How to get clients for an agency, consultancy or dev shop

Every agency owner eventually has the same bad month. A big project ends, the next one slips, and the pipeline that felt full turns out to be three people who said “let’s talk in the new year”. Then the owner types “how to get clients” into a search bar and gets a list of forty tactics with no indication of which ones produce money this quarter.

This article is the list I wish I had been handed fifteen years ago. Seven channels, in rough order of how quickly they produce signed work, with the lead time for each, the actual wording to use, and a plan for the 90-day drought. It sits at the top of the Clients hub, which follows the client from first conversation to last invoice.

One principle runs through all of it. Agency work is bought on trust. A buyer is handing over $40,000 or $400,000 and a piece of their own reputation. They will choose the option that feels least likely to embarrass them. Every channel below is a way of borrowing or building trust, and the speed of the channel depends on how much trust already exists when the conversation starts.

The seven channels and how long each one takes

The table below reflects my own experience and what I have watched other owners go through. Your numbers will differ, but the order rarely does.

Two things stand out. The channels with no cash cost are also the fastest, and most owners in a panic skip them because emailing an old client feels awkward and launching a campaign feels like work. And the slow channels are the only ones that keep producing when you stop pushing, which is why you need one of them running long before you need it.

The pipeline arithmetic

Before choosing channels, work out how many conversations you need. Here is a worked example for an 8-person shop.

Four qualified conversations a month sounds modest. Now look at your calendar for the last three months and count. Most owners I talk to find one or two, clustered in the weeks when delivery was quiet. The gap between four and one is the whole problem. The channels below are ways to close it.

Positioning comes first, because nobody can refer a generalist

“We build custom software for businesses” gives a referrer nothing to hold. Nobody hears that sentence and thinks of a specific person to introduce. Compare it with “we rebuild booking and scheduling systems for multi-location clinics”. Now every person who knows a clinic operator has a reason to mention you, and every clinic operator who hears it assumes you have seen their problem before.

A position has three parts: who you serve, what problem you fix, and what proof you have. A useful test is whether a past client could repeat it at dinner. Another test is whether it excludes anyone. If your position fits every company with a budget, it is a description of the industry.

Owners resist narrowing because they fear losing work. In practice you keep taking the general work that arrives. The position controls what you say first, what your site leads with, and what your referrers remember. A shop that is known for one thing still gets asked to do adjacent things, and at better prices, because specialists are compared with fewer alternatives. That pricing effect is covered in the article on value-based pricing.

You do not need to pick a niche forever. Look at your last ten projects. Find the three with the best margin and the happiest client. Whatever those three share, in industry or problem type, is your working position for the next twelve months.

Past-client reactivation: the fastest channel you are ignoring

Open your accounting system and export every client you have ever invoiced. For a ten-year-old agency this list is often 60 to 150 names. Most owners are in active contact with fewer than ten of them.

These people already trust you. Their systems have aged since you last spoke. Their business has changed. Some of them have moved to new companies and brought their budgets with them. A fair share of the work I have won in slow periods came from one plain email to someone I had not spoken to in two years.

Sort the list into three groups: clients where the work went well, clients where it was mixed, and clients you would prefer never to hear from again. Write to the first group individually. A message that works:

Subject: The scheduling system, two years on

Hi [NAME],

It has been about two years since we shipped the scheduling rebuild. I was looking back at it this week and wondered how it has held up as you added the new locations.

We have built a few things since then that would apply directly to your setup, mainly around reporting and the mobile side. If it would be useful, I am happy to spend 30 minutes looking at where the system is today and telling you what I would change. No charge, and no pitch unless you ask for one.

Would next week work?

What makes it work: it refers to the specific project, it offers something concrete and small, and it asks one question. Send ten of these a week. If your contact has left the company, that is two leads: the person in their new role, and their replacement who has inherited a system they do not understand.

For clients on the mixed list, a shorter note is fine. Even a 10 percent response rate on a list of 80 is eight conversations, which is two months of pipeline in the arithmetic above.

Reactivation also has a second-order benefit. Clients who hear from you twice a year refer you. Clients who have not heard from you since the final invoice have forgotten your name.

Referrals as a system

Most agencies say they get their work from referrals. What they mean is that referrals happen to them. They have no idea how many they will get next quarter because they do nothing to cause them.

A referral system has three parts.

Ask at the right moment. The best time is within two weeks of a visible success: a launch, a good set of numbers, a compliment in writing. The client’s goodwill is at its peak and the result is fresh enough to describe.

Ask for something specific. “Let me know if you hear of anyone” produces nothing because it asks the client to do your prospecting. Name the kind of company and the problem:

I am glad the launch went well. One request. We are looking for two more projects like this one this year: operations teams at companies with 50 to 300 people who are still running a core process on spreadsheets. Is there anyone in your network who fits that description? If someone comes to mind, I can write a short introduction for you to forward, so it costs you two minutes.

