Money Does Not Buy an Input That Does Not Respond to Money

Ask one question before the term sheet: does my binding constraint have a price?

If it does, raise. Servers, ad inventory, inventory finance, contract manufacturing, junior headcount: all of it arrives faster when you pay more. If it does not, the money lands somewhere other than capacity. It becomes salary, or a place further back in a queue you do not control, or a rival’s new cost base.

Almost nobody runs that test. The plan says the round removes the ceiling, the ceiling stays where it was, and eighteen months later the deck explains it as a hiring problem.

Why this ensemble

The Wire Model scores the features of a decision, routes to a small ensemble of formal models, then forces dated actions out of them. What scored:

• Price elasticity of the binding input (0.9). How many extra units arrive if you pay thirty percent more.
• Statutory or licensed quantity (0.8). Some inputs are rationed by a rule, and a rule does not read your cap table.
• Transit time from payment to unit (0.85). Training, licensure and approval have durations money shortens weakly or not at all.
• Relational dependence (0.7). Agents, distributors, regulators and reference customers commit on accumulated evidence, not on price.
• Recipe substitutability (0.6). Whether the constraint can be designed out instead of bought out.
• Activity bias (0.45). Real, and too low to earn a card.

Those route to Solow capital deepening against a fixed factor for where the money goes, base rates and reference class for how long the constraint takes to clear, and threshold dynamics for the constraints that clear by sequence. On the outcome-type map that is equilibrium, cycle-regime and complex: three types, so the three lenses fail in different directions.

Behavioral and governance are folded rather than shipped as cards. The behavioral layer is one bias, a preference for the visible action over the effective one, and naming it adds no lever the first card does not carry. Governance sits in the blind spot.

The framework: three ways a constraint refuses your money

1. The price: with one factor fixed, capital becomes wage

Add capital to a production function while the other factors stay put and each new unit contributes less than the last. The Nobel Committee’s summary of Solow’s model says it plainly: additional capital injections make ever smaller contributions to production, and raising the share of income saved cannot produce a permanent increase in the growth rate.¹ Money into an unchanged recipe buys a step, then flattens.

Goolsbee tested the sharp case. Most government R&D spending is salary for research workers, and the supply of those workers is quite inelastic, so a large share of any increase goes straight into wages rather than research. His estimate: conventional measures of R&D policy effectiveness may be thirty to fifty percent too high, with the effect strongest for specialists such as physicists and aeronautical engineers.²

The last line of his abstract is the one that concerns you. By moving the wages of scientists and engineers even at firms receiving no federal support, the funding crowds out private inventive activity.² The money did not stay inside the company that received it. It repriced the input for the whole market.

Your round is somebody else’s cost increase, and theirs is yours. Two funded companies hiring from the same forty senior backend engineers in Lagos or Nairobi do not produce eighty engineers. They produce a bidding war, two shorter runways and the same forty engineers.

Bloom, Jones, Van Reenen and Webb show where this ends if you refuse to accept it. To keep chip density doubling every two years, the number of researchers required today is more than eighteen times the number required in the early 1970s.³ Effort climbs, productivity per unit of effort falls. That is what buying through a hardening constraint looks like over decades.

Solow capital deepening, the incidence lens

Assumes: at least one factor is fixed over your planning horizon and the recipe stays roughly constant.

Fits because: input elasticity scored 0.9 and the pass-through evidence is direct.

Breaks when: the raise is what changes the recipe. Capital that buys automation or a new architecture is not capital deepening, and this card will underrate it.

Counteracts: reading a funded competitor’s hiring as proof that hiring works.

May reinforce: fatalism about growth, and treating every ceiling as permanent.

2. The clock: some constraints clear on a calendar you do not own

When the input is rationed by a rule, the useful question stops being how much and becomes how long. Answer it the way an actuary would, from the record of everyone who has stood in the same queue.

Kenya’s digital lending queue is a clean reference class. The Central Bank of Kenya has received more than 800 applications for digital credit provider licences since March 2022, and as of 14 April 2026 had licensed 227 of them.⁴ Four years, roughly a quarter cleared. No applicant’s round moved their position, and no applicant’s round could have.

The United States runs a harder version. The H-1B cap is 65,000, with 20,000 more reserved for holders of American advanced degrees. Congress sets both, and both were reached again for fiscal year 2027.⁵ A company with a billion dollars and a company with a seed round meet the same ceiling.

