Hlib Storchak · 2026-08-21 · ~11 min readQuick answer
A new B2B agency can't rely on referrals or word of mouth, the two channels that drive most established agencies' new business, because those need delivered work behind them. The fastest honest path to your first clients is a narrow ICP, a small amount of unpaid or heavily discounted proof work, and your own cold email and LinkedIn outbound run exactly the way you'd want a client to see it. Once you have one or two happy clients, switch as much volume as you can onto the warm channels and treat outbound as the backup engine, not the permanent one.
There's no referral flywheel to lean on at day zero
I'm Hlib Storchak. I build and run outbound systems for B2B founders and sales teams, and I've booked 2000+ meetings for B2B clients doing it. Some of those clients aren't SaaS companies at all, they're founders of their own small B2B service businesses, agencies, consultancies, dev shops, who came to me after a few months of waiting for referrals that hadn't shown up yet. The advice in this article is the same thing I tell them before we ever talk about a client engagement.
Most guides to landing agency clients quietly assume you already have some. They tell you to lean on referrals, ask for testimonials, get featured in a roundup, post consistently until the algorithm rewards you. All of that is genuinely good advice once it applies to you. None of it applies on day zero, when you have delivered nothing to anyone and nobody has a reason to vouch for you yet. This is a playbook for that specific, uncomfortable window, not the one after it.
What the 2026 data says new business actually looks like
AgencyAnalytics' 2026 Marketing Agency Benchmarks Report, a Typeform survey of 494 agency professionals run between February and April 2026, asked agencies where their new business actually comes from. Referrals led at 85% and word of mouth at 72%, both down from 94% and 78% in 2025, which the report reads as agencies broadening their acquisition mix rather than personal networks losing relevance outright. SEO and content marketing followed at 38%, outbound outreach at 33%, digital advertising at 26%, social media at 19%, and a newer entrant, AI-driven search results, at 11%.
Read that table again as someone with zero clients instead of someone with an established book of business, and the two biggest bars on the chart are unavailable to you. You have no past clients to refer you and no track record for anyone to mention in passing. What's left is the bottom half of the list, and outbound outreach is the only one of those that you can start today, on your own time, without waiting for an algorithm or a search engine to notice you exist.
| New business source (AgencyAnalytics, 2026) | % of agencies using it | Available at zero clients? |
|---|---|---|
| Referrals | 85% (down from 94% in 2025) | No, needs delivered work first |
| Word of mouth | 72% (down from 78%) | No, needs a network that has seen you work |
| SEO and content marketing | 38% | Slow, usually months before it compounds |
| Outbound outreach | 33% | Yes, from the first day |
| Digital advertising | 26% | Yes, but pay-to-play with no earned trust yet |
| Social media marketing | 19% | Yes, if you're already posting consistently |
| AI-driven search results | 11% | No, needs indexed history to surface at all |
Tip. Don't treat this table as "outbound is the best channel." It's the 33% of established agencies that already have referrals and content working for them too. The honest read is narrower: outbound is the one channel on this list a founder with nothing yet can turn on unilaterally, this week, at close to zero cash cost.
Step 1: pick one narrow ICP before you send anything
The single biggest mistake I see new agency founders make isn't a bad subject line, it's an ICP so broad it can't be tested. "B2B companies that need more leads" isn't an ICP, it's every company on earth with a sales team. Pick one: a company size band, an industry, a specific trigger like a recent funding round or a new VP of Sales hire, something narrow enough that you can write one email that sounds like it was written for exactly that reader, not a mail merge. Narrow also means small, and small is fine at this stage. You need 100 to 200 well-matched accounts to run a real test, not 10,000 loosely matched ones.
Narrow ICPs also make your own outbound more honest. If your offer is genuinely built around one kind of buyer's problem, the email writes itself faster and reads less like a template, which matters more for reply rate than any subject line trick.
Step 2: earn proof before you have client proof
You can't cite a client result you don't have yet, and inventing one is the fastest way to torch a brand-new reputation before it exists. What you can do is generate a small amount of real, checkable proof before your first paid client: a free or heavily discounted pilot for one or two accounts in your target ICP, a public teardown of a problem in your niche that shows real expertise rather than generic advice, or documented work from a past employer if it's honestly attributable and relevant. The goal isn't a glossy case study, it's one specific, believable sentence you can put in a cold email that isn't "trust me."
