What Is the Difference Between an Agency, a Consultancy and a Product Studio?

Ask what separates an agency from a consultancy and nearly every answer you find will have been written by one or the other. That is not a conspiracy, it is a supply problem. The people motivated to publish definitions of these models are the people selling one of them, so the definitions arrive pre-loaded. The consultancy's version makes execution sound like commodity labour. The agency's version makes advice sound like an invoice for a slide deck. The product company's version treats both as things you do while waiting to become a product company.

A definition written to flatter its author is positioning, not description. It is also useless to the two people who most need one: the buyer trying to work out what they are about to be sold, and the operator trying to work out what they have actually built.

Set self-description aside and rebuild the categories from two questions that are hard to spin, because both have observable answers that show up on invoices and balance sheets rather than on about pages.

• Revenue mechanism. What event has to occur before money arrives? An hour worked, a deliverable accepted, a judgement delivered, a subscription renewed, a stake sold.
• Ownership. When the engagement ends, what does the company still hold? A relationship and a permission to describe the work, or an asset that can earn again without a new client.

These two questions cut where the marketing language blurs. A firm can describe itself as strategic and still be paid only when something ships. A firm can describe itself as a product business and still find that every pound it earns is tied to a named person's calendar. Neither of those is dishonest, exactly. It is just that the description was chosen and the mechanism was not.

Both tests have a limit worth naming before they are used. Revenue mechanism tells you what a company is structured to do; it says nothing about whether the company does it well. There are consultancies that ship better than agencies and agencies whose thinking is sharper than any deck. Where you find that, the mechanism has been overridden by unusual people, and the model has stopped being predictive. That is precisely the condition under which the frame in this piece is wrong. It is also rarer than the firms claiming it would have you believe, and it is worth asking for the evidence rather than accepting the claim, because the claim costs nothing to make.

Four models, then, separated by mechanism rather than by tone of voice. Most real companies sit near one and borrow from another. The label still matters, because it tells you which of the two the company is organised, staffed and priced to do well, and which one it does on the margins when a client insists.

Advise and execute: where a consultancy and an agency actually diverge

A consultancy sells judgement. The thing purchased is a position taken: an assessment, a diagnosis, a recommendation the client did not feel able to reach alone. Revenue is triggered by the delivery of that judgement. The client then acts on it, or does not, and either way the transaction is complete. The artefact, whether report, model, workshop or roadmap, is packaging. The substance is the view.

An agency sells execution. The thing purchased is a change in the world: a site, a campaign, an identity, a running system, a set of files that did not exist before. Revenue is triggered by the work being made and accepted. Advice happens constantly inside an agency engagement, but it is given in service of the build and it is rarely the line item.

The cleanest test for the difference between an agency and a consultancy is what happens when the recommendation is unwelcome. A consultancy's product survives the client disagreeing with it. The judgement was still formed, still delivered, still the thing bought. An agency's product does not survive it, because disagreement halts the build, and a halted build is not a deliverable. That asymmetry explains a great deal of behaviour that otherwise looks like temperament. Consultancies can afford to be blunt. Agencies are structurally encouraged to find a way to proceed.

A second test is accountability for outcome. A consultancy hands back a decision and its reasoning, and the result depends on execution it does not control, which is why consultancy engagements so often end with the honest and unsatisfying position that the advice was sound and the implementation was not. An agency hands back a made thing whose quality is visible and attributable, and whose commercial effect usually is not, which is why agency engagements so often end with a well-built artefact and an argument about what it achieved.

The characteristic failures follow from the same asymmetry. A consultancy fails by being correct and unimplementable, producing a recommendation that assumes capacity, appetite or authority the client does not have. An agency fails by producing a beautifully finished answer to a question nobody stopped to check, because checking the question is unpaid work that delays the paid work.

Product studio, defined explicitly, and how it differs from a venture studio

A product studio earns from products it owns. Not from hours, not from artefacts handed over and forgotten, but from things it builds, keeps and remains responsible for after release. Naming it a product studio at first mention matters, because the word studio on its own carries almost no information. Design studios sell client execution. Development studios sell teams. Game studios sell titles they may not own. Studio describes a working culture, not a revenue mechanism, and a buyer who reads it as a mechanism is guessing.