Property gives the longest series. American house prices have risen roughly two percent a year since 1950, but the reason changed. Before 1970 the rise tracked construction costs and quality. After 1970 it tracked the growing difficulty of getting regulatory approval to build, with zoning and land use controls the dominant cause of expensive housing.^{6, 7} Half a century of capital chasing a permitting constraint produced prices, not houses.

Training pipelines are the version African founders meet first. The IFC projects 230 million digital jobs in Sub-Saharan Africa by 2030, translating into nearly 650 million training opportunities including retraining.⁸ You can fund a place in a pipeline. You cannot compress it below the length of the course.

So build your own reference class before you build the hiring plan. Find five companies that hit your exact constraint, ask each how many months passed between deciding and having, take the median. That number is the schedule you are on.

Base rates and reference class, the clearing-time lens

Assumes: your constraint belongs to a class with a visible history and no rule change since.

Fits because: statutory rationing scored 0.8 and transit time scored 0.85.

Breaks when: the queue is discretionary rather than procedural. Where a regulator has latitude, preparation and relationships genuinely move you, and a base rate hides that.

Counteracts: plans built from the date you want to launch backwards.

May reinforce: waiting, when the honest move is to change the recipe and leave the queue.

3. The sequence: relational inputs clear by threshold, not by budget

Granovetter’s threshold model covers the constraints that look social and behave mathematically. Each actor has a number: the proportion of others who must have committed before they will. Feed in the distribution of those numbers and the aggregate outcome falls out.

His demonstration still lands. Take 100 people, thresholds spread evenly from 0 to 99, and everyone eventually joins. Now replace the person whose threshold is 1 with a second person whose threshold is 2. The two crowds are indistinguishable on any average you could measure. The second produces one actor and stops, because the difference lies entirely in the aggregation and in one gap in the distribution.⁹

That is your agent network, your distributor list, your first three enterprise references and the regulator’s private confidence in you. Each party is waiting on evidence about who else has committed. Money pays for a signature. It cannot supply the evidence, because the evidence is other people’s behaviour.

The mobile money industry has the numbers. In 2024 there were 28 million registered mobile money agents worldwide, twenty percent more than the year before. Ten million of them were active in any month.¹⁰ Eighteen million signed agents who do not trade is the exact size of what money buys and trust does not.

Safaricom understood this while building the network everyone now cites. The IFC case study records that agent training and management, not agent count, was the binding lesson of the M-PESA pilot, and that the company deliberately avoided flooding the market with agents whose profitability could not be sustained.¹¹ They were managing a threshold distribution, not a purchase order.

The lever this gives you is sequence. Find the lowest-threshold party in your market, the one who will commit on the least evidence, and spend your effort there first. Signing the largest distributor before anyone else has moved is the most expensive way to learn what a threshold is.

Threshold dynamics, the sequence lens

Assumes: parties observe each other, and each commits only after enough others have.

Fits because: relational dependence scored 0.7 across agents, distributors, references and regulators.

Breaks when: the parties are isolated from one another. With no observation there is no cascade, and paying a premium works exactly as advertised.

Counteracts: ranking prospects by size instead of by threshold.

May reinforce: patience past the point where the distribution itself is the problem.

The price-response test

Run this on the constraint you would name if a partner asked what is holding you back. Five questions, answered in writing.

1. The thirty percent question. If you paid thirty percent above market for this input starting tomorrow, how many additional units arrive within ninety days? If the honest answer is the same number at a higher price, elasticity is near zero and the round is a wage increase.
2. The queue question. Does a third party control the throughput, and is their throughput independent of your budget? Then you are buying position, at best.
3. The clock question. What is the shortest path from money to a usable unit, measured in months rather than in effort? A course, a licence period and an approval cycle all have floors.
4. The evidence question. Does someone have to watch you for a period before committing? Then the currency is track record and sequence, not price.
5. The recipe question. Can you redesign the product so this input stops being required? This is the only question whose yes converts an inelastic constraint into an elastic one.

Two or more zeros in questions one to four, and the raise does not do what your plan says it does. That is not a reason to skip the round. It is a reason to change what the round is for.

GEER: the levers, once you know the elasticity

Five channels carry an inelastic constraint: the recipe, demand triage, geography, the pipeline, price. Pull the reversible ones first, and notice where price sits.

