How to Get Clients for Your Agency: The Complete Playbook

Every agency question eventually reduces to this one. Positioning, pricing, delivery, tools: all of it is decoration on the only machine that matters, which is the one that turns strangers into clients. This is the complete version of that machine, the same system taught in our free Academy, compressed into one article you can execute from.

A warning before the tactics: none of this works as a one-week sprint. It works as a system you run for an hour a day, every working day, whether you feel like it or not. The agencies that stay small do not lack talent. They lack a pipeline habit.

Start with the math, not the mood

Sales stops being emotional the moment it becomes arithmetic. Before you send a single message, know roughly what it takes to create one client, because that number turns "I hope someone says yes" into "I am forty messages away."

Realistic starting numbers for well-targeted, personalized outreach to local businesses:

• Roughly one reply per ten to fifteen good messages. Generic blasts do far worse.
• About one booked call per two to three replies, if booking is effortless.
• One close per three to four calls early on, improving as your calls improve.
• Net: on the order of 75 to 150 quality touches per client when you start.

Work it backward. Two new clients a month at starting conversion rates is roughly 10 to 15 touches per weekday. Each personalized message takes a few minutes once your list exists. That is the whole secret: not talent, an hour a day with a list.

Build a list worth working

A pipeline starts with a list, and for local niches this is almost unfairly easy, because the prospects are public, listed, and conveniently ranked by how badly they need you.

Where to find prospects

• Local search and maps: "[niche] in [city]". The businesses ranked below the top handful need you most, and the map pack shows exactly who is winning and who is invisible.
• Ad libraries (Meta and Google): a business already paying for ads understands marketing spend. A broken site underneath a live ad campaign is the single best prospect profile that exists.
• Industry directories and trade association member lists. Every member is a warm-ish prospect.
• Adjacent referrers: suppliers, distributors, and complementary trades who serve your niche all day and know who is growing.

What to capture for each one

• Business name, the owner's name if findable, phone, email, and the current website URL.
• One specific, true problem: slow load, broken on mobile, no call to action, dead SSL, a five-year-old copyright date, no booking option. This note becomes your opener.
• A temperature score: hot (running ads plus a broken site), warm (visible problem), cool (fine site, future timing play). Work hot to cool.

Fifty prospects with a specific note each beat five hundred names with nothing to say. And this is exactly the kind of work to hand to an AI agent: pull the businesses for a niche and city, check each site for speed and mobile rendering and forms, find the owner's name, and output a scored table with a one-line problem note per prospect. You review the notes, because you are the one who has to say them out loud later. Keep the list in one place with a next-action date per prospect; the leads pipeline or a single sheet both work. A prospect with no next action is not a prospect, it is a name.

Outreach that gets replies

Cold outreach fails for one reason: it is about the sender. "Hi, I build websites, want one?" gets deleted because it is generic and self-centered. Outreach that works opens with something specific and true about them, ties it to money, and makes a small ask.

The four-line structure

1. A specific observation about their business. Proof you actually looked.
2. What it costs them, in money terms. Missed calls, lost jobs, wasted ad spend.
3. A small, easy ask. Not "buy a website." A two-minute video, a quick look, a yes or no question.
4. Optionally, one line on who you are, in their language.

In practice, that reads something like: "Hi Mike, I pulled up your site on my phone while looking at roofers in Fairfield. It takes about eight seconds to load and the phone number is not tappable, which usually means people give up and call the next roofer on the list. I recorded a two-minute video showing the three things costing you calls. Want me to send it over?"

Notice what it does not do: no "hope this finds you well," no paragraph about your passion for design, no link dump, no meeting request. The only decision the reader makes is "do I want the video," and that is an easy yes because it costs them nothing.

The rules that make it work

• One prospect, one specific observation. If the first line could be sent to anyone, delete it and start over.
• Short. Four to six sentences on email, two or three in a DM. Respect the phone screen it will be read on.
• Ask for a micro-yes, never a meeting. The meeting comes two messages later.
• Plain text, no images, no attachments, one link at most. You are a person, not a campaign.
• Send from your normal address at normal volumes. Deliverability infrastructure is a scale problem; you have a starting problem.

Follow-up is where the replies live

Most replies come from the second, third, or fourth touch. Silence is not rejection; it is a busy owner on a roof. Follow up three to four times over two weeks, and make each touch add something new instead of "just bumping this." Day three: send the video anyway. Day seven: a new observation, like a competitor with worse reviews outranking them. Day fourteen: a graceful close that leaves the door open and your calendar link on the table.

The teardown video: your best cheap asset

A teardown is a two-to-three minute screen recording of you looking at a prospect's website and narrating what is costing them customers. It converts because it is proof of competence they can watch with zero commitment, and because nobody else in their inbox is doing it.

