How Do Agencies Add a Service Line Without Winning New Clients?

How do agencies add a service line without winning new clients?

By selling something new into relationships that already exist. The expensive, slow, uncertain part of agency growth is convincing a company that has never bought from you that you are competent and safe. That cost has already been paid for every name currently on your book. A second service line changes what those relationships are allowed to be worth, and it does so without adding a single conversation to the pipeline.

This is not a fringe strategy. In the 2026 AgencyAnalytics benchmarks, which drew on responses from 494 marketing agency professionals in a survey run between February and April 2026, 56% of agency leaders reported that more than 40% of their new revenue already comes from clients on the books. The same report found that only 6% of agency clients buy a single service, while 56% engage for three or more. Multi-service relationships are the normal state of the industry, not an advanced move.

So the interesting question is never whether it is possible to increase revenue per client. It obviously is, and most agencies are already doing it. The question is which line, delivered by whom, and what happens the first time it misbehaves. The shorter, buyer-facing version of the argument below is on our page about the arithmetic of adding a service line, and the mechanics of the model itself, which this article deliberately does not restate, are in what white-label voice AI is and how agencies resell it.

Why is agency growth almost always diagnosed as a lead problem?

Because that is what it feels like from inside. When revenue is flat, the visible shortage is new logos, and the visible remedy is more outbound, more content, more referrals. The same AgencyAnalytics survey found acquiring new clients ranked as the single biggest operational challenge at 30%, described in the report as the number one operational challenge for the fourth year running. Retaining existing clients ranked at 7%.

Read those two together and the diagnosis looks misplaced. Agencies are not losing clients: the same report puts 62% of clients at two years or more with their agency, 43% at two to five years and a further 19% past five. The book is stable and long-lived. What is scarce is not relationships. It is the number of things each relationship is permitted to buy.

That is a surface-area problem, and it has a different remedy. A lead problem is solved with more attention at the top of a funnel. A surface-area problem is solved by changing what an account is allowed to be worth, which is an operations decision dressed as a sales one. It also explains why the remedy gets refused so often: the people who would benefit sit on the commercial side, the people who would carry it sit on the delivery side, and nobody volunteers to be the second group.

Does adding a service line actually make an agency better off?

Not automatically, and the industry evidence is uncomfortable enough that any page skipping it is selling you something.

Promethean Research's 2026 State of Digital Services report is built on 119 completed responses from digital agency owners and managers, surveyed in February 2026, with an average agency size of 31 employees. Its headline finding on service mix runs directly against the instinct to add: agencies that reduced services grew 13% on average and posted 30% net margins, while the industry-wide average after-tax net margin for 2025 was 13%. The report's own summary line is that narrowing service mixes outperformed. Focus won.

The academic work on cross-selling says something similar from another angle. Cross-selling is usually modelled as free money, on the assumption that a customer who buys more is worth more. Shah, Kumar, Qu and Chen tested that against the customer databases of five firms and found the opposite for a substantial slice: 10% to 35% of customers who cross-buy are unprofitable, and they account for 39% to 88% of the firms' total loss from its customers. The pattern is a downward spiral: customers with persistent adverse traits, such as excessive service requests or heavy promotion-chasing, get worse as they buy more. The practitioner version of the same research ran in Harvard Business Review as "The Dark Side of Cross-Selling".

Put those together and the conclusion is not "do not add a line". It is that adding is a load-bearing operational decision whose failure mode is not lack of demand. It is a line that consumes more of the business than it returns, sold hardest to the clients who were already expensive to serve.

Which services are adjacent, and which are a second business wearing your logo?

The test is not whether the service is related in a category sense. It is whether you are already having the conversation.

A performance agency spends part of every quarterly review on leads that were generated and never called back. That argument is already in the room, it is already partly the agency's problem in the client's eyes, and a workflow that answers the phone attaches to it without any new education. Bookkeeping for the same client does not, however logically adjacent it looks on a services page. Nobody has ever blamed their media buyer for a late reconciliation.

The distinction is older than the category. Ansoff set it out in 1957 in the article that introduced the product-market grid: "There are four basic growth alternatives open to a business. It can grow through increased market penetration, through market development, through product development, or through diversification." Ansoff's market axis is deliberately not the customer. He defines it as a product's mission, the job the product is meant to perform, precisely because "a customer usually has many different missions, each requiring a different product", and he defines product development as the strategy that retains the present mission. So the familiar agency comfort, that a new service sold to an existing client is safely product development, does not follow from the grid. If the new service performs a job you were never performing for that client, Ansoff puts it in the diversification box however familiar the logo on the invoice, and he was explicit about what that costs.

