Haute Lumière

Commerce · I.10 · MMXXVI · daylight

La Bourse  /  Volume I  /  Nº I.10

Scaling Without Losing the Thing

Volume I — Transition: From Here to the Living Economy

---

THE PLATE

---

THE LETTER

You have something that works. That is the position this chapter assumes, and it is a good one to be in. One shop, one clinic, one crew, one fund, one farm, one team — a unit of something you built that does what you hoped it would do, and now people are asking whether there can be more of it.

The question is not whether to grow. You already want to; that is why you are reading. The question is the one nobody asks early enough, and it is this: which of the things that make this work will survive being copied, and which of them are only working because you are standing here?

Most growth advice answers a different question. It tells you how to add units. Adding units is the easy half, and there are four well-understood ways to do it — build them yourself, franchise them, federate them, or licence the name and let go. Each one is a real instrument with a real body of law, real accounting and real precedent, and each one does something different and permanent to the thing being scaled.

What this chapter does is put numbers on that. Not on how fast you can grow, but on where quality per unit breaks, and at what size coordination stops paying for itself. Both of those have an arithmetic, both can be computed from figures you already have, and the second one has an answer that is smaller than almost anyone expects and a fix that is not the one people reach for.

The good news arrives early and it is worth holding onto: the organisations that have scaled the furthest without hollowing out are not the fastest ones. They grew at rates that look almost embarrassing on a slide. Mondragón added cooperatives at under seven percent a year for sixty-seven years. That is the whole trick, and it is available to you on Monday.

— The Editors

---

DISCOVERY

What is already working

Four families of organisation have solved this, each in a different way, and all four are open to inspection.

Mondragón, Basque country, 1956 to now. Five graduates of a technical school in Arrasate started one cooperative making paraffin heaters. Sixty-seven years later the federation holds on the order of ninety-five cooperatives and something near seventy thousand people. Run the compound rate on the unit count and it is 6.8 percent a year — roughly one new cooperative for every fifteen already standing, every year, for six and a half decades. Nobody sprinted.

What actually scaled was not the heater business. It was a financing organ: Caja Laboral Popular, founded in 1959, three years in, before there was anything much to finance. The bank held the deposits of the members and the Empresarial Division inside it did the thing that makes the federation work — it incubated, staffed and supervised each new cooperative, and it could see across all of them at once. The unit that replicated was not the factory. It was the relationship between a new cooperative and an institution that had already done this forty times.

Credit union leagues, United States, 1934 to now. Individual credit unions are small and legally independent; there is no head office. What they built instead was a two-tier structure: state leagues for representation and training, corporate credit unions for liquidity and payments, and later CUSOs — credit union service organisations — jointly owned companies that run the things no single small institution can afford to run alone. Card processing, mortgage servicing, compliance, core banking. A two-hundred-member credit union gets the back office of a large bank and keeps its own board.

Community land trusts. The Burlington Community Land Trust, founded in 1984 with municipal funding and now the Champlain Housing Trust, holds roughly three thousand homes and apartments. The mechanism is a resale formula: the household owns the building and leases the land, and on sale keeps a stated share of the appreciation — twenty-five percent is the common figure. The remainder stays in the home. What replicated across some two hundred-odd trusts in the United States was not a housing development. It was a clause. A ground lease with a resale formula in it is about forty pages, it is free, and it carries the entire mechanism.

Franchising. Set aside the aesthetics and look at the structure, because it solved a genuine problem that neither ownership nor licensing solves. Paul Rubin's 1978 analysis is still the clearest statement of it: the franchise contract exists because the local operator has knowledge and effort that head office cannot observe, and the only reliable way to buy unobservable effort is to make the person who supplies it the residual claimant on the unit's profit. The operator keeps what is left. That is why franchised units are so often better run than company-owned ones in the same chain, and it is a real finding about human beings, not a piece of ideology.

Four families, four mechanisms, one shared move: in every case what scaled was not the product. It was a small, portable piece of structure — a bank, a shared back office, a lease clause, a residual claim — that made the next unit possible without making it dependent.

Notice what none of them scaled. None of them scaled a founder. None of them scaled a culture by describing it. And none of them went fast.

---

THE ARITHMETIC

What works, what does not, and where the line sits

Three calculations. The first tells you what a licence costs. The second tells you how many hands the thing can pass through. The third is the one that decides everything, and it has an uncomfortable answer.

