The complete guide to growing your web development agency
The agencies growing right now have changed what they sell rather than how fast they sell it. This guide covers how to do that: picking a narrower niche, getting found by both search engines and AI assistants, pricing for recurring revenue, and keeping clients long enough that they’re worth the cost of winning.
A basic marketing site that cost $5,000 three years ago now sells for under $1,500. Goodfirms surveyed more than 300 agencies across 31 countries this year and found 98% of them using AI somewhere in their workflow, 60% quoting $1,000 to $3,000 for a basic site or MVP, and 71% delivering those builds in under four weeks. Meanwhile 66% of web designers surveyed by 20i said tighter client budgets were a real problem for their business last year.
That's the market. If your agency still sells hours to build brochure sites, the floor is dropping out from under you.
Table of contents Direct link to this section
- How do you choose the right niche?
- How do agencies get found now?
- How should you price agency work?
- What do clients actually expect?
- How do you win the work?
- How do you keep clients?
- How do you keep the team sharp?
How do you choose the right niche? Direct link to this section
Pick a niche narrow enough that you can name the next three companies you want to work with. If you can’t name them, it isn’t a niche, it’s a category.
We’d also recommend that you start with the clients you already have. Sort them by margin, not revenue, then look at what the top quartile share. Usually it’s not industry. It’s something more specific: they all have an in-house marketer who updates the site weekly, or they all run multi-location businesses with near-identical page templates, or they all had a bad experience with an offshore build and now over-index on responsiveness.
Those patterns are more useful than a demographic profile, because they tell you what you’re actually good at. An agency that repeatedly rescues abandoned WordPress builds has a niche whether it advertises one or not.
Once you can describe the pattern in a sentence, rewrite your site around it. Your homepage should make a prospect in that niche think “this is for me“ and make everyone else bounce. Losing the wrong leads is the point. A narrower positioning means fewer inbound enquiries and a higher close rate on the ones you get, which is a better use of your sales time than qualifying out tyre-kickers for six weeks.
Two things worth checking before you commit: is the niche big enough to sustain you, and can you reach it? A niche of forty companies worldwide is a consulting practice, not an agency. And a niche you can only reach through cold outreach will cost more to serve than one that congregates at a conference or in a Slack community.
How do agencies get found now? Direct link to this section
Three ways: referrals, search, and AI assistants. The third one is new enough that most agencies haven’t done anything about it.
Referrals are still the best channel Direct link to this section
Referrals close faster and negotiate less, and they cost nothing to acquire. Most agencies treat them as luck. Treat them as a process instead. Ask for one at the point of maximum goodwill, which is usually two weeks after launch when the client has just seen their first traffic report, not six months later when they’re annoyed about a support ticket.
Partner programs work the same way. If you build sites on a platform, get listed in that platform’s directory. We run the CloudCannon Expert Directory for exactly this reason, and the agencies who get the most out of it are the ones with a specific listing rather than a generic one.
Getting found in search means being answerable Direct link to this section
Traditional SEO still applies. Do keyword research with Google Trends, Ahrefs, or Semrush, find the questions your prospects type, and write pages that answer them properly.
What’s changed is that a growing share of those questions never reach a search results page. Someone evaluating agencies now asks ChatGPT or Perplexity “who builds Astro sites for healthcare companies in the UK“ and gets a shortlist of three names. If your agency isn’t in the training data or the retrieved sources, you’re not in that shortlist, and you’ll never see it happen in your analytics.
Being cited by an AI assistant comes down to a few practical things. Have your specifics written down in plain text on your own site: the technologies you build with, the industries you serve, the size of client you take. Publish case studies with named clients, real numbers, and dates. Structure pages so each section answers a discrete question, with the answer in the first sentence rather than buried three paragraphs into a narrative. Keep them current, because recency weighs heavily in AI retrieval and stale pages drop out of citations.
There’s a second opportunity here. Most of your clients have the same problem, and almost none of them have addressed it. Auditing a client’s site for AI visibility, then fixing the structure and content, is a service you can sell today at retainer rates.
Outbound still works if you’re specific Direct link to this section
Cold outreach fails when it’s generic. It works when you’ve done ten minutes of homework and can open with something true about the prospect’s site: their LCP is 4.2 seconds on mobile, their blog hasn’t been updated since 2023, their contact form is broken on Safari.
Conferences and trade shows in your client’s industry beat web development conferences, because your competitors are at the latter and your prospects are at the former. If you’re going to run paid ads, start at $500 to $1,000 a month, track cost per qualified enquiry rather than clicks, and give it three months before you judge it.
How should you price agency work? Direct link to this section
Move off hourly billing and toward retainers and value-based pricing. Hourly billing punishes you for getting faster, and everybody is getting faster: 71% of the agencies Goodfirms surveyed now turn around a basic site or MVP in under four weeks, and AI site builders finish nearly two thirds of their jobs inside two weeks.
Retainers are where the sustainable money is. Goodfirms found that nearly 58% of the market charges monthly retainers between $500 and $3,000. That’s not glamorous, but ten retainers at $2,000 is $240,000 of predictable annual revenue that doesn’t depend on winning a new project every quarter.
The trick is making retainers worth paying for. A retainer that covers “up to 5 hours of changes“ is a maintenance contract, and clients cancel maintenance contracts the first time cash gets tight. A retainer that covers performance monitoring, quarterly content reviews, conversion testing, and AI visibility audits is a growth service, and those get renewed.
