What it means
When planning a new project, humans naturally fall into the trap of optimism bias, assuming everything will run smoothly and according to plan. Reference class forecasting acts as a reality check by ignoring the unique promises of the current project plan and looking at hard data from past efforts.
It asks a simple question: when people like us tried to do things like this before, how long did it actually take, and how much did it really cost? This technique was pioneered to combat chronic cost overruns and delays in large public infrastructure projects, but it is equally powerful for business managers.
By treating your new project not as a one-off miracle, but as a statistical sample of a broader category, you remove personal ego from the numbers. You start by identifying a reference class of past projects that share core traits with your new proposal.
You gather their actual outcomes, noting the average variance between initial estimates and final results. You then adjust your current forecast to match historical reality.
For non-finance managers, adopting this habit protects your credibility and your budget. It shifts your planning from wishful thinking to evidence-based forecasting, ensuring your team is not constantly surprised by preventable overruns.
It is like baking a cake for a dinner party. Even if you buy the finest ingredients and follow the recipe carefully, past experience tells you that cleaning the kitchen afterwards always takes twice as long as you initially hope.
In practice
Real-world examples
Example 1
A tech startup plans to launch a new mobile app in three months with a $20,000 budget. Reference class forecasting of ten similar past app launches shows an average duration of six months and an average cost of $45,000. The founders revise their plan, phase the launch, and tell investors the realistic timetable.
Example 2
A local manufacturing firm wants to expand its warehouse space, estimating a six-week build time. By reviewing data from fifty past regional warehouse expansions, they discover the average actual build time is ten weeks. They adjust the promises made to clients, avoiding late-delivery penalties.
Example 3
A restaurant chain plans to open a new city-centre branch, budgeting $50,000 for marketing. Historical data from twenty previous openings reveals initial marketing spend always exceeds budget by 40%, leading them to set a safer budget of $70,000. The area manager now uses that figure in every opening plan.
Think of it
It is like baking a cake for a dinner party. Even if you buy the finest ingredients and follow the recipe carefully, past experience tells you that cleaning the kitchen afterwards always takes twice as long as you initially hope.
Formula
Calculation
Adjusted Forecast = Initial Estimate x (1 + Average Historical Overrun Percentage).
Worked example: your team estimates a software upgrade will cost $10,000. Past data for similar upgrades shows an average cost overrun of 50%. Adjusted Forecast = 10,000 + (10,000 x 0.50) = $15,000.
Here is how to find the average overrun. A manager reviews five similar past projects, which overran their estimates by 20%, 30%, 40%, 50% and 60%. The total is 20 + 30 + 40 + 50 + 60 = 200, and 200 / 5 = a 40% average overrun. A new $50,000 project is therefore forecast at 50,000 x 1.40 = $70,000, and the same method can be applied to timelines.Case study
Seen in the real world
This is a fictional, illustrative scenario. Oak Tree Logistics decided to upgrade its fleet management software, with the internal project team promising a swift delivery within four months and a budget of $80,000. The finance director, wary of past delays, applied reference class forecasting. She reviewed twenty previous software implementations within the mid-sized transport sector.
The historical data revealed a stark contrast: similar projects averaged seven months to complete and cost $130,000 due to unforeseen data migration challenges. Armed with this outside view, Oak Tree revised its budget and timeline before signing contracts. When the project eventually finished in six and a half months at a cost of $125,000, leadership was prepared. The revised forecast prevented a cash flow crisis and maintained stakeholder trust, proving the value of learning from industry peers rather than internal optimism.
The comparison with the original plan was striking. The $130,000 reference figure was 62.5% above the team's $80,000 estimate (50,000 / 80,000). Against the internal budget the project would have been $45,000 over, a 56% overrun, but against the revised $130,000 forecast it came in $5,000 under, and the board noticed the difference.
Watch out
Common mistakes
- Treating your project as completely unique and assuming past data does not apply to you.
- Selecting a reference class that is too narrow, resulting in a skewed sample size.
- Failing to update your historical database as your organisation gains new experience.
Questions
People also ask
How is this different from traditional budgeting?
Traditional budgeting builds estimates from the ground up based on current plans, which often leads to optimism bias. Reference class forecasting starts from the top down using historical reality.
What should I do if my project has no direct historical matches?
Broaden your reference class. If you cannot find data on launching a specific type of digital cafe, look at general retail store openings or standard software deployments for a baseline.
Will using this method always increase my budget and timeline?
Usually yes, because most projects suffer from optimism. However, it simply aligns your plans with historical truth, reducing nasty surprises later.