Forecasting is hard.
Everyone has an opinion about the best forecasting methodology.
Historical time-series forecasting looks back at historical indicators as predictors of future events
Stage-based forecasting attempts to assign a probability of a close to the stage of a current deal
Velocity forecasting tries to leverage the deal’s age as a driver to determine how close to a close you are
Rep-driven forecasting uses a “boots on the ground” methodology where we assume reps know everything going on in the deal
The list goes on and on…
At SourceDay, we struggled to adopt an accurate methodology for forecasting the business. Until 2021, we primarily ran a rep-driven forecast (with manager override). We leveraged Aviso’s predictive analytics tool for a couple of years but found the insights too shallow, drastically shifting the “AI call” every week.
After missing our number too often due to “black box analytics,” we decided to take back our forecasting methodology.
Phase 1: Re(Instrumenting) our Sales Funnel
From 2019 to 2022, we trained our sales team using a combination of Challenger, SPIN, and MEDDIC programs. When we decided to regain control of our forecasting from the black box analytics tools, we knew the makeup of “the perfect deal” and the various attributes that won those deals. What we lacked was discipline to ensure all of those attributes were well incorporated into all of our deals.
So, our first step was to update the CRM to track and monitor each critical attribute.
We created a custom “Milestone” concept, where reps were prompted to add color to conversations during a critical meeting. Each milestone asked our field teams different questions. Because we required this extra level of detail we (RevOps) knew if the solutions engineers gave a custom-tailored demo to a customer, but their IT team didn’t attend. We knew if the team presented an ROI, but no financial stakeholder was involved yet.
These indicators highlighted risk, not just with gut instinct from the reps, but with correlatable data. Deals with X had an outcome of Y. Deals without X had an outcome of Z.
These indicators manifested into what we called our “BINGO Card.” Either you had all the squares filled in and won, or you were missing something you still needed. The BINGO card is a story for another time, but know that we visualized the health of our deals to support our forecasting.
Phase 2: Gating
After we instrumented our deals, we revisited our gating strategy. Every stage of a deal had specific milestones that needed to be completed. Before reps could exit a stage, they needed to have everything in the green on their BINGO card for that deal.
Obviously, we needed to exercise some judgment regarding these gates, as not every deal always met 100% of the requirements. Although this was not often the case, it needed to be defensible when it did.
There was no excuse if you came to a forecast call with your BINGO square for executive champion blank and were working a deal for a $500M manufacturer. But if you were working a deal for a $20M manufacturer and were only threaded into three individuals instead of five, sales managers were more forgiving.
Our gates also aligned with our expectations around how many BINGO squares could be completed by specific stages. Pricing shown in Stage 1 made no sense when reps wouldn’t scope SOWs until Stage 3. Reps were encouraged to complete squares early and often but were not required to do so until they wanted to move to certain next steps.
This is all relatively normal for B2B SaaS sales motions. But timing the requirements on the BINGO cards with stages helped us better understand if and when we would encounter headwinds and risks.
Phase 3: Triangulating the Forecast
Once everything was tracked and moving through our funnel in the same way, the final step was to develop a method for triangulating our forecasted number.
By default, we took a basic rep roll-up approach. Reps would provide us with their best case, worst case, and most likely case based on their open pipeline. Managers would then take the same pass across all deals and do the same for their roll-up. This would provide us with all of the “gut” numbers.
From here, RevOps took over.
RevOps would provide numbers from two more models each week: the weighted forecast and the time-weighted forecast.
Our weighted forecast also looked at the current pipeline by best case, worst case, and more likely case. But instead of using the gut call, we looked at all of the BINGO squares across deals and, based on their completeness, gave a read on every deal. If you were in the negotiation stage but missing the required conversations with the finance leadership, things didn’t look as good for you. We weighed every attribute against the stage of the funnel.
The time-weighted forecast provided the same level of granularity on the call with one more layer of complexity. The time-weighted forecast also factored in how much time (average) was still required to complete all remaining squares on the BINGO card. If the quarter was closing in 13 days and you still needed an average of 29 days to complete the remaining actions, we calculated that into our number.
When we finally rolled the forecast up to the executive and finance teams each week, we gave them three numbers, just like we had done everywhere else:
Best Case
Most Likely
Worst Case
However, each of these three numbers was also triangulated based on three different models:
Manager Call
RevOps Weighted Call
RevOps Time-Weighted Call
This method gave us nine unique “call” values that let us see nearly all possible outcomes for the quarter.
Once we had gotten our modeling to a place where we were happy with it, which took about two-quarters of tweaking, the result was a repeatable model that allowed us to forecast the quarter with 80-90% accuracy by the second week of every quarter.
