Bad assumptions in early-stage techno-economic analysis don't just produce wrong numbers. They produce wrong decisions. And those decisions compound through FEL gates until the cost of correction is hundreds of thousands of dollars — or a shelved project.

In this on demand webinar, Crystal Bleecher, P.E. (SVP Process Engineering, Roebling) and Frederick Twigg (UC Berkeley) unpack how it happens: how assumptions go untracked, how single-scenario models create false confidence, and what it actually means to layer AI on top of engineering analysis responsibly.

Top 5 Takeaways

1. A single scenario creates false confidence. One configuration under one set of assumptions tells you nothing about what happens when conditions change. And conditions will change. A single-scenario model doesn't de-risk a project. It hides the risk.

2. Making assumptions isn't the failure. Losing track of them is. Every early-stage TEA requires assumptions. That's the nature of the work. The problem starts when those assumptions go untracked and unversioned, then surface as surprises during FEL-2 or FEL-3 when rework costs real money.

3. Assumptions compound, and that's where projects break. Crystal walked through a project where a downstream chromatography step nearly sank the economics. The model carried three assumptions for one unit operation: resin binding capacity, resin cost per liter, and resin longevity (how many cycles the resin can run before it needs to be replaced). When the team tested resin cost by itself, the impact on project economics looked modest, because the model assumed the resin lasted a long time and the cost was spread over many cycles. But when they tested resin cost and shorter resin longevity together, the compounding effect on OPEX was severe enough to force a complete rethink of the downstream process. Single-variable sensitivity analysis missed it entirely.

4. Not every input matters equally if you're wrong by 50%. Being off on a pump cost probably doesn't move the needle at commercial scale. Being off on a chromatography system or spray dryer can break the entire business case. Know which assumptions carry weight — and pressure-test those first.

5. AI amplifies whatever foundation it sits on. Rigorous deterministic model with transparent data? AI is a genuine multiplier. Bad assumptions and missing context? You get garbage out faster and with more confidence. That's worse, not better.

How Roebling Helps Engineering Teams Tackle These Challenges

These problems are why we built Roebling.

Roebling is an AI-enabled modeling platform that connects process simulation, equipment sizing, cost estimation, and financial analysis in a single live environment. Every assumption is tracked. Every output traces back to a deterministic model. Scenario analysis that used to take weeks happens in hours.

Integrated modeling. Change a process input and watch it propagate through equipment sizing, CAPEX, OPEX, and financials automatically. No manual rework. No version control issues.

Transparent, data-backed assumptions. Every data source is visible, modifiable, and tagged with uncertainty ranges. So your team knows exactly where confidence is high and where it isn't.

Scenario and sensitivity analysis at scale. Run hundreds of configurations across technology, scale, site, and cost assumptions to identify which inputs actually drive the investment decision.

Book a strategy session with our team →