One. The franchise quality wedge.

A franchise takes its royalty on gross revenue, not profit — typically six percent, plus a marketing fund of around two. Take a unit turning over £900,000 at a twelve percent operating margin before royalty:

  unit revenue                          900,000
  operating profit before royalty       108,000   (12%)
  royalty 6% + ad fund 2%                72,000   (8% of gross)
  royalty as a share of operating profit    66.7%

Two-thirds of the unit's operating profit, and it is indexed to the wrong line. Now watch what that does to quality. The operator considers spending £20,000 on something that makes the thing better — a better ingredient, a slower process, another pair of hands on a Friday — and it lifts revenue two percent:

  incremental revenue                    18,000
  franchisee keeps 92% of it, less spend  -3,440   a loss
  franchisor takes 8% of it, at no cost   +1,440   a gain
  franchisee break-even lift               2.42%   of revenue

The franchisor is ahead from the first pound. The operator is not ahead until the lift clears 2.4 percent. Between those two lines sits every quality decision in the chain, and the contract has quietly told the only person who can make it not to. That gap is not a flaw in franchising. It is franchising. You buy speed with it, and the price is paid in exactly the currency this chapter is about.

Two. Fidelity per hop.

A thing is taught by a person to a person. Call the fraction that survives one teaching hop φ. Be generous and set φ = 0.90 — nine-tenths transmitted, which is better than most organisations achieve.

  hop 0   1.000        hop 4   0.656   <- below the floor
  hop 1   0.900        hop 5   0.590
  hop 2   0.810        hop 6   0.531
  hop 3   0.729        hop 7   0.478

Half the thing is gone in seven hops — ln(0.5) / ln(0.9) = 6.58. Set a floor at 0.70, below which a unit is no longer recognisably the thing, and you get the operating rule: three hops, then retrain from source.

That rule has a price, and it is worth stating plainly. Let every trained unit train four more a year. After five years the cascade holds 3,125 units — and 125 of them, four percent, are at or above the floor. Source-training only, with central capacity for four a year, gives you 21 units, all of them good. The cascade produces six times as many good units and three thousand bad ones wearing your name. Both halves of that sentence are true and the second one is the one that closes companies.

Three. The coordination ceiling — and this is the honest negative.

A federation of peers returns a shared benefit per unit that saturates: joint purchasing, shared back office, risk pooling, a common brand. Most of it is captured early. And it carries a coordination cost that does not saturate, because a federation of peers is a mesh — n units hold n(n−1)/2 relationships, so each unit holds (n−1)/2 of them.

  B(n) = 120,000 · (1 − e^(−n/12))      shared benefit per unit, GBP/yr
  C(n) = 2,800 · (n − 1) / 2            coordination cost per unit, GBP/yr

  n = 12   benefit  75,854   coord  15,400   net  +60,454
  n = 24   benefit 103,760   coord  32,200   net  +71,560
  n = 60   benefit 119,191   coord  82,600   net  +36,591
  n = 86   benefit 119,907   coord 119,000   net     +907
  n = 87   benefit 119,915   coord 120,400   net     −485

The optimum is n* = k·ln(2B/kc) = 23.6 — twenty-four units. And above eighty-seven, a federation in which every member deals directly with every other member costs its members more than it returns to them. Not slows down. Costs more. That number is smaller than nearly every federation's ambition, and it is why so many associations of peers feel, past a certain size, like a tax their members pay for a newsletter.

Four. The move.

Every instinct at this point is to ask what the right size is. That is the wrong question, and the arithmetic says so. Cluster the units — groups of about √n, each group meshing internally, the groups meshing with each other — and change nothing else. Same benefit, same cost per relationship, same people:

  n = 86     groups of 9    coord  13,186   net +106,721
  n = 900    groups of 30   coord  41,953   net  +78,047
  n = 2,500  groups of 50   coord  69,972   net  +50,028
  two-tier net turns negative at  n = 7,141

The ceiling moves from 87 to 7,141 — eighty-two times — and not one thing about the units changed. The ceiling was never a size. It was a shape.

