Agency specialization: why do-everything agencies stall
Agency specialization by the numbers: why a do-everything agency grows slower, three axes for choosing a niche and a test that keeps your current clients.
Should an agency specialize? In most cases, yes. An agency that does everything for everyone has nothing to show a buyer that proves it knows the buyer’s problem, so the sale takes longer and the conversation turns to price. In Promethean Research’s study of digital agencies, specialists grew 43% faster than average in 2024, and 84% of agencies identified as specialists that year. The condition: specialization is a choice of what the firm is known for. Cutting the list of services alone does not guarantee growth.
In this post I’ll show three axes you can specialize on, a way to pick a niche from projects you’ve already done, and a test that doesn’t cut off your current clients. At the end I’ll describe what we did with this at our own firm and what we don’t know yet.
What agency specialization means
Agency specialization is a decision about what the firm is known for in the market: one client industry, one problem it solves, or one type of service. The other half of that decision is the work the firm turns down. David Maister put it this way: “Strategy is deciding whose business you are going to turn away.”
A do-everything firm has nothing to prove its expertise with
The symptoms tend to be similar: sales conversations drag on for weeks, the client compares your proposal with three others and negotiates the rate, and the portfolio is a collection of projects that came through referrals. In my view this comes from the way services are bought. A buyer can’t judge the quality of the work up front, so they look for proof that the firm has already done what they need. An agency with a “bit of everything” portfolio has that proof scattered across many topics.
Maister wrote about this in “Strategy Means Saying ‘No’”: to be differentiated, you have to be known for something in particular, and you can’t build a reputation on something you only do occasionally. A firm that never says no, in his words, ends up as “one more undifferentiated firm, trying to do a little bit of everything.”
Data from Promethean Research’s 2025 report points the same way. Specialists grew 43% faster than average in 2024, and the report’s authors link this to an easier sales cycle and an easier time winning work when an agency is seen as an expert in a niche. Specialists also tend to earn higher net margins because, as the authors put it, clients are buying expertise, not hours, and look less closely at the price. The survey skews toward larger agencies, so it shows a correlation, not proof that specialization by itself raises growth.
You choose agency specialization on one of three axes: industry, problem, service
A niche agency can do many things, as long as one element is narrow and that’s what it’s known for. Here are the three axes:
| Axis | What it means | What you show the buyer | Risk |
|---|---|---|---|
| Industry | you serve companies from one industry or one type (e.g. service firms), with different services | you know their market, vocabulary and typical problems | dependence on one industry’s economic cycle |
| Problem | you solve one business problem for clients from different industries, choosing services to fit it | you know how that problem ends when nobody solves it | harder to describe in one sentence than an industry or a service |
| Service | you do one type of work for clients from different industries | you do it better and faster than a firm that does it occasionally | easy to compare you with other providers of the same service |
In Promethean’s study for 2024, specializing by service was more common (77% of agencies) than specializing by industry (57%); some agencies specialize on both at once. The report doesn’t measure the problem axis separately.
You choose one axis as the main one. The other elements can stay broad: a software house specialized in logistics can still build web and mobile apps.
How to choose a niche: start from projects you’ve already done
You can usually see the niche in your project history; nobody has named it yet. I recommend three steps:
- A list of projects from the last 12–24 months. Input: invoices, contracts, CRM. Output: a table where each project has the client’s industry, the problem it solved, the type of service, your assessment of the margin, and whether the client came back with another engagement.
- Looking for repeats. Input: the table from step 1. Output: one to three candidate niches, each on one axis, with a list of projects that support it. Don’t consider a candidate with no projects behind it.
- The niche statement. Input: the chosen candidate. Output: one sentence “for whom, with what problem, with what service” and a “what we don’t do” list. The element from the chosen axis is narrow in the sentence; the others can stay broad.
Niche statement template:
| Field | What you write |
|---|---|
| Axis | industry, problem or service |
| For whom | type of client |
| With what problem | what the client wants to solve |
| With what service | what you do |
| What we don’t do | work you don’t take on, and what you do with such an inquiry |
| Projects that support it | projects from the table in step 1 |
A fully filled-in example. This is an illustration, not a real firm: take a software house that noticed in its project table that clients in logistics bring the best margins and the most repeat work.
| Field | Example (illustration) |
|---|---|
| Axis | industry |
| For whom | logistics and freight forwarding companies |
| With what problem | various: from shipment tracking to settlements with carriers |
| With what service | web and mobile apps, integrations with the client’s systems |
| What we don’t do | online stores and games; we pass such inquiries to firms that specialize in them |
| Projects that support it | projects for logistics clients from the table in step 1 |
The niche statement from this example: “We build web and mobile apps and integrations for logistics and freight forwarding companies, from shipment tracking to settlements with carriers.” The industry is narrow; the problem and the service stay broader.
Once you have a niche statement, describe the client it refers to. I covered that in How to create an ICP for a B2B service firm.
Test the niche on new clients before you change anything for current ones
The biggest fear about narrowing is losing revenue. That’s why you test the niche on new sales and keep doing for current clients what you’ve been doing. In practice:
- prepare a separate service page for the niche and talk about it in new sales conversations,
- for a few months, note for every new conversation whether the client is from the niche, how long it took to reach a decision and whether they negotiated the price,
- pass on new inquiries from the “what we don’t do” list from the first day of the test; accept other out-of-niche inquiries and log them the same way as niche conversations, because they are your comparison group.