“Much more than other growth alternatives, they require a break with past patterns and traditions of a company and an entry onto new and uncharted paths.”

Most service-line proposals pitched to agencies are diversification described as product development. The giveaway is the sales motion: if the line needs its own positioning and its own audience before anyone will buy it, it is a second business with shared branding and should be resourced like one. If it can be raised in an existing account review as a fix for a complaint the client has already made, it is an adjacency. Our page on appointment setting for marketing agencies covers what that looks like on the performance side.

Where does the capacity to deliver a second line actually come from?

This is the question that quietly decides the outcome, and it is the one most proposals answer with a shrug.

A line delivered by the people who already deliver the current service is not new revenue. It is the same team doing more work. The AgencyAnalytics data puts 68% of agencies at 60% or more of their team's time on billable activities, with 18% at 80% or above. Those are not organisations with a spare afternoon. The same survey lists allocating time and billable expenses as the top team-management challenge at 28%, with employee burnout and upskilling employees tied at 15% each.

What happens next is predictable and worse than a slow launch. The new line borrows hours from the old one, the original service slips, and the client paying for it reconsiders an invoice they were previously happy with. The agency has not added revenue. It has traded a proven service for an unproven one, and the trade only becomes visible when the renewal conversation arrives.

There are three honest answers. You hire for it, which is a real investment with a payback period, a hiring risk, and a cost profile that is heavier in year two than in year one: what building it in-house actually costs by the second year. You stop doing something else, which is the Promethean finding in disguise and is often the right call. Or the capacity comes from outside the agency entirely, the only one of the three that does not compete with work you are already being paid for. The trade-offs between operating a platform yourself and buying a line that arrives already operated are in done for you versus do it yourself and in in-house versus managed service.

Who carries a phone line at 2am?

Every service an agency currently sells is asynchronous. A report is late, a campaign underperforms, a page renders wrongly, and each is repaired during working hours by somebody who read about it in a queue. Nothing an agency delivers has to be correct at three in the morning on a Sunday.

A phone number is a different category of object. It is either answering or it is not, and the person who discovers the difference is the client's customer rather than the client. The failure is public, it is immediate, and it lands on the brand whose name is on the line.

Round-the-clock cover is a staffing problem with a floor under it, and the clearest public working of that arithmetic is in Google's Site Reliability Engineering chapter on being on call. It reports that dealing with an on-call incident, including root-cause analysis, remediation and follow-up, takes about six hours, and concludes that "the maximum number of incidents per day is 2 per 12-hour on-call shift". On the staffing side it is blunter still: "the minimum number of engineers needed for on-call duty from a single-site team is eight", on a rule that no more than 25% of an engineer's time should be spent on call.

Almost half the AgencyAnalytics sample, 49%, lead agencies of ten full-time employees or fewer. Those agencies do not have an eight-person rotation to spare, and building one is not a service line, it is a second company. This is the item that converts an attractive idea into a bad quarter, and a candid proposal names it first rather than last. It is also why the reliability of the underlying system stops being a technical curiosity and becomes your commercial exposure: where the agent's rules live and what is verified before it ever speaks to a caller determine how many 2am events there are to carry in the first place.

Why is the first sale of a new service line an internal one?

Because the person who has to raise it on the next client call is the first buyer, and nobody sells something they cannot explain or support.

This is documented in the new-product literature, and it is the part of a launch plan that is easiest to leave out. Reviewing what makes a launch succeed inside a sales force, Atuahene-Gima wrote that some firms take commitment as a given, "seemingly adopting the attitude, 'If we build it, they will sell'", when management has no such guarantee: salespeople may simply not sell a new product, or may misrepresent its benefits for a short-term result. The recommendation is the useful part: treat the people who have to sell it as a first line of customers, and launch to them as seriously as to the market.

For an agency the practical form is small and testable. Can the account manager say, without opening a document, what the service does, what it explicitly refuses to do, and who to call when it misbehaves? If they cannot, the line is not ready to meet a client, and enthusiasm at partner level will not close the gap. What they fear is not the sale. It is being handed a problem at a client meeting they have no way of resolving, and they avoid that by not raising the subject.

Is it really cheaper to sell to a client you already have?

Almost certainly yes in a professional-services relationship, and the statistic everybody quotes for it does not survive being checked. That is worth saying out loud, because the rest of this article leans on evidence and it would be dishonest to lean on this one.