Make it easy. Write the forwardable email yourself. Three sentences: who you are, what you did for the referrer, why it might be relevant. The referrer adds one line and sends it. If you leave them to write the introduction from scratch, it sits in their drafts for a month.

Then track it. A spreadsheet with four columns is enough: client, date asked, names given, outcome. Aim to make one specific referral request per active client every six months. With ten active clients that is twenty requests a year. If one in four yields an introduction and one in three introductions becomes work, that is one or two extra projects a year from a habit that takes ten minutes a month.

On referral fees: paying clients for referrals usually makes them uncomfortable, because it turns a favor into a transaction. A handwritten thank-you and a genuinely good dinner works better. Paying partners is different, and I cover that next.

The other half of referrals is being referable after the project ends. A client who had a smooth start remembers it. The client onboarding process is where that impression is made.

Partnerships with adjacent agencies

An adjacent agency sells to your buyer and does something you do not. A brand studio whose clients need a web application built. A marketing agency whose client needs a data pipeline. A strategy consultancy that writes recommendations nobody on their team can implement. An infrastructure shop that gets asked for product work.

Each of these firms turns away work every month, and each has clients who ask “do you know anyone who does X?”. You want to be the name they give.

How to set one up:

1. List ten firms within reach that serve your kind of client and do not compete with you. Ten is enough.
2. Contact the owner or the head of delivery directly. The opening line that works is an offer of work in their direction: “We regularly get asked for brand and identity work and have nobody to send it to.”
3. Meet once. Agree what each of you is good at, the minimum project size worth an introduction, and what happens commercially.
4. Send them something first, even a small lead. Reciprocity does the rest.

The commercial arrangement can be simple. Three common models:

Put the arrangement in writing, even as a one-page email, and include a non-solicitation line so that neither side approaches the other’s client for the other’s service. If you subcontract, read the ownership terms carefully: who owns what you build is decided by the contract above you. The article on when to partner instead of hire covers the delivery side of these arrangements in more detail.

Expect a quarter before the first referral arrives and expect two or three of your ten partners to produce most of the work. That is a normal distribution. Keep the productive ones warm with a call every couple of months and a lead when you have one.

Directories and marketplaces, honestly

These are two different things and owners often confuse them.

Directories list agencies by specialty, location and budget range. A buyer searches, reads profiles and reviews, and contacts a shortlist of three to five. You are found by someone who is already looking.

Bidding marketplaces post jobs and collect proposals. You compete with dozens of others, often on price, often for a buyer who has never hired an agency.

When a directory is worth it

A directory listing is worth the effort when three things are true. First, buyers in your niche actually use it, which you can check by asking your last five clients how they built their shortlist. Second, you have at least five past clients who will leave a detailed, verified review, because a profile with no reviews sits at the bottom. Third, your profile states a narrow specialty and a minimum project size, so that the inquiries you receive are ones you want.

The listing is a slow channel. Plan on one to three months before inquiries start and three to six before one becomes signed work. The reviews are the asset. Ask for them the same way you ask for referrals: after a success, with a specific request, and with the link ready.

Be careful with paid placement. Some directories sell sponsored positions for meaningful monthly sums. Before paying, work out the cost per qualified conversation. If a placement costs $1,500 a month and brings two qualified conversations, each one cost $750. At a one-in-four close rate, that is $3,000 of acquisition cost per client. On a $50,000 project with a 30 percent margin, that is a fifth of your profit on the first engagement, which is acceptable if clients tend to stay and poor if they tend to leave after one project. Run a free or basic listing for three months first and see whether the inquiries match your position.

When a marketplace is worth it

Bidding marketplaces suit three situations: a new shop with no portfolio that needs its first five projects, a team with idle capacity that would otherwise earn nothing, and productized offers with a fixed scope and price. For an established agency selling $50,000 projects they are mostly a distraction. The buyers anchor on the lowest bid, the platform takes a cut, and the platform owns the relationship.

If you do use one, treat it as outbound. Ignore jobs with twenty proposals already. Reply only to posts where the buyer has described a problem you have solved before, lead with that specific experience in the first two lines, and move to a call quickly. Leave as soon as referrals can replace it.

Outbound that is specific

Cold outreach has a bad name because most of it is bad. A message that starts with “I hope this finds you well” and offers “custom software solutions” is deleted in a second. The sender bought a list of 5,000 names and wrote to none of them in particular.

Specific outbound is a different activity. You pick a small number of companies that fit your position, find one observable problem at each, and write to the person who owns that problem.

Observable problems are everywhere if you look: a booking flow that breaks on mobile, a job posting for a role that suggests a stalled internal project, a public product that has not changed in three years while competitors shipped, a customer review complaining about the same bug repeatedly. Fifteen minutes of research per company is enough.

