When Forecasts Get Real: Moving Beyond Single-Point Thinking

You are in a supermarket, confidently filling your basket with the exact ingredients for the perfect recipe you have chosen for your eight dinner guests this evening — they have all accepted your invitation, and you’re absolutely sure they will show up. Reality hits when you open the front door: one of your friends brings an unexpected plus-one, another brings their new vegetarian partner, and another brings their own dessert! Your perfectly planned dinner transforms into an impromptu cooking challenge.

So often, this is how we approach forecasting in product development — with an almost adorable level of certainty that would make weatherpeople laugh. Just like our dinner party planning, we convince ourselves we can predict exactly how things will play out, only to find ourselves rapidly adapting to new realities: unexpected stakeholder requirements (the plus-one), changing user needs (the vegetarian partner), and competing solutions we hadn’t anticipated (the surprise dessert).

The Seductive Math of Maybe

Our brains are masterful storytellers but terrible probabilistic thinkers. Like a GPS that only shows the fastest route without accounting for traffic, construction, or the coffee stops you’ll inevitably need along the way, we default to single-point forecasts that ignore the messy reality of product development.

Think about that last project plan you saw. Did it look something like this?

• Design: 2 weeks
• Development: 4 weeks
• Testing: 1 week
• Launch: Done! 🚀

Seems reasonable, right? It’s like assuming you’ll hit every green light on your commute because it happened once last month.

The Three Flavors of False Certainty

The “Just Like Last Time” Fallacy

Remember that project that went smoothly last quarter? Your brain certainly does — and it’s playing a clever trick on you. In crystal-clear hindsight, that project looks like a well-oiled machine, every part moving in perfect synchronization. But let’s be honest: when was the last time two projects were genuinely identical? Hand on heart, has a project ever been “just like last time”? Think about it. Even if you’re building the exact same feature (you’re not), with the exact same team (unlikely), using the exact same tech stack (probably evolved), for the exact same customer segment (their needs have changed), in the exact same market conditions (impossible) — you’re still dealing with different humans, different contexts, and different moments in time. It’s like claiming you’ll have the same conversation twice. Sure, you might cover the same topic, but the dynamics, the energy, and the outcomes will inevitably differ. Our minds love to smooth over the messy details of past projects, creating a highlight reel that edits out the uncertainty, the pivots, and the late-night troubleshooting sessions. We remember the successful launch but forget the three architectural debates that preceded it. We remember hitting the deadline but forget the scope adjustments that made it possible. We remember the positive customer feedback but forgot the two rounds of usability issues we had to solve first. This sanitised version of history becomes dangerous when we use it to predict the future. “The last payment integration took six weeks” becomes a promise rather than a data point, ignoring that this time, we’re dealing with different APIs, different compliance requirements, and different user expectations.

The “We’ve Thought of Everything” Illusion

Have you ever watched a detective show where they’ve built the perfect case and accounted for every detail, only to have something completely unexpected blow their theory apart? That’s product development. The illusion isn’t just that we’ve thought of everything — it’s that we could possibly think of everything in a world where customer behaviours emerge, technologies shift overnight, and market dynamics play out in ways that would have seemed impossible six months ago. Think about it: in 2019, did any product roadmap account for a global pandemic that would transform how we work? Of course not. But that’s just the dramatic example. The reality is that every week brings micro-versions of these unpredictable shifts: a key customer discovers a novel use case that changes your feature priority, a competitor releases something that reshapes market expectations, or your seemingly simple integration reveals complex edge cases that no amount of upfront planning could have surfaced. In complex domains like product development, the idea that we can think through all possibilities upfront isn’t just optimistic — it’s a fundamental misunderstanding of how complex systems work. It’s not about failing to be thorough; it’s about recognizing that the territory changes as we walk it. The map we drew yesterday might not match today’s reality, let alone next month’s.