1. Re-cut the work. Audit where the scarce input spends its hours and strip out everything that does not need it. Recovering thirty percent of a senior engineer’s week is a hire you did not have to win.
2. Triage demand. Rank customers by scarce-input hours per unit of gross margin, then stop selling to the bottom of that list. Uncomfortable, reversible, immediate.
3. Re-price. If the constraint is real, your price is too low. Raising it rations demand and funds the constraint at once.
4. Buy calendar early. File the application, enter the training programme, start the licence clock before the plan depends on it. Queue position is cheap in advance and unbuyable late.
5. Move the input geographically. Remote seniority, a second engineering base, a distributor in an adjacent corridor. Costs coordination, and it is the fastest real capacity increase available.
6. Substitute the recipe. Re-architect so the constraint stops binding. Slow, expensive, and the only lever that changes elasticity itself.
7. Build the pipeline. Train, certify, licence, grow your own. Justified only when the constraint is permanent and the input is core.
8. Pay above market. Last, deliberately. It moves price reliably, quantity rarely, and it teaches competitors your new number.

No-lever flag: if levers one through six all return nothing, your company sits inside a market whose capacity has to grow before yours can. That is a market-development problem wearing a fundraising costume. Write it down in those words and re-plan the year around a smaller footprint rather than a bigger round.

RADAR: what to settle before the round closes

DO NOW, by T+3 days. Reversible and worth doing under every scenario.

1. Name the single binding constraint in one sentence. If two people in the company name different ones, that disagreement is the first finding.
2. Run the price-response test on it, in writing, with numbers rather than adjectives.
3. Rewrite the use of funds so each line states which constraint it relieves and what that constraint’s elasticity is.

HEDGE, by T+14. Cheap protection against the ceiling not moving.

1. File every application whose clock you might need within eighteen months, including the ones you hope to avoid.
2. Build the five-company reference class for your constraint and record the median months from decision to capacity.
3. Run one substitution experiment: a single workflow rebuilt so the scarce input is not required. Small scope, real measurement.
4. Identify the lowest-threshold partner in your relational network and secure them before approaching the largest.

DEFER AND TRIGGER. Irreversible, so hold and pre-commit the observable that releases the spend.

1. Defer: a step change in compensation bands, a second office, an in-house training academy.
2. Trigger to spend: the substitution experiment fails on measurement, and the reference-class median exceeds your runway. That combination means the constraint is both real and permanent, and building the pipeline becomes the cheapest remaining option.
3. Counter-trigger: by T+28, if the test returned two or more zeros and the plan still promises capacity from the raise, cut the target size and lengthen the timeline instead. Raising more against a fixed input buys inflation and dilution together.

Underwriting this from the other side. DO NOW: for each growth plan, ask which constraint the money relieves and demand the elasticity estimate rather than the hiring plan. HEDGE: price wage inflation explicitly in any category where you have funded two companies competing for the same input, since you are on both sides of that bid. DEFER: writing the larger cheque until the founder shows a substitution path, because a bigger round into a fixed factor raises the entry price of your own next round.

CHAIN: what has happened to companies standing here before

Three histories share this structure while sharing none of its surface: public R&D money into a fixed pool of researchers, housing demand into a permitting queue, mobile money capital into an agent network built on trust. In all three the money arrived, the quantity barely moved, and the price of the scarce input rose.^{2, 6, 10} That is the base rate, and it is unkind.

Present conditions bend it slightly. African tech funding rose about twenty-five percent in 2025 to just over four billion dollars, but equity grew eight percent and deal count moved one percent.¹² More money, the same number of companies. Even the capital market has an input that does not respond to capital.

Now remove what would have happened anyway. Part of the wage rise your round appears to have caused was already running, driven by remote hiring, currency moves and one large employer entering your city. Do not credit your round with a market you merely joined. The share you caused is the share to plan against, and it is smaller than it feels.

Matrix-break flag. Elasticity is not a property of the input. It is a property of the input in a given year. Work that was inelastic because only senior people could do it turns elastic the moment a tool lets a mid-level person do it to the same standard. The constraint you built the plan around stops binding, and something else, usually relational or regulatory, takes its place. Re-run the test every two quarters. Never inherit last year’s answer.

What none of these three can tell you

Three gaps, and they matter.

Whether the constraint you named is the binding one. Every model here assumes you identified it correctly. The loudest constraint is the one with the most internal advocacy, and it is frequently not the one setting the ceiling.

The option value of the money. A round buys time, and time is what lets a queue clear and a threshold tip. These models price capacity and undercount survival, which is worth buying even when the ceiling holds.

Governance under a ceiling. A board that has just funded a growth plan does not enjoy being told the plan assumed elasticity that does not exist. The models are silent on who says that out loud, and when.

