The startup versus SME debate has been going on in Bangladesh for as long as I can remember. I have heard it at events, read it in newspapers, and watched it play out on Facebook, which somehow manages to generate more heated discussions than most actual policy issues in this country. I have been meaning to put my own thinking in writing for a while because writing forces me to think clearly and learn to fill my gaps. This is that attempt.
My take will probably frustrate people on both sides of the debate, but I want to take the risk. If you have a different opinion, I would love to hear your views.
To that end, I want to be upfront about where I stand. I do not think the distinction between a startup and an SME is as hard and fixed as most people in this debate assume. I think it is fluid, especially in the early days of a company. Moreover, I also feel that in many instances, SME is not a type of business; it is a stage of a business.
I think the canonical definitions that get cited in this debate often get misinterpreted and, in some instances, are built on a specific model of business, the Silicon Valley software company, that does not cleanly describe most of the interesting businesses being built in Bangladesh, and in other cases, businesses that Bangladesh needs.
I think the definition matters most in the context of capital decisions, and even there, professional investors do not actually use the startup or SME label as their decision criterion.
And I think the most consequential and least discussed dimension of this debate is what happens when the government tries to encode a fixed definition of a startup into policy. That, to me, is where the real danger lies.
These are the things I want to work through here.
I will start with what the serious thinkers and practitioners in the field have actually said, and then I will make the case that those definitions are more fluid than the way they are presented in our market and that many of those definitions rest on assumptions that need examining before we apply them in this context verbatim.
Then I will offer my own position on the fluidity of the distinction, and what it actually means for capital decisions. And then I will spend some time on the policy question, because I think it deserves more careful attention than it has received.
There is no single authoritative definition of a startup. What we have is a bunch of definitions from people who built startups, invested in them, or studied them closely, and each one explains something the others miss. Let's look at some of the most prominent ones.
Steve Blank: A temporary organization searching for a model
Steve Blank, the Silicon Valley entrepreneur, Author, and Stanford professor who formalized the Customer Development methodology, defines a startup as: A temporary organization formed to search for a repeatable and scalable business model.
This is my favorite definition. Partly because I agree with it. Partly because I think it is more meaningful.
Temporary — because a startup is a phase, not a permanent identity. Once you find the model, you stop being a startup and become a company.
Searching — not executing. A startup operates in the domain of discovery. It does not yet know who its customer is, what they will pay, or how to reach them reliably. A company already knows these things and is optimizing a known model.
Repeatable and scalable — because finding something that works once is not sufficient. The goal is a model that can be replicated and grown without proportional cost increases.
Blank also drew the clearest line in the literature between startups and small businesses. "While large companies execute known business models," he wrote, "startups are temporary organisations designed to search for a scalable and repeatable business model."
He also identified six distinct types of ventures that all get called startups: lifestyle startups, small business startups, scalable startups, buyable startups, large-company startups, and social startups.
This taxonomy is the most honest thing in the entire debate, because it names genuine diversity rather than flattening it.
Eric Ries: Extreme uncertainty as the defining condition
Eric Ries, whose 2011 book The Lean Startup brought Blank's ideas to a global audience, defines a startup as: A human institution designed to deliver a new product or service under conditions of extreme uncertainty.
Ries strips the definition to its essential condition: uncertainty. He says nothing about size, technology, age, or investment. What makes something a startup is the epistemic situation it is operating in. It is running experiments to resolve unknowns, not executing a plan built on knowns.
This framing is time-sensitive rather than type-based: a business is a startup for as long as it faces fundamental uncertainty about whether its model can work.
Paul Graham: Designed to grow fast
Paul Graham, co-founder of Y Combinator and one of the most influential voices in the space, offers the most stripped-down definition: A startup is a company designed to grow fast. Being newly founded does not in itself make a company a startup. Nor is it necessary for a startup to work on technology, or take venture funding, or have some sort of "exit". The only essential thing is growth.
The most famous and arguably most influential thinker in the space. Everyone loves to refer to this line: startup = growth, but you wouldn't hear anyone going into the nuances Graham explains in his essay.