Keep the pilot small and time-boxed, two to four weeks, with a clear deliverable both sides agree on up front. An open-ended free engagement rarely converts to paid work, it just becomes free work with no end date.
Step 3: run your own cold email the way you'd want a client's to look
This is the part most new agency founders skip or rush, and it's the one I actually watch most closely with clients who came from running their own scrappy version of this first. If you're going to sell outbound, sales enablement, or any adjacent service, your own cold email is the first thing a prospect judges you on, whether you say so or not. That means a dedicated sending domain separate from your main inbox, proper SPF, DKIM and DMARC set up before you send a single message, and a short list you've actually researched rather than a scraped mega-list. For infrastructure and sending at this stage, I default to Salesforge and its Infraforge and Warmforge stack because it's what I run for clients day to day, but Instantly, Smartlead, and Lemlist all cover the same job well. Pick one, set it up properly, and don't send from your personal Gmail.
Keep volume modest and personal at this size, 20 to 40 emails a day from one or two mailboxes is plenty for a founder testing an ICP alone. Instantly's 2026 Cold Email Benchmark Report puts the platform-wide average cold email reply rate at 3.43%. At founder-led volume with a genuinely narrow ICP you can usually do better than that average, but plan around it, not around the outlier numbers vendors put in their marketing.
Step 4: pair it with founder-led LinkedIn, not automation at scale
Cold email gets the message delivered. LinkedIn gets you recognized when it lands, because a prospect who's seen your name or a post from you before opening your email is a warmer read than a total stranger. Expandi's 2026 benchmark, built on 13.2 million connection requests, puts average connection acceptance at 28.5% and post-connect message reply at 10.4%. At this stage, run it yourself, manually, at low volume, connecting with the same narrow list you're emailing, rather than automating hundreds of connection requests a day on a brand-new profile with no history. A restricted or flagged founder profile in month one is a much worse outcome than a slower start.
Post occasionally about the specific problem you solve, not generic industry commentary. You're not building a personal brand yet, you're building enough of a paper trail that a prospect who searches your name before replying finds something real.
Step 5: price the first few deals to buy proof, not margin
Your first three to five paying clients aren't where you make your margin, they're where you buy the case studies, testimonials, and referral base that make every deal after them easier. That doesn't mean working for free or wildly underpricing, it means being honest with yourself that the goal of deal one is a client who'll let you use their name and result once you've delivered, not the highest number you can extract. A smaller scope with a clear, deliverable outcome beats a bigger scope you can't execute well enough to earn the testimonial.
Say so plainly to the prospect if it's true: "you'd be one of my first clients, here's what that means for price and for how closely I'll be paying attention." Founders respond better to that honesty than to an agency pretending it has ten years of case studies it doesn't have yet.
Step 6: turn the warm channels on the moment you have one client
The day you deliver a real result for your first client, your channel mix should start shifting back toward the AgencyAnalytics numbers in the table above, not stay parked on outbound forever. Ask for a specific, permission-based introduction, not a vague "let me know if you hear of anyone." Write the testimonial request yourself and let the client edit it down, most people are happy to approve something rather than write it from a blank page. Put the result, however modest, somewhere public. Every one of those moves either activates referrals, word of mouth, or the case-study content that eventually starts pulling search and AI-search traffic too.
Outbound doesn't go away once referrals start, it becomes the channel you keep running underneath the warm ones so growth isn't entirely dependent on who happens to mention you this month. That's the same architecture I build for established clients: outbound as the controllable floor, everything else as upside on top of it.
What the first 90 days actually cost
Here's a cost model built on stated assumptions, swap in your own numbers. Assume a solo founder spending 8 hours a week on list building, writing, and sending, valued at nothing extra since it's your own time, plus roughly $200 to $400 a month in tooling: a sending platform, a domain and mailbox provider, warmup, and a light data source for list building. Over a 90-day window that's $600 to $1,200 in hard cash cost, not counting your own time, to run a genuine test of one ICP across cold email and manual LinkedIn outreach in parallel.
