Change one thing in an industrial process and consequences pop up everywhere. A techno-economic analysis that works as an ongoing decision tool, owned across R&D, engineering, and finance, catches the surprises before they kill your project. 

“It feels like a high-stakes game of whack-a-mole,” says Crystal Bleecher, Roebling’s SVP of Engineering. "You make a change over here, and almost immediately something surprising happens over there." 

She's talking about a techno-economic analysis (TEA) that's doing its job well. Every data point and assumption in the model is wired to all the others. Change a feedstock concentration, and equipment sizing has to change, along with the margin on every kilogram of product. Move a site, and CapEx, utility loads, financing options, equipment dimensions, and the project's carbon footprint move with it.

Bleecher has two decades as a process engineer scaling first-of-a-kind technologies, mostly in industrial biomanufacturing: early years in operations at biofuel startups, a decade leading process design at an engineering firm with dozens of clients. 

From the inside, she was seeing a pattern she couldn't fix. Engineering and finance lived in separate models and reconciling them took serious manual effort, so it happened sporadically and in low fidelity. "We would execute on our scope of work, only to see the project fall apart because they couldn't get it financed," says Bleecher, who joined Roebling to help break that pattern.

TEAs are often produced as contracted deliverables, to satisfy a stage gate or a funder. They produce snapshot numbers for a specific decision. But engineers and scientists make hundreds of design decisions over the life of a project, and that snapshot keeps getting cited long after it has stopped describing what they're building.

Without a live model showing how each choice ripples through the economics, those decisions come down to experience and intuition. "Most of the time we're right," says Bleecher. "But sometimes we miss a mole."

That’s why a great TEA cannot be just a deliverable. It must be an ongoing decision tool.

The TEA you commissioned six months ago is already wrong

Sometimes a TEA is DOA. 

Bleecher recalls a recent Roebling project where her team was brought in to review a client's existing TEA. “It looked pretty robust. Multi-tab spreadsheet, great calculations – clearly the work of a competent engineer,” she says. Under the hood, though, critical data were stale. Equipment cost data in the CapEx section was pulled, unadjusted, from a decade-old project. Feedstock prices were two years old, missing a recent price hike.

When the team refreshed the inputs, without touching any of the math, the project came in 50% more expensive than the original number. The IRR collapsed, forcing the company to prioritize a different product for their launch. "They would rather have made that decision a year earlier," says Bleecher.

Bad news early is good news to get

Before joining Roebling, Bleecher watched a client finish FEL-2 and then decide to switch sites for a co-location partner. The expectation was a quick update to the outside battery limits – the utilities and supporting infrastructure around the core process.

But the new site's feedstock was more dilute. That changed everything downstream. Equipment had to be resized. Chilled water wasn't available on site, so chillers had to be added. Energy use went up, raising the carbon footprint, which was a problem for this project's commercial case.

The engineering firm spent roughly 60% of the original FEL-2 cost just to redo the package for the new site. "What seemed like it would be a quick study ended up having a much bigger impact than any of us, at a high level, had anticipated," says Bleecher. A dynamic TEA could have explored these consequences in hours. Before the rework, not during.

Sometimes the TEA arrives in time. Take an early-stage precision fermentation company that came to Roebling with just bench-top data for a dye they were looking to produce. The first TEA Roebling built revealed the project simply wasn’t commercially viable. "I felt like I was a project killer!" Bleecher says. But six months later the company came back with a new host organism, a higher titer, and a more valuable target molecule – the levers the dynamic TEA had flagged.

Two audiences, two different priorities

The engineering team needs the TEA model to hang together: balances that close, equipment sized for commercial scale, sensitivity work focused on the assumptions that, if wrong, kill the project. The business audience needs the financials to stack up. Is the process viable? What's the margin of safety? What's the downside case? They read the TEA to decide where capital goes.

In most companies, the two audiences aren't reading the same document at all. Process engineers own the mass and energy balance in one spreadsheet. A cost estimator owns CapEx in another. Finance owns a third model that depends on outputs from the first two but isn't dynamically linked to them. (And let’s not forget that mysterious pump-sizing spreadsheet from 1992 that no one fully understands, built by someone now retired.)

"That's the big problem," Bleecher says. "Different teams on these different models, not talking to each other." Too many TEAs fail at the seams between teams.

Whack-a-mole with the lights on

A dynamic TEA has three properties most traditional TEAs don't:

• Integration. Process, cost, and finance live in one model. Change a unit operation, and the financial model updates. Change a site, and CapEx ripples. Nobody has to re-key numbers between spreadsheets, so nobody forgets to.

• Continuous updating. The model evolves with every design decision and every new piece of data: a fresh vendor quote, new bench data, a swapped unit operation. "Engineering becomes a continuum," says Roebling co-founder Brentan Alexander, with results updated continuously as uncertainties narrow, from R&D through to the final investment decision.

• Cross-functional ownership. Each team owns its piece of the model: process engineers the mass and energy balance, cost estimators the CapEx, finance the project economics. A single source of truth lets them work seamlessly together. "These teams need to own different sections of the TEA," says Bleecher, "and it all comes together as a powerful decision tool."

Roebling’s platform was built around this problem. Process simulation, equipment sizing, CapEx, and the financial model live in one environment. A change anywhere ripples through the whole model. AI agents handle the mechanical work: sizing equipment, running sensitivity analyses, regenerating deliverables when the design basis shifts. The engineer provides direction and judgment.

As anyone who works in industrial process design knows all too well, the moles never stop coming. The only question is whether you spot them in time.  

About the author
Sean O’Neill is a seasoned editor, journalist, and science communicator. He spent 14 years at New Scientist and served as the magazine’s People Editor. Sean has written for Google DeepMind, served as communications lead at the Cambridge Centre for AI in Medicine at the University of Cambridge, and worked as a Science Writer for The Alan Turing Institute in London. You can also find his byline on AI stories and profiles on the Bentley blog.