The easiest thing to see about an accelerator is the cheque, and it is the least interesting thing about it. A cheque is fungible. If capital were the whole of what a program changed, a founder could raise the same amount from an angel, skip the six months of scheduled obligation, and be no worse off. So the question that actually decides whether a program earned its place in the story is a counterfactual one: hold the money constant, take it out of frame, and ask what six months inside a cohort changes about a startup that the identical money outside a cohort would not. This is a note on our own answer to that question, drawn from the program we went through while building the raster-log models behind VeerNet, and it is a deliberately narrow claim. It is not about which thresholds a program gates investor-readiness behind, and it is not about the equity we traded a venture partner for the engineering capability itself. It is about the part of the accelerator that the cheque does not buy: the cohort, the network it points into, and the cadence that turns that reach into something a founder can actually use.
Subtract the money and something is still left
Start with the subtraction, because it is the whole method. Strip the capital out of a program and you are left with two things that money does not directly supply. The first is reach: the network the cohort belongs to, which for our program was a base of more than 1,500 startups supported over its tenure, a team of more than 120 people, and a footprint across more than 50 countries. The second is a delivery schedule: more than 100 hours of interactive curriculum, up to 12 hours of mentorship, and six monthly mastermind sessions of two hours each across a six-month program. The first is an asset. The second is a machine for spending it. The counterfactual lift, the thing a founder could not have bought with the same money in isolation, is the reach, and the reason it is not just a directory you could scrape is the cadence that keeps making you use it.
This distinction is not our invention. The founding account of the accelerator as a category, from Cohen and Hochberg, defines it precisely by these two features rather than by the capital, which incubators and angels also provide: a cohort of peers moving through a fixed, time-limited window together, and an intense, scheduled dose of mentorship [1]. The capital is the entry ticket; the cohort and the schedule are the product. Later work made the mechanism sharper. Hallen, Cohen, and Bingham asked directly whether accelerators work and, if so, through what, and isolated the program's schedule of forced, broad, intense interaction as a mechanism that changes venture behaviour independently of who the program selected or how much money it attached [2]. That is exactly the claim the subtraction makes tangible: the schedule is doing work the cheque cannot.
Reach is latent until a cadence spends it
The trap in celebrating the network is treating the reach as if it were the lift. It is not. A base of 1,500 startups and a presence in 50 countries is a stock of latent access, and latent access changes nothing about a startup on its own. Every founder who has ever collected a stack of business cards at a conference and never emailed one of them knows the failure mode. Reach converts to lift only when something forces the conversion, repeatedly, on a schedule the founder did not set and cannot quietly skip. That forcing function is the cadence: the curriculum you have to complete, the mentor hours you have to book, the six masterminds where you sit in a room with a rotating set of other founders and have to say out loud what is not working. The mastermind in particular is the cohort made concrete. It is not content delivered at you; it is the peer network delivered on a clock.
The exhibit below is that argument made into a single reading. It splits the program's lift into two stacked contributions and lets you run the counterfactual by hand.
Drag the cadence lever to zero and the network contribution collapses, leaving only the money-and-access floor: the counterfactual where you took the cheque and skipped the room. Drag it up and the orange band opens, and that band is the entire point of the instrument. It is the distance between what the money bought and what the delivered program bought, and it grows only as you convert more of the schedule. The reach counts on the right do not change as you drag; they are fixed and sourced. What changes is how much of that fixed reach the cadence actually lands, which is why the accelerator's real product is the cadence and not the network. The network is the same whether or not you show up. The lift is almost entirely in showing up.
Faster, not just bigger
There is a second-order effect the subtraction reveals that a founder feels but rarely names, and it is worth stating plainly because it cuts both ways. Compressing a network's worth of introductions, a curriculum's worth of hard questions, and a cohort's worth of peer comparison into six months does not only make good outcomes arrive sooner. It makes every outcome arrive sooner, including the bad one. Yu's study of accelerator effects found exactly this: programs speed ventures up, and one of the things they speed up is failure, because the compressed exposure surfaces a fatal problem in months rather than letting it hide for years [3]. From inside the room this is a feature, not a defect. The cadence that would have delivered a market or a hire faster is the same cadence that tells you faster when there is no market to reach, and for a pre-seed team with a fixed runway, learning that in month four instead of year two is the most valuable thing the schedule can do. The network is what a founder gains access to; the speed is what the cadence does to the clock.