For Graham, the structural growth ceiling is the test. A business whose economics require costs to grow proportionally with revenue is not a startup in this sense, however new or innovative it is. A startup has a model designed to grow disproportionately to its cost base.
Peter Thiel: A plan to build a different future
Peter Thiel, co-founder of PayPal and venture capitalist, offers a more philosophical definition: Positively defined, a startup is the largest group of people you can convince of a plan to build a different future.
In his broader framework in Zero to One, the most valuable companies create something genuinely new; they go from zero to one, rather than copying and expanding what already exists. A startup begins with a vision of a future that does not yet exist and the hard work of convincing others that this future is achievable.
Before I offer my own position, I want to make a point that often gets lost in our local debate. The canonical definitions reviewed above are not the problem. If anything, read carefully; they support the fluidity argument. They do not draw a hard line between startups and SME. The problem is how they get selectively read here.
Look at what these thinkers actually say.
Blank does not define a startup as a technology company. He defines it as any organization searching for a repeatable and scalable business model, and then goes further, identifying six distinct types of startups, most of which have nothing to do with software.
A lifestyle business is a startup, by his definition. A neighbourhood shop genuinely searching for a model it can scale is a startup by his definition. He is explicit that the scalable technology venture is one type among many, not the archetype against which all others are judged.
Ries is equally clear. His definition says nothing about technology. The defining condition is uncertainty, and uncertainty is a universal condition of any new business in any sector. There is no SME exception written into his framework.
Graham is the one most often cited to argue for a narrow, high-growth-only definition of startup. But read what he actually writes: "Nor is it necessary for a startup to work on technology, or take venture funding, or have some sort of exit. The only essential thing is growth."
He is not saying the company must be a software business. He is not saying it must raise venture capital. He is saying it must be designed to grow fast, a criterion that a mudi dokan with a serious franchise ambition could plausibly meet.
These definitions, read in full and in good faith, do not make the hard startup/SME distinction that our ecosystem debate keeps trying to enforce. They describe a set of conditions, uncertainty, search for a scalable model, and orientation toward growth that can apply to a wide range of businesses.
They are more inclusive and more fluid than how they tend to get deployed.
The problem in our market is not the definitions themselves. It is that people extract only a fragment of those definitions, the fragment most associated with Silicon Valley software companies, and then use that fragment as if it were the whole picture. In some instances, I guess people do this misreading out of a lack of reading, understanding, and critical thinking. In some other instances, I suspect it is driven by personal interest.
Marc Andreessen's influential 2011 essay "Software Is Eating the World" is the key exhibit here. Andreessen's argument was specific: software companies, because of the near-zero marginal cost of serving one additional user, have an economic profile unlike anything that came before. The economics become exponentially more favourable as you grow, which is what makes the venture capital model work for those businesses. This is a real and important observation. But it is an observation about a specific type of business, pure software, not a universal theory of what a startup is or should be.
When this thesis gets imported into local startup discourse and fused with Graham's growth framing, the result is a working definition that goes roughly like this: a startup is a technology company designed to grow exponentially by exploiting near-zero marginal costs. Everything else is an SME. That definition is not in any of the canonical texts. It is a selective assembly of fragments, pulled from their context and hardened into a gatekeeping tool.
The consequences of this selective reading are visible in the Bangladesh ecosystem. Founders of hybrid businesses, companies that use technology but have significant physical operations, or manufacturing or trading businesses, are told their unit economics are wrong because they do not look like SaaS metrics.
Investors trained on Silicon Valley models struggle to value businesses where marginal costs are real and operational intensity is high.
Policymakers reach for an innovation-and-technology test when defining startup, implicitly encoding the software company archetype into regulatory frameworks, while the original thinkers they cite never intended any such narrowness.
Bangladesh has pure software companies and pure technology businesses. But the most significant ventures in this market are almost all hybrid. They use software, often sophisticated software, but they are not pure software businesses. They have physical operations that are central to what they do, and that carry real costs.
A grocery delivery company has software, the app, the ordering system, the inventory management, but it also has riders, dark stores, cold chains, and fresh produce that spoils. Every additional delivery requires a real person on a real motorcycle navigating real Dhaka traffic.