And then the part that has already been paid for. The tier you add to break the ceiling is the tier that breaks. American credit unions built exactly this structure — small independent institutions clustered under corporate credit unions that held their liquidity — and in 2009 that second tier failed, five corporates were placed into conservatorship, and the cost was assessed back across the very institutions the tier existed to protect. The mesh could not scale. The hub could, and it concentrated into one point everything the mesh had kept apart.

So the honest sentence is both halves at once: you can raise the ceiling by a factor of eighty by changing shape, and each shape you add is a new single point of failure that must be capitalised, governed and stress-tested as such. Mondragón knows this; Caja Laboral was supervised as a bank from the beginning. And Mondragón carries the other half of the honest ledger too: the great majority of its plants outside Spain are ordinary subsidiaries with ordinary employees. The federation scaled magnificently to seventy thousand people and did not cross the border as itself.

---

DREAM

What becomes ordinary

In the version of this that has already happened, an organisation knows its own ceiling and has written it down.

There is a number in the constitution. Not a cap on ambition — a threshold, with a rule attached: at twenty-four members we cluster; at eighty-seven we stop admitting until we have. It is reviewed each year against measured figures, the way a covenant is reviewed, because B and k and c are all things a federation can measure about itself once somebody decides to. Nobody experiences this as a limit. They experience it as the reason the meetings are still worth attending.

Fidelity is a measured quantity. A new unit is assessed against the source, not against the unit that trained it, and the assessment is not an audit — it is the same assessment the source runs on itself, which is why nobody minds it. Units know their hop count the way a piece of equipment knows its service interval. When a unit reaches three hops, somebody comes and works alongside them for a week, and that week is in the budget as a line called transmission, sitting where depreciation sits, because it is the same idea.

Growth rates are chosen rather than pursued. A board can say we could open forty and we are opening nine, and give the arithmetic, and be heard. The question in the room is never how fast — it is how many hops from the source is the fortieth unit, and who is going to go and stand in it.

The four instruments are held as instruments, not identities. An organisation franchises the part of itself that is a system, federates the part that is a profession, licenses the part that is a standard, and owns outright the part that is still being invented — and it can say which is which. Nobody argues about whether franchising is good. They argue about which of the four this particular thing is, which is a much better argument and one that can be settled.

And the unit that is forty kilometres away and three hops from the source is doing it right, on a Tuesday, with nobody watching. Which was the whole point and turns out to have been an engineering problem.

---

DESIGN

The structure that gets there

First: separate the four things that want to travel.

Take your thing and sort it into four boxes, because they scale by different instruments and mixing them is the most common error in this chapter.

The diagnostic is one question: can a stranger tell, from the outside, whether it was done right? If yes, it is a system or a standard and it can travel by contract. If no, it is a profession or an invention, and it travels only by ownership or by presence — which is to say by federation or by you being there.

Most things people try to franchise are professions, and most things people try to federate are systems. Both errors are expensive and both are visible in the first four units if you look.

Second: build the organ before you need it.

Caja Laboral existed three years in. The CUSO exists before the credit union needs the card processing. The ground lease is written before the second house. The portable piece of structure is built at unit three, not unit thirty, because at unit three you can still see all of it, and because building it later means building it while it is load-bearing.

What is that organ, concretely? It is whichever of these your units cannot afford alone and will not build separately: the financing, the training, the measurement, the standard, the shared back office. Pick the one that is already costing each unit the most and build that one first.

Third: cap the hops, not the growth.

Write the fidelity rule into how units are created. Every new unit is trained by a unit no more than two hops from the source. A unit at three hops trains nobody until it has been retrained. This sounds restrictive and is not: it caps depth, not breadth. A source that can train four units, each of which trains four, each of which trains four, gives you eighty-five units all within the floor — and that is more units than your coordination arithmetic will let you hold in one mesh anyway.

Fourth: cluster before you have to.

At n*, cluster. Not at the ceiling — at the optimum, which is about a quarter of the ceiling, because clustering while things are working is a design decision and clustering at the ceiling is a rescue. Groups of about the square root of your target size. Each group meshes internally and sends one person to the between- groups mesh, and that person's job is explicitly translation, not representation.

Fifth: capitalise the tier.

Whatever you build as the second tier — the shared services company, the liquidity pool, the training academy — is now systemically important to everyone who depends on it. Give it its own balance sheet, its own board, its own capital buffer sized against the loss it could impose on the units, and an annual test that asks one question: if this fails, what happens to the smallest member?

