Forecasting models are evidence‑starved

Most sales forecasts still model what sellers say about a deal, not what buyers have actually done. Spotlight.ai and Atlantic Growth Solutions both argue that stage changes should be tied to proof events such as a successful proof of concept, legal review, or procurement engagement. (spotlight.ai) (atlanticgrowthsolutions.com) Sales forecasting is the practice of estimating future revenue from open deals, and many teams still do it with manager roll-ups, spreadsheet updates, and fixed stage probabilities. Spotlight.ai wrote on February 13, 2026 that traditional systems ask reps “how they feel” about a deal, while evidence-based systems score buyer behavior pulled from customer relationship management records, emails, and calls. (spotlight.ai) That critique lands in a market where accuracy is already weak. Gartner figures cited by Spotlight.ai, Outreach, and other sales-operations vendors say only 7% of sales organizations reach forecast accuracy of 90% or better, while 69% of sales operations leaders say forecasting is getting harder. (spotlight.ai) (outreach.ai) (demandgenreport.com) The core idea is simple: a forecast gets stronger when a stage means the same thing every time. Atlantic Growth Solutions wrote in its Q2 2026 “Revenue Engine Manifesto” that teams need one shared qualification protocol, enforceable gates, and a single handoff standard between marketing and sales. (atlanticgrowthsolutions.com) In practice, that means a seller cannot move a deal forward just because a champion sounded positive on a call. The deal advances when a buyer completes a concrete step, such as bringing in procurement, confirming budget, expanding stakeholder access, or finishing a proof of concept. (spotlight.ai) (salesmotion.io) The argument also reflects a broader shift in revenue operations toward automated data capture. Spotlight.ai says its platform inspects calls, emails, and customer relationship management data to produce “bottom-up forecasting,” while AskElephant and other vendors make the same case that delayed or incomplete customer relationship management updates distort the forecast before any model runs. (spotlight.ai) (askelephant.ai) Atlantic Growth Solutions frames the problem less as bad rep judgment than bad system design. In its March 2026 manifesto, the firm pointed to 60% of leads dropping out at discovery, 4-to-6-month revenue lag after headcount increases, and roughly 30% forecast inaccuracy in technology and Internet of Things sales tied to customer relationship management leakage and poor stage hygiene. (atlanticgrowthsolutions.com) Not everyone describes the fix the same way. Some vendors pitch artificial intelligence models that learn from historical wins and losses, while others stress governance first, arguing that no model can rescue a pipeline built on inconsistent definitions and stale fields. (spotlight.ai) (apollo.io) (alturagtm.com) The common thread is that “commit” numbers are being pushed closer to buyer evidence than seller confidence. If that shift holds, the forecast meeting becomes less a weekly opinion poll and more an audit of whether the customer has taken the next real step. (spotlight.ai) (atlanticgrowthsolutions.com)