So do the part that is not silent. Name the constraint this week, run the five questions in writing, and put the answers in front of whoever is about to fund the plan. You are not under-capitalised. You are supply-constrained. Those two diagnoses take the same money to a different place, and only one of them survives the next eighteen months.

Sources and notes

1. The Royal Swedish Academy of Sciences, press release for the 1987 Prize in Economic Sciences awarded to Robert M. Solow. Source for the statements that, due to diminishing yields, additional capital injections make ever smaller contributions to production, that an increase in the proportion of income saved cannot lead to a permanent increase in the rate of growth, and that only a small proportion of annual growth could be explained by increased inputs of labour and capital. Press release.
2. Goolsbee, A. “Does Government R&D Policy Mainly Benefit Scientists and Engineers?” NBER Working Paper 6532, April 1998; American Economic Review 88(2), 298-302. Labour supply of R&D workers is quite inelastic, so a significant fraction of increased government R&D spending goes directly into higher wages; conventional estimates of R&D policy effectiveness may be 30 to 50 percent too high; and by altering the wages of scientists and engineers even at firms receiving no federal support, government funding directly crowds out private inventive activity. Abstract and working paper.
3. Bloom, N., Jones, C. I., Van Reenen, J., and Webb, M. “Are Ideas Getting Harder to Find?” NBER Working Paper 23782, September 2017. Research effort is rising substantially while research productivity declines sharply; the number of researchers required today to achieve the doubling of chip density every two years is more than 18 times the number required in the early 1970s. Abstract and working paper.
4. Central Bank of Kenya, press release “Licensing of Digital Credit Providers,” 14 April 2026. More than 800 applications received since March 2022; 227 digital credit providers licensed as at that date, with other applicants at different stages of the process. Press release PDF.
5. United States Citizenship and Immigration Services, “H-1B Cap Season.” The congressionally mandated regular cap is 65,000, with a further 20,000 exemption for holders of United States advanced degrees; both were reached for fiscal year 2027. Cap season page.
6. Glaeser, E. L., Gyourko, J., and Saks, R. “Why Have Housing Prices Gone Up?” NBER Working Paper 11129, February 2005; American Economic Review 95(2), 329-333. House prices have risen by almost two percent a year since 1950; between 1950 and 1970 this reflected rising quality and construction costs, and since 1970 it reflects the increasing difficulty of obtaining regulatory approval to build. Abstract and working paper.
7. Glaeser, E. L., and Gyourko, J. “The Impact of Zoning on Housing Affordability.” NBER Working Paper 8835, March 2002; published as “The Impact of Building Restrictions on Housing Affordability,” FRBNY Economic Policy Review 9(2), 2003. Housing price exceeds physical construction cost in only a limited number of areas, and in those areas zoning and other land use controls play the dominant role in making housing expensive. Abstract and working paper.
8. International Finance Corporation, “Digital Skills in Sub-Saharan Africa: Spotlight on Ghana” (2019). Projects 230 million digital jobs in Sub-Saharan Africa by 2030, translating into nearly 650 million training opportunities by 2030 including required retraining. Report PDF.
9. Granovetter, M. “Threshold Models of Collective Behavior.” American Journal of Sociology 83(6), 1978, 1420-1443. Source for the definition of a threshold as the proportion of others who must act first, and for the 100-person example in which a uniform distribution of thresholds from 0 to 99 produces full participation while replacing the individual with threshold 1 by a second individual with threshold 2 leaves a single actor, the difference resulting from aggregation and a gap in the frequency distribution. Text-layer mirror PDF, hosted by the Chinese University of Hong Kong.
10. GSMA, “State of the Industry Report on Mobile Money 2025,” reporting 2024 data. 28 million registered mobile money agents in 2024, 20 percent more than in 2023, of which 10 million were active on a monthly basis; 755 registered agents per 100,000 adults in mobile money countries, double the 2021 ratio. Report PDF.
11. International Finance Corporation, “M-Money Channel Distribution Case: Kenya, Safaricom M-PESA,” IFC mobile money toolkit, World Bank Group. Source for the finding that agent training and management emerged from the pilot as critical to the service, that the in-house training team was too small for national launch, and that Safaricom was careful not to flood the market with agents whose profitability could not be maintained. Case study PDF.
12. Partech, 2025 Africa Tech Venture Capital Report. Total equity and debt funding rose around 25 percent year on year to just over US$4B; equity funding grew 8 percent while equity deal count moved 1 percent. Report page.