On this page
Every Monday morning across thousands of commercial organisations, the same ritual takes place. The revenue operations team pulls pipeline data from Salesforce, HubSpot, or an AI forecasting platform like Clari. The system produces an objective, calculated projection for the quarter: the exact operational baseline meant to provide one number the commercial team shares.
Then the sales leaders sit down to review it.
Deal by deal, category by category, they make adjustments. An enterprise deal stuck in legal gets nudged forward because the account executive insists the buyer gave a verbal promise. A region falling short of target receives an extra 5% bump in commit because the regional vice president is confident the team will pull through. By the time the forecast reaches the chief revenue officer and the board, it is no longer the machine’s number. It is the number the commercial leadership decided to call.
The implicit belief behind this weekly exercise is that managerial intervention adds signal. Sales leaders believe their qualitative proximity to deals, their reading of buyer intent, and their institutional experience improve the forecast.
The empirical evidence says the opposite.
When human judgment overrides a quantitative forecasting baseline without strict constraints, it systematically degrades accuracy. And in the largest study ever conducted on the subject, the damage was not random. It had a clear, predictable mathematical signature: direction dominates size.
Why do manual managerial overrides systematically distort revenue forecasts?
While commercial software vendors produce endless white papers on forecast hygiene, the academic literature on judgmental forecasting has spent decades measuring what actually happens when human managers adjust statistical baselines.
The definitive field study was published in the International Journal of Forecasting by Fildes et al. (2009). The researchers analysed 68,984 complete forecast triples: the initial statistical baseline, the manager’s adjusted forecast, and the actual realised outcome: across four large corporate supply chains over two to three years.
Figure 1Directional impact of judgmental forecast adjustments
The mirror isolates the asymmetry: upward overrides overshoot reality in every organisation more often than downward cuts do.
Source: Fildes et al. (2009), International Journal of Forecasting 25(1), Tables 1, 5a and 5b.
The findings overturn standard management intuition in three specific ways:
- Upward adjustments are overwhelmingly wrong-signed. Across the manufacturers, 66% of upward adjustments resulted in a final forecast that was higher than actual demand. At the retail firm, 83% of upward adjustments overestimated reality. When managers pushed numbers up, they produced inaccurate, inflated numbers three to four times out of five.
- Downward adjustments consistently add value. Negative adjustments overestimated demand only 46% of the time across the entire dataset. In other words, when managers cut a forecast, they were right more often than not. Downward overrides acted as an effective correction to overly optimistic baselines.
- Small adjustments destroy time without improving accuracy. Managers spent considerable effort making minor adjustments (under 5% to 10%). Fildes and colleagues found that small adjustments yielded marginal or negative accuracy gains. As the authors noted in §5.3, preventing smaller adjustments “would marginally enhance accuracy, but also, more importantly, it would free up the time of those involved in the forecasting process.”
Why the asymmetry exists: the B2B translation
Fildes et al. studied SKU-level demand forecasting in supply chains. A B2B revenue pipeline is obviously different: deals are discrete, deal values are heterogeneous, and sales cycles stretch across months.
However, modern revenue operations platforms replicate the exact architecture Fildes evaluated: a quantitative baseline derived from stage progression, historical conversion rates, and engagement telemetry (the very layer examined when measuring what AI changes in revenue operations), followed by a layer of human managerial override.
The reason the directional bias travels from supply-chain planners to sales vice presidents is that the human loss function is asymmetric.
When a sales manager considers an upward adjustment, they are responding to quota anxiety, executive pressure, and optimistic self-persuasion. An AE who needs one more deal to hit their target will convince themselves that a procurement delay is merely a formality. The manager, wanting to present a viable path to quota on Monday’s executive call, endorses the override. The cost of admitting the gap today feels higher than the distant risk of missing the number at the end of the quarter.
Downward adjustments happen under the opposite psychological conditions. Sales leaders rarely cut a committed deal unless they possess hard, unmodeled negative information: the champion left the company, a competitor dropped their price by 40%, or the budget was frozen by the CFO. Because negative adjustments are socially and professionally uncomfortable, they are only made when the evidence is incontrovertible.
As a result, unrestricted upward adjustments inject noise and wishful thinking, while downward adjustments inject genuine private information.
