2B sales cycle to help leadership evaluate marketing performance before revenue closed.
INetU Managed Hosting operated in a complex B2B environment where leads could take up to 18 months to close, making it difficult to know which marketing channels were actually creating future revenue.
Create a predictive lead scoring and attribution model that translated early sales intelligence into projected pipeline value, giving Marketing and leadership a faster way to evaluate campaigns, allocate budget, and manage CAC/LTV targets.
The model gave the company earlier visibility into expected campaign value, improved decision-making around paid media and trade show spend, and helped align marketing investment with a target CAC of less than 10% of customer lifetime value.
INetU Managed Hosting provided enterprise-level web hosting and infrastructure solutions to customers with complex technical requirements, high trust needs, and long buying cycles.
The sales process was not transactional. Prospects often needed extensive consultation before committing. They had to evaluate technical fit, reliability, security, migration risk, service expectations, and long-term partnership value. As a result, leads could take up to 18 months to become customers.
That created a serious measurement problem for Marketing.
Campaigns could generate leads immediately, but actual revenue might not appear for more than a year. By the time closed-won revenue was available, the original campaign decision was often too old to optimize. Paid search budgets had already been spent. Trade show decisions had already been made. Future event contracts might already be due for renewal.
Traditional attribution was technically possible, but operationally too slow to guide timely decisions.
The company needed a way to evaluate marketing performance before the sales cycle fully played out.
The core issue was not a lack of data. It was a lack of actionable timing.
Marketing and leadership needed to know which investments were likely to generate profitable customers, but the available performance signals were either too early to be trusted or too late to be useful.
Several structural problems were working against effective attribution:
This meant marketing performance could easily be misread.
A campaign with high lead volume might look successful while producing low-value or poor-fit opportunities. A campaign with fewer leads might appear weak while producing stronger prospects with higher projected value. Without a predictive framework, both Marketing and leadership risked making decisions based on incomplete signals.
In practical terms, the question became:
The strategic insight was that early sales conversations contained useful predictive data.
Sales representatives could often tell, after an initial discovery conversation, whether a lead was a poor fit, a possible fit, or a strong opportunity. They could also estimate the likely monthly contract value if the opportunity closed.
That information was not perfect, but it was valuable. More importantly, it was available early.
The opportunity was to convert that early sales judgment into a standardized data model that Marketing could use for attribution, forecasting, and budget optimization.
The strategy was built around three principles:
The goal was not to create a theoretical analytics model. The goal was to build a practical decision-making system that helped Marketing evaluate spend while there was still time to adjust.
The first step was creating a shared framework between Sales and Marketing.
I partnered with the Sales Manager to define a simple lead classification system that sales representatives could apply during or shortly after initial discovery. The model needed to be easy enough for Sales to use consistently and structured enough for Marketing to analyze.
We introduced three qualification tiers:
The simplicity was intentional. A more complex system might have looked more sophisticated, but it would have been harder to use consistently. The value of the model depended on adoption, so the framework had to fit naturally into the sales process.
This alignment also helped create a shared language. Marketing could no longer evaluate success only by lead count, and Sales had a clearer way to communicate lead quality back into campaign analysis.
Lead classification alone was not enough. A hot lead with low contract value and a warm lead with high contract value could represent very different business opportunities.
To account for this, sales representatives were also asked to enter an estimated monthly contract value based on the initial conversation.
This created two early-stage inputs: likelihood to close and projected monthly revenue.
Together, these inputs made it possible to estimate expected value before the deal closed.
The model did not require Sales to predict the future perfectly. It required Sales to capture informed early judgment in a consistent way. Over time, that created a dataset that could be compared against actual outcomes and used to improve marketing decisions.
Using historical CRM data, I developed a predictive attribution model that assigned weighted probabilities to each lead category.
The model combined:
This produced a predicted revenue value for each lead and allowed Marketing to aggregate expected value by campaign, channel, or event.
Instead of waiting for closed-won revenue, Marketing could evaluate the expected value of campaign activity much earlier in the funnel.
For example, a trade show could be evaluated not only by badge scans or raw lead count, but by the projected value of the leads generated. Paid search campaigns could be assessed not only by cost per lead, but by expected revenue contribution and CAC/LTV efficiency.
This shifted attribution from a backward-looking reporting exercise to a forward-looking planning tool.
The company could compare channels based on expected value rather than waiting for final revenue. This was especially important for trade shows, where rebooking decisions sometimes had to be made well before closed revenue would be available.
If a trade show generated fewer leads but produced several high-value hot opportunities, the model could show that value early enough to support rebooking during discounted periods. If a paid media campaign produced lead volume without quality, the model could reveal that weakness before months of additional spend accumulated.
The model supported decisions around:
The key value was timing. The model gave the team useful directional insight while decisions were still active.
The predictive lead scoring and attribution model changed how marketing performance was evaluated.
Results included:
Most importantly, the model reduced the decision lag created by the 18-month sales cycle.
Before the model, Marketing had to wait months or years to fully understand whether a campaign had produced valuable customers. After the model, the team could begin evaluating campaign quality within weeks of launch.
That did not eliminate uncertainty. It made uncertainty manageable.
The business could now make faster, more informed decisions using the best available early signals, rather than relying only on delayed revenue outcomes.
This case was not just about attribution. It was about building a decision system for a business where standard marketing reporting was too slow to be useful.
The work helped INetU:
For long-cycle B2B companies, attribution cannot only answer, “What closed?” By the time that answer is available, the budget decision may already be over.
The more valuable question is:
This model helped answer that question early enough to matter.
This case demonstrates my ability to solve ambiguous marketing measurement problems, build practical attribution systems, collaborate across Sales and Marketing, and turn incomplete early-stage data into actionable business intelligence.
It also reflects one of the strongest patterns in my work: I do not treat marketing analytics as reporting for its own sake. I use analytics to improve decisions. When standard metrics are too slow, too shallow, or too disconnected from business value, I look for better signals and build systems that help teams act sooner with more confidence.