Selling Outcomes, Not Software — Questions and Experiments for Newburgh
This memo attempts to frames what we need to learn before deciding how far to move from selling software toward selling outcomes. It lays out the model options, the open questions, and a triaged set of experiments. It does not pick an answer.
It is July 2026. J-51 will be live by Q1 2027, maybe sooner. Once it opens, deals get developed — by us or by someone else. A model that shows up after deals are already being structured is worthless. There is no prize for the best strategy if it lands late.
And the clock is already running. With every passing day, owners start talking to OEMs and service providers about their projects. They make investments in those relationships — switching costs, sunk effort, and feelings to manage all accrue. Every day we don't have a model, we make our own future work harder, because we're prying owners loose from commitments instead of shaping them from the start.
J-51 is a multi-billion-dollar market opening. We saw it coming earlier than anyone. Now everyone else is waking up to it too — and shame on us if we're not ready when it hits.
So this is not a call for more study. It is the opposite: find the fewest decisions and experiments that matter, and be running them by year end — refining against real deals as the window opens, not still arguing frameworks. Everything here favors tests we can start now, on buildings already in front of us, that pay off in weeks.
The question that matters: what can we have in-market and generating signal by December 2026, so that when the J-51 floodgates open we are sharpening a working motion instead of inventing one?
The four canonical delivery methods (Construction Management Association of America, Owner's Guide), ordered by how much design control the owner keeps and how much risk shifts to the builder — not a ranking. The question for each is what role Cadence would play in it.
Design-Bid-Build (DBB). Owner hires a designer, then separately bids the finished design to contractors. Owner holds design risk; sequential and slower; prone to disputes when field conditions differ from the drawings. — Does DBB fit sub-$20k/DU retrofits at all, or is its soft cost and speed exactly the problem we exist to fix?
Construction Management at Risk (CMAR). A construction manager advises during design, then converts to an at-risk builder at a guaranteed price (often a GMP). Owner still holds a separate designer. — Would we ever be the at-risk party — and when is that attractive, financeable, insurable, and operable for us?
Design-Build (DB), incl. Bridging. One entity owns both design and construction, so design risk shifts to it. In the Bridging variant, the owner keeps its own consultant to write the requirements and performance spec, and the design-builder finishes the design and builds. — If our performance spec is the bridging document, does that make us the owner's consultant, the designer-of-record, or something in between — and how much risk comes with each?
Integrated Project Delivery (IPD). Owner, designer, and builder share risk and reward under one multiparty contract. — High-trust and largely legally untested; is it even relevant to us near-term?
"Myth of Third-Party Design" (Jason): the engineer's "design" is mostly the equipment schedule; the real detail happens in contractor submittals — pointing to a performance-spec-plus-referee setup. A hypothesis to test.
"Buying a Transformation": frames the owner risk stack (below) and says turnkey comes closest to covering owner risks but "is hard to scale." That scaling claim is an assumption — test it.
"Earning the Position": proposes a staged path to a market-maker role. Worth debating, not adopting.
Whatever model we pick, its job is to take these off the owner's plate:
Before yes:Pathway (too many options), Narrative (selling a cautious board), Financing (a real way to pay), Supply (a credible installer at a fair price).
After yes:Change orders (field surprises), Schedule (permits, interconnection, resident access), Coordination + accountability (who holds it together and who's liable when something breaks).
What does "selling outcomes" mean, and what's the first product? A confident pathway, a quoteable project, a financing plan, a bid-ready package, an equipment reservation, a managed procurement, an accountable delivery, or a finished project?
What does "buy through Momentum" actually promise? Which version is Bomee picturing — and which is legal, doable, and sellable in 2026?
The language test. At what point does each phrase start implying real construction liability: "helps you buy the project" → "manages the process" → "manages delivery" → "stands behind the project" → "one throat to choke" → "sells the installed project"? This is where liability creeps in — through the words sales uses.
Procurement spread vs. construction spread. Carrying equipment/working-capital risk (where we may have an edge) is a different animal from carrying change-order/schedule/GMP risk (where we don't, until we have data). Separable? Do one first?
Is the procurement spread even worth building around? Assignment 1 suggests it's low single digits and dwarfed by the price differences between OEMs. Real lever or distraction?
What owners want vs. what we can do — kept separate
Which risks do owners most want gone? (pathway, board narrative, price, contractor choice, change orders, schedule, resident access, warranty, performance, savings, financing, operations)
Which can we credibly handle now? For each risk: reduce it with process / price it with judgment / price it with data / hand it to a partner / insure it / contract around it / avoid it for now.
Scope and gate
The minimum quoteability gate. Before any firm price or reservation, what must we know — and for each input, is it mandatory, sampled, unit-priced, or an allowance? (HP count, apartment mix, window fit, existing outlets, panel and service capacity, demand data, backgrounds, resident access, J-51/incentive eligibility, OEM lead time, contractor availability, warranty.)
Does writing a performance spec make us the designer? Owner's-requirements doc (low liability) or designer-of-record (high)? Legal input needed.