Value-based pricing takes more nerve. You set the price against the outcome rather than the effort, which means you need to know the outcome before you quote. If a client’s site converts at 1.2% and does $2M in revenue, taking that to 1.8% is worth $1M a year to them. Charging $40,000 for that work is a bargain and both of you know it. (Here are 22 ways to deliver more value to your web development clients if you need a starting point.)
To make either work you have to measure and report. Track Core Web Vitals, organic traffic, conversion rate, and whatever else the client’s business actually runs on. Send the numbers monthly whether or not they ask. Clients who can see what they’re getting don’t renegotiate at renewal.
One warning. Raising prices without changing your offer is how agencies lose their best accounts. Raise the price and the scope at the same time.
What do clients actually expect? Direct link to this section
More for the same budget. Goodfirms’ data shows that while simple sites collapsed in price, complex builds held steady or went up, because buyers now treat things that used to be add-ons as table stakes.
The specific expectations that have shifted:
- Performance is assumed, not sold. Nobody pays extra for a fast site any more. They just leave if it’s slow.
- Accessibility to WCAG 2.2 AA is increasingly contractual, especially for anyone selling into government, education, or the EU.
- Analytics has to be instrumented properly from day one, not bolted on when someone asks for a report.
- Clients expect to edit their own content without filing a ticket and waiting two days.
You’ll also field a version of this question in almost every sales call now: “couldn’t AI just build this?“ It’s a fair question, these days, and it deserves a straight answer rather than defensiveness. AI writes code quickly. It doesn’t decide what the site needs to do, own the consequences when it’s wrong, or answer the phone in eighteen months if something breaks. The agencies losing this argument are the ones whose deliverable really was just typing.
The clearest way to demonstrate the difference is to show the client the parts of the work AI can’t touch: the discovery that reframed their whole information architecture, the accessibility audit that kept them out of a lawsuit, the performance work that halved their bounce rate.
How do you win the work? Direct link to this section
Listen more than you pitch. That’s more a sales technique than it is a personality trait, and it’s learnable.
Most losing proposals fail because the agency answered a question the client didn’t ask. Before quoting anything, get the client to tell you what happens if the project succeeds and what happens if it doesn’t. The second answer is usually more useful. A client who says “if this doesn’t work we lose the contract with our biggest distributor“ has just handed you your pricing, your timeline, and your entire pitch.
Ask about the last agency they worked with and what went wrong. You’ll learn what to avoid and what they’re anxious about, and anxious clients buy reassurance more readily than they buy features.
Then write the proposal in their words, not yours. If they said “our team can never update anything without calling someone“, your proposal should say that back to them before it says anything about static site generators. Technical detail belongs in the appendix.
Two habits that pay off consistently: send the proposal live on a call instead of emailing it and hoping, and always quote three options rather than one. Three options changes the conversation from whether to hire you to which version to buy.
How do you keep clients? Direct link to this section
Set expectations you can beat, then make the handoff good enough that the client doesn’t need you for trivia.
Retention is mostly a communication problem. Clients rarely leave because the work was bad. They leave because they didn’t know what was happening for three weeks, or a launch date moved and nobody told them until the day before. Give larger projects a single point of contact, and when something slips, say so early with an alternative rather than late with an apology.
Build quality assurance into the process rather than the last week of it. At minimum, before any launch, check that the site:
- Works without errors across browsers and devices
- Meets WCAG 2.2 AA and your client’s data privacy obligations
- Hits your performance targets on a real mobile connection, not a desktop test
- Has analytics and conversion tracking firing correctly
- Has a documented rollback state if something goes wrong
Handoff is where a lot of agency relationships break. If your client has to email you to fix a typo, one of two things happens: they resent paying for it, or you stop responding quickly and they resent that instead. Neither ends well.
This is the argument for giving clients real editing access to their own site. With a Git-based CMS like CloudCannon your team keeps the codebase, the build process, and the version history, while the client gets a Visual Editor they can use without touching anything they shouldn’t. Custom Permission Groups let you decide exactly what each person on their side can change. Client Sharing lets them review a build before it goes live without you exporting screenshots into a slide deck. And View As mode lets you see the whole CMS using their permission set before you hand off.
The commercial effect is that your retainer stops being about typos and starts being about strategy, which is the version clients keep paying for.
Finally, ask for feedback in a structured way. A ten-minute call sixty days after launch will tell you more than any survey. Ask what was frustrating, not whether they were satisfied. Then publish the good parts, with permission, as case studies with real names and real numbers.
How do you keep the team sharp? Direct link to this section
Book the time. Professional development that isn’t on the calendar doesn’t happen, because client work always looks more urgent.
Give people a fixed budget and fixed hours, then let them choose. A few places worth the money:
- Frontend Masters for deep front-end work
- Google Skillshop for analytics and ads certification, which is free
- HubSpot Academy for the sales and marketing side, also free
- Institute of Code for structured web development courses
Run internal sessions where whoever solved something interesting last month explains it to everyone else. It’s cheap, it spreads knowledge that would otherwise leave with one person, and it surfaces skill gaps faster than a performance review.
Pair the two people on your team with the widest skill gap on the same project occasionally. It slows that project down and speeds everything after it up.
And look at what you’re building on. If your stack means every client change is a developer ticket, no amount of training fixes the economics of that. Sometimes the constraint on growth is a tooling decision from four years ago.
Scale without the headaches
If you’re building sites for clients and want your agency’s work to scale without your support load scaling with it, the CloudCannon Partner Program gives your team commercial benefits on every client site you bring across, plus access to our partner support channel.