A logistics platform has software, but it has vehicles, warehouses, and last-mile operations. An agri-input company or a manufacturing company has technology, but it has field agents and physical distribution networks. An education company has digital content, but its value often depends on real teachers conducting real sessions with real students.
These are not inferior businesses because they have physical operations. They are the businesses appropriate to Bangladesh's economy and stage of development. A country where most transactions still happen in cash, where road infrastructure shapes everything, and where the majority of economic activity involves physical goods and physical labour will produce hybrid businesses. This is the right response to the actual market. The business our market needs should rightly be distinct and unique to our market. These businesses don’t necessarily need to agree to definitions from other markets. Investors who follow the Bangladesh market and want to invest here, I assume, understand this reality.
In fact, the canonical definitions acknowledge this about these businesses; they don’t disregard these businesses. Read carefully. Blank's definition includes them. Ries's definition includes them. Graham's definition could include them if their model is genuinely oriented toward growth.
The case I want to make is that the selective, tech-inflated reading that dominates local discourse does exclude them, and that exclusion has real costs. The remedy is not to discard the definitions but to read them properly: in full, in context, and without importing assumptions the original authors never made.
After more than 11 years of covering businesses in Bangladesh and speaking with hundreds of founders, operators, and investors, here is what I actually think.
In the early days of a company, there is often not as much difference between a startup and an SME as the debate implies.
The structural differences that the canonical definitions point to — the relationship to uncertainty, the aspiration toward scalable growth, the founding vision — are real. But they are not fixed at birth. They develop over time, as companies and founders develop.
Treating them as binary categories assigned on day one is a serious misunderstanding of how businesses actually grow.
More importantly, I feel that SME is not always a type of business, even startup is not always a type of business, these are stages of businesses.
One point I want to draw attention to is that it may feel like some of the canonical definitions treat the intention toward scalability as something a founder either has or does not have from the beginning.
Paul Graham's definition implies that the design is present at the founding. But in Bangladesh, and in most markets, this assumption does not hold cleanly.
Many of the most interesting entrepreneurs I have spoken with over the years did not start with a clear vision of building something large. They started with a problem they could see in front of them, or an opportunity within reach, or simply a need to earn a living. The ambition to scale came later, sometimes because they met someone who showed them what scale looked like, sometimes because a market opened up they had not anticipated, sometimes because they got better at what they were doing and could see further ahead than they could at the start.
This is not a failure of vision. It is how exposure works. Someone who grew up in a small town and never watched a large business being built does not start a company with the same conception of what is possible as someone who grew up around entrepreneurs, or studied business at a university, or spent years working inside a large organization.
The ambition to build at scale is not a fixed trait of character. It is shaped, over time, by what you see, who you meet, what you learn, and what you come to believe is achievable.
Scale, in other words, is a function of ambition, and ambition is a function of exposure.
This matters enormously for how we think about the startup versus SME question. If ambition develops through exposure rather than being fixed at founding, then the question of whether a business is a startup or an SME cannot be definitively answered on day one. It is an evolving question, answered differently as the founder develops and as the business develops.
Consider a small general store, a mudi doka,n operating in a neighbourhood in Dhaka. By every conventional criterion, this is an SME. It is executing a known model in a defined local market. Its economics are linear: more customers means more inventory, more staff, more space. It is not searching for a scalable business model in the Blank sense. It is running one that has been validated millions of times.
Now consider what happens if the owner of that shop, at some point in his journey, starts to see a different possibility. He introduces simple inventory software. He negotiates a supplier arrangement that gives him meaningfully better unit economics than his competitors. He realizes that the model he has built in one neighbourhood could be replicated in ten others with relatively small additional investment per location. He raises some capital, opens more outlets, standardises his operations, and starts treating the whole thing as a system rather than a single shop.
That business has changed. It is no longer simply executing a known model in a single local market. It is testing whether a systemised, technology-enabled retail model can scale across a city or a country. The ambition shifted. The economics shifted with it. And the appropriate category shifted too.