How can commercial teams apply the Asymmetric Override Protocol to protect forecast accuracy?
Telling sales leaders to stop adjusting forecasts entirely is neither realistic nor desirable. Private information does exist in sales: legal roadblocks, executive sponsor departures, and sudden procurement vetoes cannot always be captured immediately by CRM stage rules.
The solution is not to ban judgment, but to install asymmetric friction.
| Forecast Dimension | Standard Operating Practice | Asymmetric Override Protocol |
|---|---|---|
| Upward Adjustments (Pushed into Commit) | Based on AE verbal confidence or managerial gut feeling. Low friction. | High Evidentiary Friction: Requires verifiable, external proof (e.g. approved redlines, completed security audit, executive sign-off). Mandatory written justification. |
| Downward Adjustments (De-committing Deals) | Discouraged during pipeline reviews; perceived as sandbagging or lack of grit. | Zero Friction: Immediate and penalty-free. Encouraged whenever deal velocity slows or champion engagement stalls. |
| Micro-Adjustments (under 10%) | Constant weekly tweaking across dozens of mid-funnel deals. | Banned: Overrides below a 10% threshold are locked to prevent wasted managerial bandwidth. |
| Accountability & Tracking | Only the final called number is tracked against final actuals. | The Three-Column Audit: CRM logs System Baseline, Manager Override, and Actual Outcome to score managerial batting averages over time. |
Table 1The RevOps Asymmetric Override Protocol
Replacing symmetrical managerial discretion with structural friction: high proof standards for upward optimism, zero friction for risk reduction, and systematic logging of override efficacy.
Operating framework adapted from Fildes & Goodwin (2007) and Fildes et al. (2009).
To implement this on your commercial team:
1. Require evidentiary hurdles for upward adjustments
Never allow a deal to be moved into commit or adjusted upward in value during a forecast call based on conversation alone. Establish three non-negotiable proof criteria:
- Documented confirmation of budget allocation from an identified economic buyer.
- Completion of technical and legal milestones (e.g. security questionnaire approved, Master Services Agreement in redlines).
- A mutual close plan agreed in writing with the counterparty.
If these criteria are absent, the statistical baseline stays untouched.
2. Remove friction from downward adjustments
Celebrate early warnings. When a rep or frontline manager identifies that a deal is stalling and cuts it from commit, accept the change immediately without treating it as a performance failure. The earlier a realistic shortfall is acknowledged, the earlier marketing and outbound capacity can be reallocated to fill the pipeline.
3. Maintain an override ledger in your CRM
In your CRM or BI layer, maintain three distinct numbers throughout the quarter:
Forecast_Baseline: The unadjusted quantitative or algorithmic prediction.Forecast_Final: The number called after managerial adjustments.Actual_Revenue: The realised bookings at quarter close.
At the end of every quarter, calculate the Mean Absolute Percentage Error (MAPE) of both columns against actuals. If a sales leader’s adjusted forecast has a higher error rate than the unadjusted baseline, their overrides are destroying information.
As Fildes and Goodwin established in their 2007 Interfaces study, the simple act of requiring managers to document their reasons and systematically auditing their override track record eliminates over half of all value-destroying adjustments. Furthermore, because forecast accuracy can hide offsetting errors, tracking stage adjustments prevents positive and negative errors from canceling each other out in the headline percentage.
The next time your leadership team meets to call the number, remember what 68,000 forecasts prove: when you feel the urge to push a number up, you are probably guessing. When you cut it, you are probably right.
Boundary
Boundary. The source studies SKU forecasting, not B2B pipelines. Treat the asymmetric override rule as a testable transfer hypothesis and compare adjusted forecasts with an untouched baseline.
References
- Fildes, R., Goodwin, P., Lawrence, M., & Nikolopoulos, K. (2009). Effective forecasting and judgmental adjustments: An empirical evaluation and strategies for improvement in supply-chain planning. International Journal of Forecasting, 25(1), 3–23. https://doi.org/10.1016/j.ijforecast.2008.11.010
Updated 1 change
- Refresh on 20 August 2026: The source studies SKU forecasting, not B2B pipelines. Treat the asymmetric override rule as a testable transfer hypothesis and compare adjusted forecasts with an untouched baseline.