A venture studio is a different mechanism wearing a similar word. A venture studio builds companies rather than products, and its return comes from the ownership stakes it holds in those companies rather than from what the companies charge their customers. The unit of production is an entity with its own shareholders, its own hiring and eventually its own separation from the studio that started it. The time horizon is measured in ownership events. This is a genuinely different business from a product studio, and the two are not usefully interchangeable, whatever the naming convention suggests.

The distinction is not pedantry, because it predicts what a buyer can expect from the relationship. A product studio's interest is in the product being used and continuing to work, because that is where its revenue lives. A venture studio's interest in any individual product is contingent on the company holding that product remaining worth holding. Neither interest is disreputable. They simply point in different directions when a product turns out to be modestly useful rather than spectacularly so.

What a company sells when it sells access to its own product

A product company sells access. Subscription, licence, seat, usage, whatever the meter, the unit sold is not a person's time, and serving the next customer does not require finding another person to serve them. That single fact reorganises everything else about the business.

Two consequences follow, and both are visible from outside. The first is that most of the work happens before the sale rather than after it. What a client of an agency buys is work not yet done; what a customer of a product buys is work already done, which is why the sales conversation is about fit rather than about scope. The second is that the sale is not finished at signature. Access is renewed or abandoned, and the company therefore carries an obligation with no natural end date: the thing has to keep working, keep being maintained, and keep being worth the next payment.

That standing obligation is the real cost of the model, and it is treated properly in the piece on the standing cost of a product rather than here. It is worth being concrete about where this article is published. Zealsync is a parent technology company that develops four products, Acrosite, Flidu, Nichevio and Prooflin, and operates Intense Path, its brand, growth and technology agency. That places two revenue mechanisms inside one entity, which is why these distinctions are not academic here. What the combination costs in practice, and how capacity gets allocated between the two, belongs to the discussion of client work and product work.

Five words get used as though they were synonyms for owns. They are five different claims, and only some of them are claims about legal structure at all. Getting them wrong is not a stylistic slip, because two of them appear on documents that have to be accurate.

• A parent company owns and operates. It sits above other things and is answerable for what they claim, how they behave and what they promise to a market. The word carries an operating responsibility, not merely an ownership position.
• A holding company owns, and the relationship can stop there. It holds shares or assets, and its involvement may be purely financial. A holding company that never touches the operations of what it owns is behaving entirely normally; a parent company that does the same has misdescribed itself.
• A subsidiary is a separate legal entity. It has its own registration, its own liabilities, its own contracts, and in principle it can be sold without selling the company that owns it. It exists whether or not any customer has heard of it.
• A brand is a claim made to a market. A name, a promise, an accumulated set of expectations. It cannot sign a contract, cannot hold a liability, and cannot be sued. It can be owned, licensed and retired, but it is not an entity and does not become one by being written in capital letters.
• A product line is a grouping of things sold. It is an internal organising device that customers may never see and need not care about. It carries no legal weight and no market promise of its own.

Conflating them produces statements a company cannot support. Describing a brand as a subsidiary asserts a legal entity that does not exist. On a website that reads as loose writing. On a contract, an invoice, a privacy page or a set of terms, it is either wrong or unenforceable, because the counterparty named cannot be the counterparty. The error runs in the other direction too: describing a genuine subsidiary as merely a brand understates who is actually on the hook, and a buyer who discovers the difference later has grounds to feel misled even where nothing was intended.

The structure behind this site is a worked example. Intense Path is a brand of Zealsync Private Limited rather than a subsidiary of it. It has its own name, its own market and its own site at intensepath.com, and it has no separate legal existence whatsoever. Every agreement it enters is entered by Zealsync Private Limited. Saying so plainly costs nothing and prevents a category of confusion that is expensive to unwind later. The separate question of when a brand should get its own name at all is set out in the piece on brand architecture for a small company.