There is nothing unusual about this trajectory. It is the origin story of many of the most interesting businesses in Bangladesh's recent history: companies that started as conventional service operations and found, over time, a way to build something more systematic and more scalable.
The point is that the transition happened through the development of the founder and the business, not through a decision made on day one. A definition that tries to classify the business at birth would have gotten this wrong.
The startup versus SME distinction is often invoked most forcefully in the context of capital, with the implication that startups deserve venture capital and SMEs should stick to bank loans. This framing rests on a real underlying logic, but it also fundamentally misrepresents how investment decisions are actually made.
Venture capital firms and private equity firms invest in companies. Not in startups. Not in SMEs. In companies.
When a VC firm evaluates a potential investment, they are asking specific questions: How large is the addressable market? What do the unit economics look like today, and how do they improve as the company scales? Is there a defensible competitive position? What is the quality of the team? What does the return profile look like across a range of scenarios?
The label, startup or SME, does not appear in that analysis. It is not a decision criterion. Moreover, there are all kinds of investors who invest in all kinds of companies. They have their own thesis, priorities, mandate, and approach that guide their investment decisions.
A mudi dokan that has found a systemised, replicable model and is growing rapidly with improving unit economics will attract venture interest. A company that calls itself a tech startup but has poor unit economics, a small addressable market, and no clear path to defensibility will not attract serious investment, whatever it calls itself.
The label is a decoration. The economics are the substance.
This matters for the debate in a direct way: if the people making actual capital allocation decisions have already moved past the startup/SME distinction in favour of specific economic and strategic criteria, then the debate is taking place at a level of abstraction that is disconnected from how capital actually flows.
Young founders who spend energy figuring out whether they qualify as a startup, in the hope that the answer unlocks funding, are solving the wrong problem.
The question that unlocks funding is not "am I a startup?" It is "do my unit economics improve at scale, how large is my market, and can I build a defensible position in it?"
The distinction between startup and SME does still matter for capital decisions, but in a more practical way. Different capital instruments suit different business economics.
A business with linear growth and stable cash flows is better served by debt or revenue-based financing than by equity investment that creates exit obligations and return expectations the business cannot meet.
A business genuinely searching for a scalable model, burning cash to resolve uncertainty, needs risk capital that can tolerate the possibility of total loss.
Getting that match right, between business economics and capital type, is important. But you can get it right through an honest assessment of your economics, without ever resolving the definitional question of whether you are a startup or an SME.
In policy circles in Dhaka, the working definition of a startup has converged on something like this: an innovative business, primarily technology-enabled, with high growth potential.
The people who use this definition are not wrong that such businesses exist and deserve support. But they want to make this definition binding, to use it as the basis for who gets access to startup-specific policy benefits, regulatory relaxations, and government support programmes.
I find this deeply problematic for several specific reasons.
Any binding definition creates a gate. You are either in or out. And in practice, who decides whether your business clears the "innovative" and "high growth potential" bar? Some official or committee will have to make that judgment. And judgment calls in policy systems, in Bangladesh as everywhere, tend to favour the already-favoured.
The businesses that will reliably pass an innovation-and-technology test are the ones already visible to policymakers: the companies in the Dhaka tech ecosystem, the VC-backed ventures, the founders who attended the right universities, and can present their company in the right language.
The mudi dokan owner, who is digitizing his supply chain and starting to franchise across the district, will not naturally present as a "tech startup with high growth potential." He may be doing something more genuinely interesting than many companies that use the label comfortably. But he will not clear the gate.
There is an additional layer to this problem that connects to the software critique above. If the innovation-and-technology definition is implicitly built on the software company model, it will systematically disadvantage the hybrid businesses and many other businesses that are most characteristic of and important for Bangladesh's economy.
A logistics company is not a software company, even if it uses sophisticated software. A grocery delivery operation is not a software company, even if its technology is excellent. An agri-input distribution business is not a software company, even if it has a mobile app that farmers use.
If the policy definition encodes a Silicon Valley conception of what a startup looks like, the businesses best positioned to build real scale in Bangladesh's actual economy will be excluded from it.