The “Average is Normal” Trap

If your team typically delivers features in two weeks, you might think planning for two weeks is realistic. But that’s like saying the average family has 2.5 children — technically precise, but not accurate, and not particularly useful for buying groceries. This is what Stanford professor Sam Savage calls “The Flaw of Averages,” which he brilliantly illustrates with this example: a statistician who drowns crossing a river with an average depth of three feet. Just as the “average” depth tells you nothing about the deep spots that might sink you, your team’s average cycle time hides the peaks and valleys that make up real project life. As Savage memorably puts it, “plans based on averages fail on average.” Think about it: if Jeff Bezos walks into a bar, the average net worth of everyone there suddenly becomes astronomical — but that doesn’t make everyone in the bar a billionaire. Similarly, when we use average delivery times to predict future performance, we’re planning based on a number that, by definition, is wrong at least half the time.

Moving Beyond Single-Point Thinking

Let’s face it: single-point forecasting is like betting everything on a single dice roll. Instead of playing fortune-teller with fixed dates, we must embrace a more sophisticated approach that acknowledges the inherent uncertainty in knowledge work.

Embracing Probabilistic Forecasting

The future isn’t a single point on a map — it’s a range of possibilities, each with its own likelihood. When we forecast probabilistically, we:

• Accept uncertainty as a fundamental part of the process, not a flaw in our planning
• Communicate in ranges and confidence levels: “We’re 85% confident we’ll deliver on or before August 15th.”
• Use tools like Monte Carlo simulation to generate distributions of possible outcomes based on actual historical data

Focus on Process Predictability

Better forecasting starts with better processes:

• Monitor and maintain stable throughput
• Manage variability by identifying and addressing bottlenecks
• Use pull policies to reduce work-in-progress and minimise context-switching
• Track and actively manage Work Item Age — this is your leading indicator of problems and the key causal metric for Cycle Time. The older items get, the more likely they are to age even further.

Making Better Forecasts

Intelligent forecasting isn’t about being more precise — it’s about being more realistic:

• Work with shorter timeframes where uncertainty is naturally reduced
• Reforecast regularly as new information emerges
• Use your actual throughput data rather than relying on estimates
• Pay attention to your process assumptions and adjust accordingly

Avoiding the Common Traps

Steer clear of these forecasting pitfalls:

• Don’t rely on averages — they hide the variability that matters
• Minimise estimation — it often consumes more value than it creates
• Avoid linear projections — they rarely reflect reality
• Don’t force-fit data to predetermined patterns (often called curve-fitting)

Remember: The goal isn’t a perfect prediction — it’s better decision-making through honest acknowledgement of uncertainty. Your forecasts should help you navigate the complexity of knowledge work, not pretend it doesn’t exist.

Making the Shift

Moving from single-point to probabilistic thinking isn’t just about better numbers — it’s about fundamentally changing how we think about the future. Like a meteorologist who knows they can’t promise sunshine for your weekend wedding, we must embrace the art of educated uncertainty.

Core Principles of Better Forecasting

1. Think in Probabilities, Not Certainties

• Instead of saying, “We’ll deliver by March 15th,” try, “We’re 85% confident we’ll deliver on or before March 20th”
• Acknowledge multiple possible futures and plan for them
• Remember: a forecast without a range and confidence level is just a guess wearing a suit

2. Embrace Shorter Timeframes

• The further out you look, the more dragons live there
• Break long-term forecasts into shorter chunks where uncertainty is more manageable
• Think of it like weather forecasting: tomorrow’s prediction is more reliable than next month’s

3. Make Reforecasting Your Superpower

• New information isn’t a sign your forecast failed — it’s an opportunity to get smarter.
• Update your forecast when key assumptions change or new data arrives.
• Think of each re-forecast as upgrading your GPS with real-time traffic data

Every Good Forecast Needs Three Elements

1. A clear range of outcomes
2. The probability for the range you are communicating
3. An expiration date for the forecast

Remember: a forecast is a living thing, not a monument. When new information arrives that validates or invalidates your initial assumptions, it’s time for a refresh. This isn’t admitting defeat — it’s embracing reality.