Against that, one converted client at even a modest few-thousand-dollar engagement already clears the cash cost several times over, and the qualitative return, a testimonial, a case study, a warm intro two months from now, is worth more than the first invoice by itself. If 90 days and that budget produce zero replies worth a conversation, the more likely culprit is ICP or offer, not the channel, and it's worth revisiting step 1 before spending another 90 days on the same list.
Mistakes I see most often in month one
Sending from a personal or company-wide email address instead of a dedicated, warmed-up domain, which risks the founder's actual inbox if something goes wrong. Writing an ICP too broad to test, then blaming the channel when a generic message to a generic list underperforms. Automating LinkedIn connection requests at volume on a brand-new profile before it has any history to justify the activity. Giving away unlimited free work with no defined end point because saying no to a prospect feels riskier than it actually is. And treating the first cold email send as the final version instead of the first of several honest iterations on subject line, offer, and list quality.
When this actually starts compounding
Somewhere around your third to fifth delivered client, the mix genuinely starts to flip. You'll have real testimonials to put in a cold email, which changes its reply rate more than almost any copywriting change could. You'll have one or two people willing to make an introduction without being asked twice. That's the point where the AgencyAnalytics numbers, 85% referrals, 72% word of mouth, start becoming your numbers too, not just the industry's. Outbound got you to that point. It doesn't need to carry the whole business forever, but I'd keep at least a small, steady version of it running even after referrals show up, because the one thing every agency I've seen struggle shares is having let the controllable channel go quiet the moment the uncontrollable one started working.
Key takeaways
- AgencyAnalytics' 2026 benchmark (494 agencies) puts referrals at 85% and word of mouth at 72% of new business, both down from 2025, but neither is available to an agency with zero delivered clients.
- Outbound outreach, used by 33% of agencies per the same report, is the one major channel a founder can start alone, this week, without waiting on trust that hasn't been earned yet.
- Narrow the ICP before you send anything. A list of 100 to 200 well-matched accounts beats a broad list of thousands.
- Run your own cold email and LinkedIn outbound properly, dedicated domain, real authentication, low manual volume, since it's the first thing a prospect judges an outbound-adjacent agency on.
- Price the first few deals to buy proof, not margin, then deliberately shift volume back to referrals and word of mouth the moment you have a result to point to.
- A 90-day founder-led test typically costs $600 to $1,200 in tooling, assumptions you should replace with your own numbers before you commit to them.
FAQ
How do I get my first clients for a new B2B agency with no case studies yet?
Run a small, time-boxed pilot (two to four weeks) for one or two accounts at a discounted or free rate to generate real, checkable proof, then use that specific result, not a generic promise, as the anchor of your own cold email and LinkedIn outreach to a narrow ICP.
Should a new agency spend on ads or outbound first?
Outbound first, in most cases. Digital advertising works, per AgencyAnalytics' 2026 data 26% of agencies use it, but it's pay-to-play with no earned trust behind it yet, and it costs cash before it returns anything. Outbound costs mostly time at this stage and lets you test messaging and ICP directly with real replies.
How many cold emails should a solo founder send to land the first client?
Volume matters less than fit at this stage. 20 to 40 emails a day from one or two properly warmed mailboxes to a tightly matched 100 to 200 account list is enough to learn whether the ICP and offer are working, well before you'd need to scale sending infrastructure.
Is LinkedIn or cold email better for landing an agency's first clients?
Run both, manually, to the same narrow list. Cold email delivers the pitch directly. LinkedIn builds recognition so the email doesn't land from a total stranger. Neither substitutes for the other at this stage.
When should I stop doing outbound myself and start relying on referrals?
Don't fully stop, shift the mix. Once you have three to five delivered clients willing to give a testimonial or introduction, referrals and word of mouth will start carrying more of the load, per AgencyAnalytics's 2026 numbers they eventually cover 85% and 72% of new business for established agencies, but keep a smaller, steady outbound motion running underneath so growth isn't entirely dependent on who happens to mention you that month.