Every model is a set of trade-offs that was made once and then forgotten. The trade-offs keep operating regardless, which is why firms so often find themselves weak at exactly the thing their model was never built to do.

A consultancy optimises for the quality and independence of judgement. It protects the ability to say the unwelcome thing, and it invests in framing problems rather than finishing them. What it neglects is delivery reality. Recommendations formed at a distance from the work tend to underweight how long things take, how much of an organisation's attention is already committed, and how much of the difficulty lives in details that only surface during implementation.

An agency optimises for throughput and finish. It is built to start, make and complete, repeatedly, to a standard that survives being seen. What it neglects is the prior question. Work that arrives with a brief attached has already been defined by someone else, and the commercial pressure to begin is a pressure not to reopen the definition. The better agencies push back on briefs; the model does not reward them for it.

A product studio optimises for compounding. It builds things that can be worked on again next quarter, and it accumulates decisions, code and understanding that keep paying. What it neglects is responsiveness and near-term cash. Owned products consume capacity before they return any, and a studio that is genuinely serious about a product has less room to answer an urgent request than a pure client business does.

A product company optimises for retention. Everything bends towards the thing continuing to be worth paying for: reliability, support, the steady closing of gaps. What it neglects is the bespoke need. A request that would serve one customer well and everyone else not at all is a request the model is designed to decline, and the decline is correct even when it is disappointing.

None of these is superior. A ranking would require a shared objective, and these models do not share one. The only defensible statement is that each is well suited to what it is built for and structurally weak at what it is not, and that a buyer who needs the weak thing should buy it from a different model rather than hoping this one will turn out to be the exception.

Operators tend to experience the model as a series of independent decisions, made one at a time and reversible. They are not independent. Choose a revenue mechanism and it will quietly settle at least three other things, usually before anyone notices a choice was made.

Hiring goes first. A consultancy hires for the ability to frame a problem in front of people who are sceptical, which is a different skill from making anything. An agency hires for craft and for the temperament that gets work to a finish under a date. A product business hires for the willingness to own something for years, including the unglamorous parts of it, which is a disposition rather than a technique. Hire the wrong profile for your mechanism and you get a firm that is excellent at something nobody is paying it for.

Pricing follows. Where revenue is triggered by time or by judgement, price attaches to the engagement, and the commercial conversation is about scope and duration. Where revenue is triggered by access, price attaches to the unit of access, and the commercial conversation is about who counts as a user and how much use is included. These are not two styles of the same negotiation. They ask different questions and reward different answers, and a firm running both has to be fluent in each without letting either logic contaminate the other.

The definition of done follows last and matters most. In a consultancy, done is a decision reached and defended. In an agency, done is an artefact accepted. In a product business there is no done at all, only a release and the standing obligation that follows it. Teams that move between models without renegotiating the definition of done keep producing work that is finished by their old standard and unfinished by the new one, and the disagreement that results looks like a quality problem when it is really a definitional one.

Most companies do not sit purely inside one model. The useful move is not to catch them out but to read the label against the invoice, because the invoice is written by the mechanism and the label is written by whoever was in charge of the website that year.

The question is always what unit is being charged for. Hours or days indicates a business selling capacity. A fixed fee against a defined deliverable indicates a business selling execution and carrying the estimation risk itself. A fee against a decision, a review or an assessment indicates a business selling judgement. A recurring charge per seat, per site or per unit of use indicates a business selling access. Where a company presents one and bills another, believe the billing.

A handful of questions will usually resolve the ambiguity faster than any amount of reading.

• If we accept your recommendation and then do nothing, what have we bought? A consultancy has a clean answer to this. An agency mostly does not.
• When this engagement ends, what do we own and what do you retain? Ask specifically about repositories, accounts, source files and the right to leave.
• Who maintains this in eighteen months, and under what arrangement? The answer separates a business that expects to hand over from one that expects to stay.
• What is the smallest thing we can buy from you? Businesses selling access have a natural smallest unit. Businesses selling capacity usually have to invent one.
• Which of your revenue lines would you protect if you had to choose? This is more informative than the homepage, and it is rarely refused outright.