A message that gets replies:

Subject: Your booking flow on mobile

Hi [NAME],

I tried to book an appointment at your [LOCATION] clinic on my phone this morning. The date picker fails on step three and sends the user back to the start. I recorded it; the 40-second clip is here: [LINK].

We rebuild booking systems for multi-location clinics. The last one we did cut abandoned bookings substantially for a group about your size.

If it is useful, I can send a one-page note on what is causing this and what a fix involves. No call needed.

It is short. It proves you looked. It offers something useful with no meeting attached. And it describes relevant experience in one line without a figure you cannot back up.

The numbers for specific outbound, as a worked example: 20 researched emails a week is 80 a month. If 8 percent reply, that is six or seven replies. If a third of those become qualified conversations, that is two a month, half of the four the 8-person shop needs. The time cost is about six hours a week, and it is the owner’s time or a senior person’s, because the research requires judgment.

Follow up twice, a week apart, each time adding something new: a second observation, a relevant example. Then stop. Three unanswered messages is the limit of polite.

A note for UK and EU readers: the rules on unsolicited business email are stricter than in the US and vary by country. Check what applies before you send, particularly when writing to sole traders or named individuals.

Content: slow, and the only channel that compounds

Content is writing or speaking that demonstrates you understand the buyer’s problem better than they do. It takes six to twelve months to produce a client. It is also the only channel where work you did two years ago still brings in conversations today.

What works for a services firm:

• Teardowns and write-ups of real problems. “Why multi-location scheduling breaks at the fifth location” is read by exactly the people you want. Twelve of these in a year is plenty.
• A point of view on price and process. Buyers search for how projects are priced and what goes wrong. If you explain retainer pricing or how you stop scope creep more clearly than anyone else, you have shown how you will behave on their project.
• Speaking where buyers gather. Pick events run for your client’s industry. A room of 40 clinic operators is worth more than a room of 400 developers.
• A short email to your list, monthly. Past clients, prospects, partners. One useful observation. This is the cheapest way to stay remembered and it feeds the referral and reactivation channels.

What does not work: generic posts about why businesses need a website, daily social posting with nothing to say, and anything written for other agencies when your buyers are operations directors.

If you are repositioning toward newer service lines, content is how the market finds out. The article on building an AI consulting business covers that shift specifically.

Start content when you are busy. It is the channel you will thank yourself for in the next drought and the one you will not have time to build during it.

What to do in a 90-day drought

A drought is 90 days with no meaningful new work signed and less than two months of committed revenue ahead. Here is the order I would work in.

Weeks 1 and 2: the warm list. Email every past client where the work went well, every prospect who went quiet in the last 18 months, and every former colleague now in a buying role. Individual messages, ten a day. This is the only step likely to produce cash inside 30 days.

Create an offer that can start on Monday. A drought is the wrong time to sell a six-month project with a three-month sales cycle. Package a small, fixed-price diagnostic: a two-week technical audit for $6,000, a one-week discovery sprint for $4,500, a performance review for $3,000. Small offers get signed without procurement, they get you inside the client’s business, and in my experience about half of them lead to a larger project. A lightweight statement of work keeps even the small ones clean.

Weeks 3 to 8: widen. Add the referral asks, the partner conversations and the outbound. Chase every open proposal with a direct question:

I want to keep our planning honest on this side. Is this project still likely to go ahead this quarter? A no is completely fine and more useful to me than a maybe.

Half will say no. The other half will give you a date. Both answers are worth more than a pipeline full of maybes.

Protect your terms. A drought is when owners drop the deposit, discount by 30 percent, and accept the client everyone else turned down. That is how a cash problem in the spring becomes a collections problem in the autumn. If you discount, reduce scope along with price, and keep the deposit. The article on what to do when a client is not paying describes where the alternative leads. A client who shows warning signs during the sale will not improve after signing, and the patterns are described in the piece on difficult clients.

Cut costs early. Work out your runway in weeks: cash in the bank divided by weekly fixed costs. If the number is under twelve, make the cost decisions now. An owner with sixteen weeks of runway negotiates well. An owner with four signs anything.

Use the bench. Idle people can write the teardown articles, build the internal tool you have postponed, and prepare the case studies. That work feeds the slow channels and keeps the team sharp.

The weekly habit that prevents the next drought

Droughts are caused six months earlier, during the busy stretch when nobody did any selling. The fix is boring: a fixed block of time for business development that delivery work cannot take.

Three hours a week is enough for most owners of small shops. A workable split:

• One hour: two past-client check-ins and one referral ask.
• One hour: five specific outbound emails or one partner call.
• One hour: one piece of content, or part of one.

Then count qualified conversations every month. That single number, compared with the four or five your arithmetic says you need, tells you more about the next two quarters than your current revenue does. When it drops for two months running, act then, while you are still busy and still able to choose.