The structure: fifteen seconds on who you are and why you recorded this, then three specific problems in order of money lost, then one sentence on what fixed looks like, then a soft ask. Show, do not lecture: load their site on a phone simulator, click the broken form, run the speed test on camera. Watching it fail is worth a thousand adjectives. After ten of these you will record one in five minutes.

The demo-site hook: the strongest opener in local outreach

Everything above works. This is what makes it unfair. Instead of telling a prospect their site is costing them customers, you send a link to a better one, already built, with their name on it.

It works for three reasons. It collapses imagination: the owner does not have to picture "a better website," they are looking at it, with their logo, their services, their photos. It is proof instead of promise: every agency claims quality, you demonstrated it before asking for a dollar. And it creates gentle ownership: once they have seen "their" new site, going back to the old one stings.

Traditionally a spec build cost a designer-week, which is why nobody did this. With an agent working from a proven boilerplate, a credible demo costs well under an hour of your attention: it pulls their services, reviews, photos, and service area from public information and assembles the pages while you do something else. You spot-check it and fix the details a human notices. Ten demos a week is a realistic solo pace.

How to run it

• Aim it. Run the hook on your hottest prospects only: real business, real ticket size, visibly broken site. Your review time is the cost.
• Build on a preview link, watermarked or on your subdomain. Good enough to want, not finished enough to take.
• Send it with the same four-line structure, with the link as the ask, and be honest that it is a draft built from public information.
• On the call, walk through THEIR demo instead of a slide deck. The conversation becomes "what would we change about this," which is a buying conversation.
• If they buy, the demo becomes the first draft and you are days from launch. If not, unpublish and move on. No sunk cost.

The discovery call: questions, then a prescription

The biggest mistake on a sales call is pitching. A great first call is mostly questions. Set the frame in the first minute: you will ask about the business, then if you think you can help you will say exactly what you would do and what it costs, and if you are not the right fit you will say that too. That frame earns you permission to ask, promises a concrete answer, and signals you can walk away, which is the most trust-building move in sales.

The five questions

1. Where is the business at right now, and what is working? Listen for how they get customers today.
2. What made you take this call? This surfaces the real trigger: a slow season, a competitor, wasted ad spend.
3. What would "this worked" look like six months from now? Their answer is the outcome you sell back to them.
4. What is a new customer actually worth to you? Owners know this number, and it sets up your pricing anchor.
5. What has stopped you from fixing this already? Whatever they say is the objection you will handle later, named in advance.

Reflect it back before you prescribe: "So the ads are working, the site loses people, and a booked job is worth about eight grand." Proving you listened is the most persuasive thing you can do. Then the pitch is three sentences: here is what I would build, here is what it costs, here is when it is live.

Qualify while you listen

A deal needs three things: a real problem they agree exists, real money that can pay your price without trauma, and a real reason to act this quarter. Two out of three is a nurture, not a deal; schedule the future touch and spend your energy on the threes. And walk away from price shoppers, no-shows, and committee decisions with no owner in the room. A bad client costs you three good ones.

Close with logistics, not tricks

If discovery was honest and the offer is clear, closing is mostly making "yes" the path of least resistance. Send the proposal the same day as the call, while it is warm: their problem in their own words, exactly what is included, one price or at most two options, dates instead of durations, and one clear next step. One page. Nobody reads page seven; they skim for the price and stall.

Then ask for the deposit plainly: half up front to lock the build slot, invoice sent now, kickoff this week. No deposit, no work. Not the logo mockup "while we sort payment," not "kick off Monday, invoice Friday." Paid clients show up, send content, and treat the project as real. If you invoice through your platform, paying is one click and you see the moment it lands.

Mechanically, the whole close should take fifteen minutes, not an evening: your package already contains the scope and price, your call notes contain their words, and a quote-to-proposal flow turns one into the other and the acceptance into an invoice.

The machine, assembled

1. Fifty-prospect list with a specific problem note per prospect. Refreshed as it burns down.
2. Ten to fifteen personalized touches every weekday, four-line structure, small ask.
3. Follow-ups on day 3, 7, and 14, each adding something new.
4. Teardown videos for warm prospects, demo sites for hot ones.
5. Discovery calls that are questions, then a prescription.
6. Same-day proposal, deposit before work, four numbers tracked every Friday.

Run that loop for a quarter and you will not have a client problem. Most people will not run it, which is precisely why it works for the ones who do.

If you want it taught properly, with every script, the objection answers, and the demo-site walkthrough, the Academy sales modules cover this system lesson by lesson, free. And when the clients land, the platform gives them a portal with your name on it.