The unit-price true-up problem. Worked for countable scopes (radiators) in the Article 321 bidding pilot; but blew up on hard-to-estimate ones (pipe insulation) on small jobs — and we had no paid role to settle disputes. What structure makes unit pricing safe, and how do we get paid to referee it?
Market and partners
What do OEMs want us to be? Direct demand, channel protection, forecastable demand, early payment, fewer bid cycles, performance data — and would they discount for our demand or fear channel conflict? (Differs by OEM — see Assignment 1.)
What do contractors want us to be? Will they bid from our spec; what do they need to price firm; what will they unit-price, cap, or exclude; will they accept our change-order governance; would steady pipeline lower their price?
Engineers and consultants who do the annual BEPS filing and believe they "own" the owner — channel, delivery partner, or competitor? Give the good ones a bounded, non-judgment scope (commissioning, M&V, filing) so they help rather than block.
Which segment do we test first — for best learning, not easiest sale? And how do we spot early whether an owner wants to own accountability (the Related/Rudin tell)?
Economics and product
Which models are real businesses once we price our own effort — revenue/project, margin, labor, working capital, legal/insurance, broken-deal cost, risk carried, repeatability, data captured?
What must Momentum become, and what's essential for the first experiment vs. later?
Owner interviews. A handful of owners across type: which risks block their yes, what they'd pay to remove them, whether fee-crediting and J-51 timing move them, which message lands. → what owners actually buy.
OEM term-sheet requests. Gradient, Innova, Midea: separate and stack volume, lead time, and prepayment, plus warranty, cancellation, substitution, freight, milestones, price-lock. → the real procurement spread + a price-book template.
Contractor interviews. What they need to price firm; what they'll unit-price, cap, or exclude; warranty handling; whether they'll work under our procurement and governance; whether steady pipeline lowers price. → a contractor risk matrix.
Turnkey reality check. We've assumed "turnkey design-build doesn't scale" — but that's unproven. How do today's turnkey players (Carlton, Nova1, VRF Solutions, BES) actually scale or fail: geography, labor, capital, or demand? → facts to replace the assumption.
This quarter — needs a live project or build time #
500-unit quoteability sprint (next week's visit). What we can collect fast, what stays unknown, and classifying every risk as fixed / unit-price / allowance / exclusion / trigger — separating equipment cost from balance-of-system risk. → a draft quoteability report and first unit-price schedule with true-up rules.
Cross-OEM selection engine (v0). All-in comparison (equipment + install + balance-of-system + performance data) across all three OEMs for the 500-unit complex. → proof we can pick neutrally in a way an owner can't do alone.
"Buy" definitions + reservation prototype. Sketch the "buy through Momentum" versions side by side with their risk levels; for the reservation, draft a term sheet — what's verified first, who holds funds, cancellation, site-change handling, freight/damage, reseller exposure. → clear buy definitions + a mock reservation term sheet.
Legal / insurance / licensing review. Where advisory becomes CM becomes contractor; whether we can earn an equipment spread without being a reseller; insurance per model; language to avoid; a contract that starts risk-free with a later GMP toggle. → a boundary memo + draft contract skeletons.
Board narrative test. A board package for one building: pathway, rejected options, what it physically looks like, "why now / why not wait / why this contractor / what after approval" — and whether someone other than Marc can deliver it (the founder-heroics test). → a narrative prototype.
Delivery-model economics. One simple comparison across the models: revenue, margin, labor, risk, working capital, data, repeatability, scale limits. → a table that separates a real business from one that only looks good unpriced.
Construction-phase role pilot. On one project, test a paid post-award role — access planning, submittal and change-order review, requisition review, quantity reconciliation, closeout, warranty handoff. → directly tests the Article 321 failure: do owners value it, do contractors accept it, can we do it profitably?
Not a finished strategy — a way to decide, on the J-51 clock:
Which §4 questions must be answered before Q1 2027, and which can wait.
A ranked shortlist of §5 experiments, each with an owner and a date, weighted toward December signal. Assign the four "start now" experiments Tuesday.
Draft calls to prepare: first 2026 offer; working definition of "buy through Momentum"; first scope and segment; risks we won't take yet; revenue model; required Momentum changes.
The question isn't "should Cadence become a market maker?" It's: what must we learn, build, prove, and risk before we can credibly sell outcomes — and how much of it can we test before the end of the year, when the J-51 floodgates really open?
Strategy Notebook
Selling Outcomes, Not Software — Questions and Experiments for Newburgh
This memo attempts to frames what we need to learn before deciding how far to move from selling software toward selling outcomes. It lays out the model options, the open questions, and a triaged set of experiments. It does not pick an answer.
1. Why This Cannot Wait #
It is July 2026. J-51 will be live by Q1 2027, maybe sooner. Once it opens, deals get developed — by us or by someone else. A model that shows up after deals are already being structured is worthless. There is no prize for the best strategy if it lands late.
And the clock is already running. With every passing day, owners start talking to OEMs and service providers about their projects. They make investments in those relationships — switching costs, sunk effort, and feelings to manage all accrue. Every day we don't have a model, we make our own future work harder, because we're prying owners loose from commitments instead of shaping them from the start.