A few months is a realistic time frame, because in services a single sales conversation often takes weeks, and to compare you need several conversations from inside and outside the niche.
Changing course doesn’t guarantee growth, but standing still did worse in Promethean’s data for 2024: agencies that repositioned their offerings grew 8%, and those that changed nothing grew 1.1%.
The test is softer than Maister’s advice, which tells you to turn work away right away, because for a few months you accept some out-of-niche work. Maister answers an excuse you may well hear from yourself here: “If you can’t afford to say ‘no’ until you are successful and distinguished, then you’ll never be successful and distinguished.” That’s why the “what we don’t do” list applies from the first day.
A new request from a current client for something on the “what we don’t do” list is also new sales. Work you already do for that client continues as before, but for the new request Maister advises not saying “We don’t want your additional work,” but “We are not your best choice for that new need,” and recommending a specialist. In his view, turning work down this way builds the client’s trust.
When specialization doesn’t help
In Promethean’s data for 2024, agencies that reduced their number of services contracted slightly, and among agencies that changed their offering, those that expanded their services grew fastest, by 9.7%. Expanding services doesn’t have to mean losing specialization: an industry-focused agency can add more services for the same industry. The report doesn’t say how many of these agencies did that.
In the 2026 report, with data for 2025, the result is the opposite: agencies that reduced services grew 13% on average with 30% net margins, against an average of 7.5% growth and 13% margin. These are two different samples (1,062 agencies in the 2025 report and 119 responses in the 2026 report), so I compare the direction, not the numbers. The authors themselves say the specialization story in this data is more nuanced than “specialize more” advice. In my view, it follows that cutting the list of services alone doesn’t guarantee growth: what counts is whether the client can see what you’re known for.
Specialization also doesn’t help when:
- it’s only a claim: if 84% of agencies already called themselves specialists in 2024, the word alone sets nobody apart, and according to the report’s authors the edge comes from how clearly and credibly you define your specialization,
- the market is too small to keep your team busy, or depends entirely on one industry that is in a downturn,
- you have no projects in it and would have to promise experience you don’t have.
The data I cite comes mostly from the North American market.
Once the test confirms the niche, rewrite your homepage and offer
I’d count it as confirmation when niche conversations reach a decision faster or with less price negotiation than conversations in the comparison group. Agency positioning shows first on the homepage, so that’s when it’s worth changing three places:
- the headline says, in your niche statement, who you help and with what, instead of promising “full-service marketing”,
- niche projects go to the top of the portfolio; the rest can move lower or go,
- the service page has a short “what we don’t do” list, so a buyer from the niche sees they aren’t one of many topics.
How we narrowed B2BForce
We narrowed the market to service firms of 1 to 50 people: agencies, software houses and consultancies. We treated knowledge of how a service firm works as our advantage, not the code, which is open. We assumed there is less competition in this niche than among general AI tools for marketers. The main axis is client type, a variant of the “industry” axis: service firms of different kinds. We narrowed the service as well, even though the method in this post doesn’t require it.
Before, it was a roughly two-week sprint with a broad scope (content, SEO, competitor monitoring, prospecting). Now the basic paid package is 5 working days, Claude Code only, with a small, repeatable scope. Our firm profile has a “what we DON’T do” section, and larger organizations that need roles, audit and permissions go to a separate offer.
A limitation: here we break our own rule from step 2. We don’t have projects or case studies yet that this niche would come from, so we treat it as a hypothesis and don’t promise results in it. That’s why we test it the way I described above: on new conversations.
If you want to write your niche down in one place, you can use our repo. A repo is a folder of text files on GitHub, and a skill is an instruction an AI agent (e.g. Claude Code) follows to do a task step by step. The firm-context skill interviews you about the firm, its main services (one to five) and clients, and saves the result in workspace/firm/profile.md and in a separate file per service in workspace/firm/services/. The marketing-company-profile skill adds services and specializations from the firm’s website. Without a firm description and at least one service, the repo’s rules tell the agent not to run the marketing, sales or market-analysis skills. The repo is MIT-licensed: github.com/b2bforce/b2bforce. More on the open source page.
Frequently asked questions
Can a full-service agency be a specialist?
Yes, if the industry or the problem is narrow. An agency that does strategy, campaigns and the website only for companies in one industry is an industry specialist with a full list of services. In Promethean’s study, some agencies specialize on service and industry at the same time.
How narrow should a niche be?
Narrow enough to describe in one sentence and to show projects in it, and broad enough to bring in enough clients for your team. If you can’t see any repeats in your project table, the niche is probably too narrow, or you don’t have one yet.
What do I do with current clients outside the niche?
Keep serving them in what you already do for them. The niche test is about new sales, and a new request from a current client counts as new sales too. When a current client asks for something new from the “what we don’t do” list, tell them you’re not the best choice and recommend a specialist. Maister allows doing such work if the client insists.
Summary
If your agency takes every project, I recommend starting with a table of projects from the last 12–24 months. Find the repeats, choose one axis and write a niche statement with a “what we don’t do” list. Then test it for a few months on new sales conversations, and keep doing for current clients what you’ve been doing.
Every method on this blog also runs as a skill in the open-source B2BForce repo. About the repo