The second problem connects directly to the fluidity argument. If ambition is a function of exposure, and exposure is unequally distributed, then many of the most interesting entrepreneurial stories in Bangladesh will follow a trajectory that a binding day-one definition will simply miss.
A founder who starts a manufacturing company, builds operational rigour over three years, and then realizes she can build a national network — that trajectory is invisible to a definition that classifies him on day one. He did not start as a "high-growth tech startup." He became something with high-growth potential through the work itself, through the learning, through the development of his ambition as his exposure grew. A binding definition applied early in his journey would have excluded him.
And retrospective reclassification is not how policy support systems typically work: the benefits flow to those who qualified at the point of classification, not to those who proved the classification wrong over time.
This is why Blank's definition, an organisation searching for a repeatable and scalable business model, is more honest as a starting point than the innovation-and-technology definition.
Blank's framing is descriptive and epistemological: it is about what the organisation is doing, not what it looks like or what sector it is in.
It is, by design, broad enough to include the mudi dokan that is figuring out whether it can scale, and it does not require a bureaucrat to assess whether that process of figuring out qualifies as sufficiently innovative.
Before I sketch what a better policy might look like, I want to sit with the strongest version of the counterargument, because I think it asks an important question that needs a meaningful answer.
The counterargument goes like this: fluidity is all very well as an intellectual position, but policy needs operational definitions to function. A tax relief programme cannot say "support innovative businesses" and leave it there. Someone has to administer it. A loan eligibility criterion cannot be "you are searching for a model", every founder will claim that. An equity co-investment programme cannot be open to every company below a certain age.
Without a concrete definition, you either end up with nothing, because the bureaucracy cannot operationalise the criteria, or you end up with the same captured outcome, where whoever is closest to the people administering the programme gets the benefits regardless of what the written criteria say.
I think this argument is correct. I do not have a complete answer to it. But I think it points toward the wrong solution, a fixed, type-based definition, when the right solution is a different kind of definitional architecture altogether.
The insight I keep coming back to is that the SME framework in Bangladesh already does something useful that the startup definition debate has failed to learn from. The SME policy framework does not ask what type of company you are. It asks what stage you are at. Revenue below a certain threshold, employee count below a certain number: you are a small enterprise.
Different thresholds, different classification. Different classification, different benefits. The criteria are verifiable. Any registrar can administer them. No one has to decide whether your business is innovative enough.
A better approach to startup policy would borrow this logic and extend it, rather than replacing it with a type-based test. The design would work in two layers.
The first layer is a broad early-stage category, defined entirely by objective, verifiable criteria: age of incorporation, revenue below a threshold, perhaps registered and operating in Bangladesh. No innovation test. No technology requirement. No judgment call about growth potential.
Any company in its early stage — the mudi dokan figuring out whether it can franchise, the logistics startup searching for a replicable model, the software company building its first product — qualifies for the same basic set of early-stage benefits. Simplified registration, certain tax treatments, access to specific credit instruments, and maybe expedited regulatory responses. Benefits that reduce the friction of starting and searching are available to everyone at that stage.
The second layer is specific benefits tied to specific regulatory or economic needs or sectoral basis, with criteria defined by the nature of the benefit or the sector rather than by a general startup classification.
A fintech regulatory sandbox has its own eligibility criteria, because the regulatory question is specific: what financial experiment are you running, what consumer protection oversight does it require, what are the systemic risk implications? Those criteria are defined by the sandbox's purpose, not by whether the applying company qualifies as a startup.
An equity co-investment programme has its own criteria, because the investment question is specific: does this company's model have the characteristics that justify risk capital? Those criteria are defined by investment logic, not by a general startup label.
A talent visa fast-track for key technical hires has its own criteria, defined by what the programme is trying to accomplish.
A specific industry, say semiconductor, has it own criteria.
In this architecture, the word "startup" is doing very little work in the policy machinery.
The first layer uses stage-based criteria that any business can meet. The second layer uses benefit-specific criteria that are defined by what each program is actually trying to do.