J-51 is a multi-billion-dollar market opening. We saw it coming earlier than anyone. Now everyone else is waking up to it too — and shame on us if we're not ready when it hits.
So this is not a call for more study. It is the opposite: find the fewest decisions and experiments that matter, and be running them by year end — refining against real deals as the window opens, not still arguing frameworks. Everything here favors tests we can start now, on buildings already in front of us, that pay off in weeks.
The question that matters: what can we have in-market and generating signal by December 2026, so that when the J-51 floodgates open we are sharpening a working motion instead of inventing one?
2. The Delivery Models #
The four canonical delivery methods (Construction Management Association of America, Owner's Guide), ordered by how much design control the owner keeps and how much risk shifts to the builder — not a ranking. The question for each is what role Cadence would play in it.
Design-Bid-Build (DBB). Owner hires a designer, then separately bids the finished design to contractors. Owner holds design risk; sequential and slower; prone to disputes when field conditions differ from the drawings. — Does DBB fit sub-$20k/DU retrofits at all, or is its soft cost and speed exactly the problem we exist to fix?
Construction Management at Risk (CMAR). A construction manager advises during design, then converts to an at-risk builder at a guaranteed price (often a GMP). Owner still holds a separate designer. — Would we ever be the at-risk party — and when is that attractive, financeable, insurable, and operable for us?
Design-Build (DB), incl. Bridging. One entity owns both design and construction, so design risk shifts to it. In the Bridging variant, the owner keeps its own consultant to write the requirements and performance spec, and the design-builder finishes the design and builds. — If our performance spec is the bridging document, does that make us the owner's consultant, the designer-of-record, or something in between — and how much risk comes with each?
Integrated Project Delivery (IPD). Owner, designer, and builder share risk and reward under one multiparty contract. — High-trust and largely legally untested; is it even relevant to us near-term?
Internal docs (inputs, not consensus) #
3. The Owner Risk Stack #
Whatever model we pick, its job is to take these off the owner's plate:
4. Open Questions #
Identity, language, and risk
What owners want vs. what we can do — kept separate
Scope and gate
Market and partners
Economics and product
5. Experiments — Ranked by Speed to Signal #
The test for every one: can it produce signal before J-51 opens? The "start now" set should be assigned Tuesday, not at the retreat.
Start now — weeks, no dependencies #
Owner interviews. A handful of owners across type: which risks block their yes, what they'd pay to remove them, whether fee-crediting and J-51 timing move them, which message lands. → what owners actually buy.
OEM term-sheet requests. Gradient, Innova, Midea: separate and stack volume, lead time, and prepayment, plus warranty, cancellation, substitution, freight, milestones, price-lock. → the real procurement spread + a price-book template.
Contractor interviews. What they need to price firm; what they'll unit-price, cap, or exclude; warranty handling; whether they'll work under our procurement and governance; whether steady pipeline lowers price. → a contractor risk matrix.
Turnkey reality check. We've assumed "turnkey design-build doesn't scale" — but that's unproven. How do today's turnkey players (Carlton, Nova1, VRF Solutions, BES) actually scale or fail: geography, labor, capital, or demand? → facts to replace the assumption.
This quarter — needs a live project or build time #
500-unit quoteability sprint (next week's visit). What we can collect fast, what stays unknown, and classifying every risk as fixed / unit-price / allowance / exclusion / trigger — separating equipment cost from balance-of-system risk. → a draft quoteability report and first unit-price schedule with true-up rules.
Cross-OEM selection engine (v0). All-in comparison (equipment + install + balance-of-system + performance data) across all three OEMs for the 500-unit complex. → proof we can pick neutrally in a way an owner can't do alone.
"Buy" definitions + reservation prototype. Sketch the "buy through Momentum" versions side by side with their risk levels; for the reservation, draft a term sheet — what's verified first, who holds funds, cancellation, site-change handling, freight/damage, reseller exposure. → clear buy definitions + a mock reservation term sheet.
Legal / insurance / licensing review. Where advisory becomes CM becomes contractor; whether we can earn an equipment spread without being a reseller; insurance per model; language to avoid; a contract that starts risk-free with a later GMP toggle. → a boundary memo + draft contract skeletons.
Board narrative test. A board package for one building: pathway, rejected options, what it physically looks like, "why now / why not wait / why this contractor / what after approval" — and whether someone other than Marc can deliver it (the founder-heroics test). → a narrative prototype.
Delivery-model economics. One simple comparison across the models: revenue, margin, labor, risk, working capital, data, repeatability, scale limits. → a table that separates a real business from one that only looks good unpriced.
Later — Q4, once early signal is in #
Construction-phase role pilot. On one project, test a paid post-award role — access planning, submittal and change-order review, requisition review, quantity reconciliation, closeout, warranty handoff. → directly tests the Article 321 failure: do owners value it, do contractors accept it, can we do it profitably?
6. What a Good Retreat Produces #
Not a finished strategy — a way to decide, on the J-51 clock:
The question isn't "should Cadence become a market maker?" It's: what must we learn, build, prove, and risk before we can credibly sell outcomes — and how much of it can we test before the end of the year, when the J-51 floodgates really open?