Method
Most companies don't lack ambition. They lack a function that can tell the difference.
Between a technology that's ready and one that isn't, and the discipline to prove it before betting the operation on it. That function is what we build, embed, or run on your behalf. Here's exactly how it works.
Phase Zero
The Reading: state of the art and TRL diagnostic.
Before we recommend a single hour of lab time, we read. Every engagement opens with a structured review of four things.
- 01
The global state of the art
What's already deployed commercially. What's still stuck at pilot scale. And what's real, but still buried in a paper nobody in your industry has read yet.
- 02
Technology Readiness Level, calibrated to evidence
We score every candidate technology on the standard nine-level TRL scale. We anchor it to verifiable deployment, not a vendor's self-assessment.
- 03
Your own readiness
A separate score: what your organization, your supply chain, and your field conditions can actually execute today. Global maturity and local readiness are two different numbers. The opportunity, and the risk, live in the distance between them.
- 04
The falsifiable questions
A short list of what the program has to find out, never a technology wish-list. If we can answer it with our own data, the engagement succeeded, even if the answer is that this doesn't work here.
The output is a written reading, not a slide: what's proven, what's assumed, what's still theory, and where your organization sits on each axis. It's the only phase that happens before we ask for more resources than the reading itself takes.
The Four Gates
Every gate is judged against criteria written down before the work starts.
Theory & Technology Scouting
- What it produces
- Literature and patent review, first-principles memo, a falsifiable hypothesis, an academic partner engaged
- What has to be true to continue
- A testable hypothesis exists, and there's a viable way to measure it with the resources available
Laboratory Proof of Concept
- What it produces
- Controlled experimental data, a repeatable measurement protocol, a first patentability read
- What has to be true to continue
- The effect exists and behaves at the expected magnitude under controlled conditions
Functional Prototype at Bench Scale
- What it produces
- A prototype that works outside the lab, detailed engineering, a defensible deployment estimate
- What has to be true to continue
- Performance holds under representative conditions, and the deployment case is credible
Field Validation & Transfer
- What it produces
- An instrumented installation on real operating assets, a technical dossier, trained personnel, a verified unit cost
- What has to be true to continue
- Performance holds through at least one full operating cycle, on your own data
Gates are a resource-allocation mechanism, not a punishment for failure. They move effort toward whatever the evidence is rewarding, and free it from whatever it just finished teaching you. A well-run portfolio concentrates resources as it advances. That's only possible if each gate is judged against criteria written down before the work starts, never after.
A negative result, properly documented, is not a failed gate. Knowing that a technology doesn't work under your conditions is worth as much as knowing one that does. It keeps you from selling it, and it keeps a competitor from selling it to you unchallenged.
The Academic-Linkage Engine
We don't build the theory from scratch. We license it from the people already building it.
The early stage of almost any research line is already being done, well, inside a university lab or a national research center. It's done by people with deep domain expertise, but no access to your field conditions, your capital, or your operating data. That asymmetry is the opening. A structured partnership turns a supervised thesis into a field experiment, and that experiment into a capability you can sell or deploy. It's cheaper, faster, and more defensible than trying to originate all of it in-house. It's how nearly every piece of infrastructure you rely on today actually got built: private capital and public research meeting in the middle, long before the return was obvious.
Holden R+D maintains those relationships: research groups, national laboratories, standards bodies, industry working groups. So when a program needs a specific kind of expertise, the relationship already exists. Participating in the standards conversation is part of the same motion. It's the cheapest way we know for a company to learn where a regulation is headed before it's published.
Portfolio Discipline
Why one technology has to serve many lines of business.
No line of research earns a place on the agenda by serving a single customer. Every candidate technology has to serve more than one part of your business before it qualifies for investment. A technology that only pays off in one product line, one client, or one contract inherits that contract's politics and its calendar. A technology that serves multiple lines survives a change in strategy, a lost client, or a slow year: it was never betting on just one.
The Priority Index
How we rank fourteen good ideas when you can only fund four.
Every program surfaces more credible technologies than any organization can run at once. Rather than choose by instinct, or by whoever pitched last, we score every candidate against the same five weighted criteria.
- 35%Industry impactHow much this actually moves the operation, or the market it serves.
- 30%Technical potentialThe ceiling on the technology itself, independent of who executes it.
- 20%Business-line coverageHow many parts of your business it actually serves.
- 10%Global maturityHow much exploration risk someone else's deployment has already retired.
- 5%Resource efficiencyReturn per unit of budget and headcount committed.
The weighting is deliberate. Impact and technical ceiling dominate: a cheap technology that doesn't move anything is still a waste, and an expensive one that changes the business is still worth funding. The index doesn't make the decision for you. It only makes sure the decision is made against the same yardstick every time, and that it's still defensible in front of a board a year later.
How We Label What We Know
Every number and claim in a Holden deliverable carries one of three tags.
- Verified
- Checked against a primary source or an official publication.
- Probable
- A sound inference, or a reputable analyst's projection.
- Estimate
- Our own calculation, with the assumptions stated next to it.
Most technical proposals blend these into one confident voice. We don't. A client acting on our work needs to know which sentence to bet the operation on, and which one to verify with a pilot of their own. This also means our estimates get sharper over time, in public. We're not protecting a number we got wrong: we labeled it an estimate from the start.
Leading Indicators
How to tell if an R+D function is working before it generates revenue.
Judging a research program by revenue in its first years forces it to lie about its progress, or to stop doing research. These are the indicators that actually show whether the function is working. Every one of them is checkable by someone outside the department.
Technical output
- Gates cleared per year
- TRL advancement per technology, tracked as a year-over-year delta
- Internal measurement protocols published and reused
- Negative results, documented and formally closed
Defensible assets
- Patent applications filed, and international families built from them
- Registered and controlled trade secrets
- Internal procedures transferred into live operations
- Proprietary performance data under your own operating conditions, the kind no vendor's catalog contains
Position and capacity
- Staff certified in new technologies
- Active academic agreements and supervised theses
- Seats on standards committees
- Papers and talks at the field's own conferences
We report against these indicators from day one. Leadership agrees to the cadence before the program starts, not once results happen to look good.
Six Conditions
What decides whether this survives contact with your organization.
None of the six is a technical obstacle. All are management decisions, within your control. And all are far cheaper to fix on day one than in year two.
- 01
Autonomy from operations
The most common way an internal R+D function dies: it reports into operations, and within two quarters its budget and headcount are fighting this week's fire. The fix is structural: a direct line to leadership, plus a simple tripwire. If more than a fifth of the team's hours go to routine operational problems in any quarter, the function has already become technical support with a different name.
- 02
Retention of specialized talent
The market for these skill sets is global. People with this kind of technical depth go where the problems are interesting. Retention runs on a real technical agenda and a budget to publish, not on salary alone.
- 03
Aligned expectations with leadership
Year one brings enthusiasm. Year two brings cost without visible results. Year three brings the first validations. The only fix is agreeing on the metrics above before the program starts, and reporting them on schedule without exception.
- 04
Vendor independence
If every piece of evidence is generated on one manufacturer's equipment, the conclusion will always be that this manufacturer is the best one. Every test bench is designed multi-brand from the outset, even when that costs more.
- 05
IP settled at the start, not the close
Every academic agreement gets reviewed before signature, with ownership and commercial-use terms written down. A discovery shared without a prior agreement is a discovery you can't use.
- 06
A sequenced agenda, not a simultaneous commitment
The full agenda is a menu to choose from, not a workload to carry all at once. One team per research line means a handful of active lines is the realistic ceiling in year two, and the Priority Index exists specifically to choose which ones.