---

DESTINY

How it holds when nobody is pushing

Three things hold a scaled thing together, and only three.

The organ has its own income. A shared services company funded by a levy that must be renewed every year is a shared services company that spends its winters lobbying its own members. Fund it from what it does — a fee per transaction, a share of the saving it produces, a patronage rebate of any surplus. It should be able to survive the indifference of its members for one bad year, and not for five.

Exit is cheap and nobody uses it. A member who cannot leave is a hostage, and hostages do not contribute. Write the exit: notice period, what they take, what they leave, and above all their data leaves with them. A federation that is held together by switching costs has already lost the thing it was federating.

Somebody visits. Not audits — visits. The single most reliable transmission mechanism in every one of the four families above is a person from the source spending a week inside a unit doing the work. It is expensive, it does not scale, and it is the reason any of it scales.

Now the failure modes, named, because they are predictable.

It fails when the second tier becomes the principal and the units become branches — which happens gradually, by way of a series of individually reasonable standardisations, and is visible only in the minutes. It fails when growth is measured in units instead of units-above-the-floor, because the two numbers diverge quietly and the divergence is invisible until a customer finds it. It fails when the federation admits past its own arithmetic and the meetings stop being worth the flight. And it fails, most often, when the founder is still the transmission mechanism at unit forty — at which point the organisation has not scaled at all. It has simply spread one person thinner and called the resulting tiredness growth.

---

DELIGHT

What it feels like

You walk into the fortieth one unannounced.

Nobody knows you are coming. The lights are the same and they are not the same; someone here made a different decision about the lamp over the counter and it is better. The greeting is not the script. It is warmer than the script, and it is unmistakably the thing.

There is a particular quiet pleasure in this that people who have built one unit never get to have. It is not pride of ownership — it is something closer to the feeling of hearing a piece of music you wrote played by somebody who has made it theirs. They are not doing what you told them. They are doing what you meant, which you were never able to say properly, and somehow it got here anyway across three hops and forty kilometres.

And then the better part, on the way out: you realise you could not have done this shift. They are better at it than you are now. The thing has outgrown its source, which is the only definition of scale that has ever been worth having.

---

OPERATIONALIZE THIS

At the level of finance

The instrument: a member-owned shared services company with a capped levy, a patronage rebate, and a written ceiling clause.

This is the CUSO in American credit unions, the secondary cooperative in Mondragón's structure, the consorzio in Italian cooperative law. It is the portable organ, and it is ordinary company law.

The structure.

• Ownership. One share per member unit, one vote per member, irrespective of size. Shares issued at par and redeemable at par on exit. This is what stops the organ becoming a holding company by accretion.
• Funding. A levy, capped in the constitution as a percentage of measured member benefit — not as a percentage of member revenue. A levy indexed to revenue grows whether or not the organ is working; a levy indexed to measured benefit cannot.
• Surplus. Rebated pro rata to patronage, annually. In the United States this is Subchapter T territory and the rebate is deductible to the company; in the UK and most of Europe a cooperative or a company limited by guarantee with a patronage clause achieves the same. Ask your auditor in the first month, not the eleventh.
• Capital. A buffer sized against the loss the organ could impose on its smallest member, not its average one. Test it annually and publish the test.
• Exit. Ninety days' notice, shares redeemed at par, data returned in a documented format. Written before the first member joins.
• The ceiling clause. The constitution states the n at which the federation must cluster and the n at which it stops admitting, both computed from measured B, k and c, both reviewed annually by the board with the workings attached.

Balance-sheet treatment. The member unit holds its share as an investment at cost — immaterial by design. The levy is operating expense. The patronage rebate is income in the year received. The shared services company capitalises the platforms it builds and depreciates them over their useful life, and here is the one to watch: a shared platform's useful life is set by the federation's commitment horizon, not the technology's. If members can leave on ninety days' notice and the platform is depreciated over seven years, say so in the notes.

The counterparty. The members, first and for a long time. Take no external debt into the organ until it has two full years of audited surplus, because the first external lender will want security, and the only security the organ has is the members' obligation to keep using it — which is precisely the thing you promised not to create.