The cohort is an ecosystem you rent for six months
It helps to name what the accelerator is actually renting the founder, because the counterfactual has a broader home than any one program. The reach we have been describing, other founders, mentors, knowledge, and talent concentrated in one place, is what the entrepreneurial-ecosystem literature treats as the real substrate of venture formation: an environment of networks and support that a founder taps into, distinct from the founder's own capital [4]. A startup embedded in a rich ecosystem outperforms an identically funded one that is not, and the mechanism is the density of the network, not the size of the bank balance. An accelerator, read this way, is a six-month lease on a concentrated slice of that ecosystem, with a cadence attached to make sure the tenant does not leave the amenities unused. The cheque gets you the keys. The cadence is what makes it a place you actually live for six months rather than an address you list.
That is the whole of the claim, and it is a modest one. We did not conclude that the money does not matter; a pre-seed team without it does not get to the room at all. We concluded that the money is not what the room changed. Hold the capital fixed and the change that remains, the peers you are made to sit with, the mentors you are made to book, the reach you are made to spend, is delivered by the cadence and would not have followed the same cheque handed over in isolation. The accelerator's product is not the capital. It is the obligation to use the network the capital bought you a seat inside.
Limitations
This is a single team's reading of a single program, and it is a counterfactual argument, which means the central comparison, the same startup with the same money but no cohort, is one we reason about rather than one we ran. We did not have a control version of ourselves outside the program, so the split the instrument draws between a money-and-access floor and a network contribution is illustrative framing, not a measured return; the counts it rests on, the 1,500-plus startups, 120-plus team, 50-plus countries, 100-plus curriculum hours, 12 mentorship hours, six two-hour masterminds, and six-month length, are the real sourced figures, but the index units and the floor-versus-network share are ours to make the argument legible, not data. The cited work is evidence that the cadence and cohort are mechanisms distinct from capital in general [1] [2] [3] [4]; it is not evidence about the size of the effect for any one startup, and effect sizes in that literature vary widely by cohort, sector, and stage. Our engagement was a deep-tech, single-product build inside an oil-and-gas problem, which is unusual for an accelerator cohort, so how much the peer comparison helped is plausibly lower for us than for a room of same-stage consumer startups and should not be read as a general multiplier. And the whole argument holds the money constant to isolate the network; it says nothing about programs where the capital terms are the point, where a bad deal on the cheque can outweigh everything the cadence delivers.
Why we kept the counterfactual
The habit this left us with is to evaluate any program, accelerator or otherwise, by the subtraction rather than the brochure. Ask what it changes once the money is held constant, and the answer tells you whether you are buying capital with a network stapled on, or a network with the cadence to actually spend it. For us the second was the real purchase, and the tell was simple: the parts of the six months we would have skipped if they were optional, the masterminds, the booked mentor hours, the curriculum with a completion bar, were exactly the parts that did the work. A cheque asks nothing of you. The thing worth paying for was the program that would not let us leave the network on the shelf.
References
[1] Cohen, S., and Hochberg, Y. V. Accelerating Startups: The Seed Accelerator Phenomenon (2014). The foundational account defining accelerators by the peer cohort and the scheduled mentorship dose rather than by the capital. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2418000
[2] Hallen, B. L., Cohen, S. L., and Bingham, C. B. Do Accelerators Work? If So, How? Organization Science 31(2), 2020, pp. 378-414. Isolates the program's schedule of forced, broad, intense interaction as a mechanism distinct from selection or capital. https://pubsonline.informs.org/doi/10.1287/orsc.2019.1304
[3] Yu, S. How Do Accelerators Impact the Performance of High-Technology Ventures? Management Science 66(2), 2020, pp. 530-552. Finds programs speed venture outcomes, including faster failure, consistent with compressed learning and network access rather than added capital. https://pubsonline.informs.org/doi/10.1287/mnsc.2018.3256
[4] Stam, E., and van de Ven, A. Entrepreneurial Ecosystem Elements. Small Business Economics 56, 2021, pp. 809-832. The networks-and-support framing that treats a venture's access to founders, mentors, and knowledge as an ecosystem property a program concentrates, distinct from the founder's own cash. https://link.springer.com/article/10.1007/s11187-019-00270-6
