Business & Marketing

How I Think About Scaling an Agency From 10 Clients to 1,000

· Updated

How I Think About Scaling an Agency From 10 Clients to 1,000

The way I think about scaling an agency is to connect three systems: client acquisition, sales and fulfillment. More prospects create more conversations; more signed clients create more delivery work. Increasing one input while the next system stays full just moves the bottleneck.

If you are still doing the marketing, taking the calls and delivering the work yourself, adding more marketing is only part of the answer. I would start by finding which part cannot handle the next increase in demand, then build the people and process around that constraint.

Personalized outreach has an acquisition cost too

I put outreach on a spectrum between highly personalized work and higher-volume distribution. Personalization spends more time; distribution spends money on the system behind it. AI can change the balance, but I still want to know what each approach costs to produce a customer.

For a concrete example, imagine selling AI ad creatives to e-commerce brands. You make an ad for a specific brand before contacting its owner, then send the deliverable as an example of your work. The point is to show the service rather than make the prospect picture it from a vague description.

In the hypothetical I use, each sample costs $5 to make and 100 samples produce one client. That is $500 spent to acquire that client. Those are illustrative inputs, not a response-rate benchmark. Before increasing spend, I want to compare that acquisition cost with what the client pays and what the service costs to deliver.

Measure the result of outreach, not just its volume

On the higher-volume side, the system doing the sending is part of the work. I discuss technical setup because a bigger activity count does not automatically create more useful conversations. I would bring the decision back to the offer, the proof behind it and the cost of a booked call.

The management question is whether the economics still make sense as you increase inputs. A call-booking figure can vary with the offer and the supporting case studies. I do not want to compare two campaigns using only how many messages they sent when the actual sales opportunities are different.

Content and ads can contribute to the same sale

I have seen people tell us they watch the YouTube channel even when the first recorded click in Hyros came from an ad. That is why I am cautious about treating a self-reported channel as the whole acquisition history. The content and the ad can both have helped move the person toward a sale.

More ad spend can buy more expensive calls

I do not assume that multiplying a campaign budget produces the same multiple of sales opportunities. An ad set can work at a lower spend and become less efficient as you push it. The useful question is where the cost per result starts to deteriorate for that particular audience and creative.

Here is the arithmetic in my illustration: spending $300 a day at $100 per booked call produces three calls. If spending $3,000 a day raises the cost to $300 per call, you get ten calls, not thirty. The larger budget has increased the call count, but it has also made each call more expensive. These are hypothetical campaign inputs.

The alternative I describe is horizontal expansion: different angles for different pocket audiences, paired with landing pages that match the promise in each ad. Someone already doing cold email has a different problem from someone considering it for the first time. I would not speak to both as though they were in the same situation.

I keep an audience at the spend level where its economics work rather than keep forcing the budget upward. More angles, creative variations, landing pages and funnel types create other places to test. That is a different job from changing one budget number and hoping the same efficiency holds.

Sales capacity includes follow-up

Once marketing starts producing more conversations, the calendar can become the constraint. If a founder is already handling marketing and delivery, filling every remaining hour with calls leaves little room to do the rest of the work. Hiring salespeople also means planning how those people will manage follow-up.

In the planning example I use, I allow a closer about five calls a day rather than booking every hour. The reason is to leave space for CRM work and follow-up calls. That is a capacity assumption for the example, not a universal workload standard.

If a founder has three call slots per day and plans for ten times that availability, the target becomes thirty slots. At five calls per closer, that implies six closers. This arithmetic shows the size of the staffing problem; it does not establish how quickly those people can be hired or how many clients they will close.

Availability is more than an empty slot

I call the spread of suitable appointment choices availability variance. A prospect choosing from a sparse calendar may settle for a time that is inconvenient. More choices across the week can make it easier for the person to find a time they can actually attend.

That connects the sales calendar back to marketing economics. If fewer booked prospects attend, the cost of a conversation that actually happens becomes worse. I therefore want to look at attended calls as well as bookings when deciding whether more acquisition spend is helping.

The operating work does not stop at hiring. I describe call review, training and standups as part of managing the team. Adding calendars without managing the quality of those conversations would leave another part of the system unfinished.

Client Ascension related reading: Jake’s cold email agency account is a separate operator’s account of hiring and delivery capacity. Read it alongside this planning framework, rather than treating either account as a guaranteed path to the same client count.

Delivery needs an owner and a sequence

For agency delivery, I think in pods of the roles needed to produce the work. A content service might need scripting, editing and thumbnail work, coordinated through a client success manager. The right capacity depends on the service and how much attention each client requires.

There are different ways to connect those roles. A client success manager can have a dedicated pod, or managers can share capacity from several pods with someone coordinating the allocation. The point is to make ownership and access to the people doing the work explicit.

I distinguish that delivery role from a coordinator in a coaching business. The coordinator can route a student to an expert or to an appropriate call without being the expert on every subject. Those are different jobs, and I would not estimate their workloads as though they were interchangeable.

Map the handoffs before adding automation

I look at a delivery process as a sequence: the output of one step becomes the input of the next. Start with what happens after a client signs, then write out how the service moves from that point to the finished deliverable. That makes the dependencies visible instead of leaving them in the founder’s head.

For each step, decide whether a person or AI should do the work and who will oversee the whole sequence. Automating individual tasks still leaves coordination and oversight to solve. I want a named owner for the process across a defined group of clients, with capacity based on the actual offer.

Find the bottleneck, then remove unnecessary work

The other question I ask is whether a step needs to exist at all. In the source lesson I discuss a content-team example about dropping color grading when it held up output. The useful operating idea is to examine whether an activity is helping the result before building a bigger system around it.

If you are still the person doing everything, start with the constraint you can actually identify: producing enough opportunities, making sales conversations available or getting client work through delivery. Expanding inputs also expands the work that follows them. I want the system behind that next stage ready, rather than make the founder carry the extra volume.

Adapted by Client Ascension with AI assistance from Daniel Fazio's lesson. The first-person narrative follows the source speaker’s account. Watch the original video