The result is that a wide range of early-stage businesses, tech and non-tech, software and hybrid, founded by people with high-growth ambitions and people whose ambitions are still developing, have access to the first layer equally.
The more specific second-layer benefits flow to companies that meet criteria specific to those benefits, without a prior determination that they belong to the right type category.
There is good international precedent for this kind of design. Singapore's startup support programmes use relatively objective eligibility criteria — age of incorporation, revenue ceilings, local registration — rather than subjective innovation assessments.
France's Jeune Entreprise Innovante status does include an innovation criterion, but it uses R&D expenditure as a percentage of revenue as the proxy: verifiable spending, not a bureaucrat's judgment about whether the company is innovative. I
Many of the countries that have built functioning early-stage support infrastructure have almost all moved toward verifiable criteria precisely because subjective ones get captured by whoever is closest to the people administering them.
I want to be honest that this design does not resolve every difficulty.
There will always be edge cases where the stage-based criteria produce results that feel wrong, a large company that technically qualifies for early-stage benefits through a newly incorporated subsidiary, or a genuinely early-stage company that has grown faster than the thresholds anticipated.
Every policy framework produces edge cases.
The question is whether the framework's general design produces better outcomes than the alternative.
I believe a stage-based, verifiable-criteria architecture produces significantly better outcomes than a fixed, type-based innovation test, because it distributes benefits more fairly, is harder to game, and does not try to predict at birth what a company will become.
I am not arguing against government support for early-stage businesses. Bangladesh's government has not done nearly enough of it, and what has been done has often been poorly designed.
What I am arguing against is using a fixed, narrow startup definition as the gating mechanism for that support.
The practical implication is this: rather than beginning with "define startup," the policy conversation should begin with "what outcomes are we trying to produce, and what criteria would reliably identify the companies that need each type of support?"
That question produces different answers for different programmes, and none of those answers require resolving, once and for all, whether a given company is a startup or an SME.
Benefits tied to verifiable, objective, stage-based criteria are harder to game.
They distribute more fairly across the full diversity of early-stage businesses in Bangladesh. They do not require someone in an office in Dhaka to decide whether a logistics company in Sylhet is innovative enough to deserve support.
And they leave room for the mudi dokan that becomes a franchise network, the company whose ambition developed through exposure rather than being present at founding, to receive support at each stage of its development, rather than being classified out of the system before it has had a chance to become what it is capable of becoming.
The question "what is a startup?" has a precise enough answer in the academic literature, and I have tried to work through it here honestly. But the definitions in that literature were built for a specific context, and they need to be held carefully before being applied to a market like Bangladesh, where most of the interesting businesses live between software and the physical world.
My own view is that the distinction between startup and SME is real but fluid, especially in the early days. It is more a question of where a company is on a developmental trajectory than a fixed classification assigned at founding.
Scale is a function of ambition, and ambition is a function of exposure, and both of those develop over time, not on day one.
The mudi dokan owner who builds a franchise network was not a different kind of person on the day he opened his first shop. He became a different kind of operator through the work itself.
On capital decisions, the label matters less than the economics. Venture capital firms and private equity firms invest in companies with specific economic and strategic characteristics. They do not have a "startups" portfolio and an "SMEs" portfolio. They have portfolios of companies they believe can return the fund.
Understanding the economics of your business clearly, how your unit costs evolve as you scale, how large your addressable market is, and what competitive position you can build is far more useful than resolving whether you qualify as a startup.
And on policy, I feel strongly that a binding, fixed definition of startup, especially one built on an innovation-and-technology test that implicitly encodes Silicon Valley assumptions, would be a serious mistake for Bangladesh.
It would create a gate that favours the already-favoured, excludes the many important businesses that best reflect Bangladesh's actual economy, and freezes what should be allowed to develop.
A better path is flexible early classification tied to verifiable criteria, with definitions that evolve as companies grow and earn more specific support.
The best entrepreneurial stories I have covered in eleven to twelve years of Future Startup came from people who did not know, on day one, what they were building. They found out by building it.
A policy that tries to classify them before they have had that chance will get the classification wrong, and the cost of getting it wrong will fall on the people who can least afford it.