The number that decides it. One ratio, on the front page of every annual report the organ produces:

              measured benefit per member unit
   -----------------------------------------------------  >  3.0
    levy  +  (coordination hours x loaded hourly cost)

Worked, from a real shape of numbers: benefit £86,000, levy £24,000, and 180 hours a year of coordination at £75 an hour — £13,500.

   86,000 / (24,000 + 13,500)  =  2.29 : 1     does not clear

That federation is not failing. It is the wrong shape. Cluster it, so each member holds nine relationships rather than eighty-five, and the coordination falls to 60 hours:

   86,000 / (24,000 +  4,500)  =  3.02 : 1     clears

Same benefit. Same levy. Same people. The only thing that changed was how many relationships each member had to hold in their own hands.

The first ninety days.

---

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

1. Which of our units is furthest from where this started and still unmistakably the thing? What did we do right in that case that we did not do deliberately?
2. Think of a time somebody in a distant unit made a decision we would not have made and it was better. What let them make it?
3. What do we share across units now that none of us could afford alone — and who built it, and when?

Dream — what becomes possible

1. If we knew our own ceiling and had written it into the constitution, what would we stop arguing about?
2. Imagine a new unit opening next year that we never visit and never worry about. What would have had to be true about how it was created?
3. If growth were measured in units-above-the-floor rather than units, what would we be proud to report three years from now?

Design — what we build

1. Which parts of what we do could a stranger judge from the outside, and which could only be judged from inside? What does that tell us about which instrument each part wants?
2. What is the one thing our units cannot afford alone that we have not yet built together — and what would the first version of it cost?
3. Who here has been taught directly by the source, and who has been taught by someone who was? What would it take to know that for every unit?

Destiny — how it holds

1. If the shared organ failed tomorrow, what would happen to our smallest member — and is that acceptable to the rest of us?
2. What would make leaving this federation genuinely easy, and what are we afraid would happen if it were?
3. Who visits? Not inspects — visits. How often, and what did the last visit change?

---

WORKS CITED

Brooks, F. P. (1975). The Mythical Man-Month: Essays on Software Engineering. Addison-Wesley.

Coase, R. H. (1937). "The Nature of the Firm." Economica, 4(16), 386–405.

Cooperrider, D. L. and Whitney, D. (2005). Appreciative Inquiry: A Positive Revolution in Change. Berrett-Koehler.

Davis, J. E. and Stokes, A. (2009). Lands in Trust, Homes That Last: A Performance Evaluation of the Champlain Housing Trust. Burlington Associates in Community Development.

Davis, J. E. (ed.) (2010). The Community Land Trust Reader. Lincoln Institute of Land Policy.

Dunbar, R. I. M. (1992). "Neocortex size as a constraint on group size in primates." Journal of Human Evolution, 22(6), 469–493.

Errasti, A. M., Heras, I., Bakaikoa, B. and Elgoibar, P. (2003). "The Internationalisation of Cooperatives: The Case of the Mondragon Cooperative Corporation." Annals of Public and Cooperative Economics, 74(4), 553–584.

Klein, B. (1980). "Transaction Cost Determinants of 'Unfair' Contractual Arrangements." American Economic Review, 70(2), 356–362.

National Credit Union Administration. Annual Reports and Credit Union Profile statistics. Successive years.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Rubin, P. H. (1978). "The Theory of the Firm and the Structure of the Franchise Contract." Journal of Law and Economics, 21(1), 223–233.

Schumacher, E. F. (1973). Small Is Beautiful: A Study of Economics as if People Mattered. Blond & Briggs.

Whyte, W. F. and Whyte, K. K. (1991). Making Mondragón: The Growth and Dynamics of the Worker Cooperative Complex, 2nd edn. ILR Press.

Williamson, O. E. (1985). The Economic Institutions of Capitalism. Free Press.

Note on figures. The saturating-benefit and mesh-coordination model, the two-tier comparison, the fidelity decay, the cascade count, the franchise wedge, the CLT subsidy retention and the coverage ratio are all computed in lib/verify/I_10.py and reproducible there with their parameters stated. Mondragón unit and headcount figures are from the corporation's own annual reporting and Whyte and Whyte (1991); credit union counts and membership are from NCUA statistics and are given as approximations. Franchise royalty ranges are the common ranges disclosed in Item 6 of a Franchise Disclosure Document.