The claim is that acquiring a customer costs five times more than retaining one. Chase it back and what you find is not a study but an attribution. Keiningham, Vavra, Aksoy and Wallard, in a 2005 Ipsos Loyalty excerpt from Loyalty Myths, list it as myth number eight and trace the earliest attribution they can find to research by the Technical Assistance Research Project in the late 1980s, noting that several other bodies claimed the identical finding as their own at the same time. A 1990 Harvard Business Review article and a Tom Peters bestseller lent it further credibility, and, in their words, it "has stood unchallenged for 20 or more years". They add that they had published the fallacy themselves. Their objection is not that retention is unimportant. It is that the assumptions underneath the ratio do not hold, that the balance between acquisition and retention cost moves with where a product sits in its lifecycle, and that customer bases vary far too much for one number to mean anything.

So the ratio is folklore. The specific, measured, dated version is better anyway, and it is the AgencyAnalytics figure above: 56% of agency leaders report that more than 40% of their new revenue comes from clients already on the books. That is a survey with a stated denominator, about agencies specifically, from this year. It says the same useful thing without leaning on a ratio whose attribution several organisations claimed at once and whose assumptions its own popularisers have since disowned.

We say this partly because it is the correct thing to do and partly because it is the discipline we would want applied to us. A vendor who still prefers that contested ratio to a figure with a named sample and a date is telling you how they will describe their own results later.

What shape does a service line have to take for the arithmetic to hold?

Everything above converges on one structural answer, and it has nothing to do with the specific service.

The agency keeps what it is genuinely best at and what is genuinely hard to replace: the client relationship, the account, the positioning, the judgement about what this particular client should be doing. The operating burden goes to an organisation whose entire business is that burden, because a 24/7 rotation is indivisible and you cannot buy a quarter of one. That is not a statement about capability. A capable ten-person agency still cannot staff an eight-person rotation, and pretending otherwise is how the 2am item becomes somebody's resignation.

Two things follow. The first is that the name on the product attracts more than reputation. The EU AI Act reaches for exactly this instinct in Article 25, which provides that a distributor, importer, deployer or third party is considered the provider of a high-risk AI system where "they put their name or trademark on a high-risk AI system already placed on the market or put into service, without prejudice to contractual arrangements stipulating that the obligations are otherwise allocated". Two limits sit in that sentence. Most phone agents are not high-risk systems, so the article is not automatically engaged, and the provision leaves the parties room to allocate the obligations by contract. The drafting instinct is still the point: regulation starts from the name on the product and works outwards, which is why the contract that allocates those obligations has to exist rather than be assumed. What that means for a reseller is on our EU AI Act page for white-label voice, alongside where the call recordings actually go.

The second is that the arrangement has to survive your client's procurement questions, not only yours. They will ask whether the agent writes into their own system or into a dashboard nobody opens, which is the difference between removing work and moving it: whether an agent is inside the system of record. The shape of an arrangement like this, described without commercial terms attached, is on the white-label platform page, our partner page, and, for the version that arrives already operated, what fulfilment included actually means.

How do you check any proposed service line in three questions?

Any line, from any supplier, including the one your competitor is currently excited about.

1. Who carries it at 2am? The answer has to be an organisation and an obligation, not an intention. "We would look at it first thing" is not an answer for a service that is either answering or not answering.
2. Whose capacity delivers it? If the answer is the team that already delivers your current service, you have not added a line. You have added overtime, and the bill arrives as a slipped deadline on the work that was already paying.
3. Which existing client would buy it this quarter, without a new pitch? Name them. If nobody comes to mind, the adjacency is theoretical, and a theoretical adjacency is a second business with your logo on it.

A proposal that cannot answer all three is a distraction dressed as growth. That applies to us as much as to anyone, which is why the first thing we would rather establish is whether you can name the client.

What a useful next conversation looks like

Not a demo. A working session: about forty-five minutes on one client of yours, where we take the policies that client's front desk already follows, push on them until the edge cases show themselves, and write down what their rules turn out to be. You keep that written version whether or not anything else happens between us, and it is worth having as an account document in its own right.

If something does happen next, it is deliberately small. One client, one workflow, missed calls and after hours, roughly two weeks, and nothing else moves until that one is behaving. We do not publish commercial terms on the site, for the same reason we would not quote you a scope before hearing the workflow. If you would rather hear the thing first there is a live voice demo, and if you would rather describe the account in writing, tell us about one client.