A company running both a client practice and owned products is not being evasive by describing both. It has two mechanisms, and the honest presentation is to say so and to be clear about which one a given conversation concerns. Zealsync sets its own out across the company page and the portfolio, which is the minimum a reader should expect rather than a courtesy.

A label does work before any conversation begins. Say agency and a buyer infers that you will make the thing, that they will be shown work in progress, that there will be a date, and that the relationship probably ends when the thing is delivered. Say consultancy and they infer that you will tell them what to do, that you will charge for the telling, and that the doing is theirs. Say product and they infer that they can look at it, price it, try to understand it without a meeting, and leave without a conversation. Say studio with no qualifier and they infer nothing reliable, which means they will guess, and the guess will be whichever model they most recently dealt with.

A wrong label does not merely misdescribe you. It sets an expectation that your own delivery then violates, and the buyer experiences that as concealment rather than imprecision.

That is the whole trust cost, and it is worth being exact about why it is so disproportionate to the size of the error. The buyer is not annoyed that a word was slightly off. They are annoyed because they made a decision using that word, allocated budget and attention on the strength of it, and now have to explain internally why the thing they described to a colleague is not the thing that is happening. The first contradiction is expensive because it retrospectively reframes everything said before it as sales talk.

Mislabelling usually runs in the direction of perceived seniority. Firms that execute call themselves consultancies because advice sounds more valuable than making. The consequence is that they are then judged against a standard of independent judgement they have not staffed for, and every implementation detail they are genuinely excellent at gets read as a distraction from the thinking they promised. The reverse error is quieter but real: a firm with genuine judgement calling itself an agency, and finding that it is only ever asked to build what someone else has already specified.

If a company describes itself as a parent rather than as a holding company, it has taken on an operating responsibility and should be able to discharge it in plain sentences. The test is not whether the structure is elegant. It is whether someone can ask a direct question and get an answer that does not require a diagram.

• What the legal entity is, in full, with its registered address, and which name appears on an agreement.
• Which of the things it owns are separate entities and which are brands or products with no independent legal existence.
• What each product is for, described in terms a prospective user would recognise, without borrowing a claim from a neighbouring product.
• Who is answerable when a product does something wrong, and through which contact route a complaint actually travels.
• Where the boundary sits between what a product does today and what is merely intended, stated without implying that intention is imminent.

Some of this cannot be delegated at all. Legal identity, liability and the accuracy of public claims sit with the parent regardless of how independently a brand behaves from day to day. A brand can have its own voice, its own site and its own audience, and none of that moves responsibility anywhere. Equally, there are things a parent should decline to say on a product's behalf: capability it has not shipped, dates it has not committed to, and comparisons it cannot substantiate. Silence on those points reads as discipline. Filling them in reads well for exactly as long as it takes someone to check.

For an operator rather than a buyer, the same tests run inward. Answer these about your own company, using last year's invoices rather than this year's intentions, and the label usually resolves itself.

• What triggered each payment you received? If most were triggered by a person working, you are selling capacity, whatever the site says.
• If every client relationship ended tomorrow, what would still be capable of earning? If the honest answer is nothing, the product language is aspiration rather than description.
• What do you hire for when you are under pressure? Firms revert to their mechanism under load, and the emergency hire is the truthful one.
• When is a piece of work finished? If nobody in the company answers this the same way twice, you are probably operating two models without having decided to.
• Which of your public claims would fail if a prospective client asked for evidence this afternoon? Remove those before deciding what to call yourself.
• Does anything on your legal pages name an entity that does not exist? This is the one item on the list that is not a judgement call.

The point of the exercise is not to arrive at a fashionable label. It is to arrive at one you can defend under questioning, from a buyer who has been sold to badly before and is reading everything you publish as evidence rather than as information. If the answer to a question here is genuinely unclear from the outside, ask directly; a company that cannot answer it plainly has told you something either way.