Launching a B2B SaaS on a Small Marketing Budget: What Worked and What Didn't

In July 2026 my brother and I took Mirox out of beta and opened it to the public. Mirox is an AI that manages Amazon advertising bids for sellers and agencies. It's a niche B2B product with a skeptical audience, and we launched it with a total marketing budget of roughly €2,000 to €3,000.

That's not a typo, and it's not a growth-hack humblebrag. It's just what we had. We're two founders with no newsletter, a small LinkedIn following and no audience of our own.

Most launch write-ups come from companies that already won, so every decision looks smart in hindsight. This one doesn't. We're a few months in and still early. I'll tell you what got us real conversations, what wasted time or money, and what I'd do differently. Some of the "didn't work" list is embarrassing. That's the useful part.

• Didn't: cold outreach that asked for too much, paid search before the basics were set up, pilots that ignored switching cost and asking for stage time without an audience.
• On a small budget, one warm introduction is worth more than a month of clicks. But a conversation isn't a customer. Count both separately.
• We moved our budget from buying clicks to earning links, reviews and placements. Buyers increasingly research through AI chatbots and review sites, not just search ads.
• Our most expensive mistakes cost almost no money. They cost trust and time.

1. The setup: what "small budget" actually meant

For context, the typical private B2B SaaS company spends a median of 8% of its annual recurring revenue on marketing, according to SaaS Capital's 2026 survey of more than 1,000 companies. That benchmark assumes you have recurring revenue to take 8% of. At launch, we didn't.

Here's what a budget like ours buys in the most obvious channel. WordStream's 2026 benchmarks, covering 13,474 US search campaigns, put the median cost per lead for business services at $93.69.

So we couldn't buy our way in. Everything below follows from that: if money is the scarce resource, you spend time, relationships and credibility instead. That's also why some of our mistakes hurt so much, because those are exactly the things you can't afford to waste.

Our other constraints:

• A skeptical buyer. We ask Amazon sellers to let software control their ad spend. Trust is the whole sale.
• A niche audience. Sellers and agencies with meaningful ad budgets. That's not a mass market you can reach with broad social ads.
• No owned audience. No newsletter, no community, a modest LinkedIn following.
• EU company, global market. We're incorporated in Portugal, which matters for what you're legally allowed to do in outreach (more on that below).

2. The first four months at a glance

3. What worked

✅ Joining someone else's network

In August I got a LinkedIn-style message inviting me into a free, invite-only WhatsApp group for ecommerce founders, run by someone who connects agencies, tools and brands. My first reaction was "this is spam." It wasn't. After one call, the introductions started.

Why it worked: every introduction came with borrowed trust. An agency founder takes a call from someone their contact vouched for. They don't answer a cold DM from a stranger with an AI tool. On a small budget, borrowed trust is the cheapest thing you can get.

What I'd watch: reach numbers in community pitches are often fuzzy. The ones we were quoted changed between messages. Judge a network by the introductions it produces, not the audience size it claims. Also, some "intro" services are paid, like podcast placements. That's fine, just know which ones are which before you say yes.

✅ Publishing the one thing only we could show

Our product's difference is that it logs the reasoning behind every bid change: what data it saw, which safety limits it checked, what it did and why. We published a piece showing exactly that, with the "glass box" instead of a black box.

When we were introduced to one larger agency with a 15+ person in-house PPC team, that piece is what caught their interest. On the call, they told us their biggest time sink isn't bidding. It's reporting to clients why bids moved and arguing about it afterwards. We'd never have learned that from a landing page test.

Lesson: on a small budget you can't out-shout anyone, but you can be specific. Generic content ("10 PPC tips") competes with everyone. Content that shows the one artifact only your product produces competes with no one.

✅ A partner program people can join without us

Partners and affiliates sign up on their own, get a referral link and generate their own coupon code. No call, no contract back-and-forth. The terms are simple: 100% of the first month's fee (capped at $300), then 25% for months 2 to 12. When a partner wanted their own name on the code instead of a generic one, we simply switched it.

One decision I'm glad we made: with a small service-provider partnership, the standard commission would have been about 10% of a €39 subscription. That's not worth the tracking effort on either side. So we waived our commission and gave their customers a 15% recurring discount instead. Partners remember who made it easy for them.

✅ An honest free trial

New users start with 30 days of Simulation Mode. The AI runs read-only on their real account and records every decision it would make, without changing a single live bid. At the end they get every decision where it disagreed with their current bids, with the full reasoning and a CSV export.

What we deliberately don't give them is a projected "you would have saved €X" number. We had one, and we removed it, because a forecast subtracted from a measurement isn't a result. For a trust-first product, an offer that a skeptic can check is worth more than one that sounds bigger.

4. What didn't

❌ Cold outreach that asked for too much

We made three mistakes here, in order of how much they hurt:

1. Asking for account access in a first message. Our product needs access to a seller's Amazon Ads account. Asking for that access in a cold message looks exactly like a phishing attempt. What works better as a first touch is something they can look at without granting anything, like an anonymized example of a decision log.
2. "Join as a founding member and help us co-design." We thought this was flattering. To a busy seller it reads as: please do unpaid work for a company you've never heard of.
3. Cold email in Germany. Before scaling email outreach to German Amazon sellers, we checked the law. Unsolicited B2B marketing email without prior consent is generally not allowed in Germany (§ 7 UWG, plus GDPR), however you found the address, and warning letters ("Abmahnungen") are a real enforcement risk. We switched to an opt-in funnel. If you're selling into the EU, check this before you build a sequence, not after. (This isn't legal advice. Ask a lawyer for your case.)

❌ Paying for clicks before the basics were in place

This one is embarrassing. We started Google Ads with one catch-all campaign before our measurement was in order. We found out later that Google Search Console had never been set up for our domain. We'd been buying traffic without seeing how we did in organic search.

What we changed:

• Paused the catch-all campaign and rebuilt it as three: brand defence, high intent and competitor searches.
• Set up conversion tracking with clear actions: free signup, audit request and agency enquiry, plus paid signups imported. Agency enquiries are tracked but don't steer bidding, because their sales cycle is too long to optimize for.
• Stayed on manual bidding until we have roughly 30 conversions a month. Automated bidding needs data we don't have yet.
• Set up Search Console properly. (Tip: if your DNS is hosted somewhere other than your registrar, the verification record has to go there. That cost us a failed first attempt.)

The irony isn't lost on me. We sell software that manages ad bids, and our own first campaign broke rules we'd tell any customer to follow.

❌ Pilots that ignored switching cost

The agency that got interested through our glass-box piece proposed a pilot: Simulation Mode on 2 or 3 client accounts for 30 days, comparing our decisions with their team's. Then, after an internal discussion, they declined. The reason wasn't price or the product. They didn't want to switch internal tools, because learning another tool costs the team efficiency.

That's a lesson about agencies, not about that agency. A 15-person team with working processes doesn't buy "better." It buys "better without retraining everyone." We now treat adoption cost as part of the product. Agencies told us exactly what lowers it: exports per keyword and date range, and spelled-out names in reports instead of internal abbreviations. To their credit, they suggested a joint webinar instead, where we show the tool live on real data. A "no" that turns into a distribution offer is still worth having.

❌ Asking for stage time without an audience

We were offered the chance to speak in a well-attended ecommerce webinar series, and then it didn't happen. The organizer's criteria included things like an own newsletter and a decent LinkedIn reach. We had neither. I told them openly we'd come back once we did.

The takeaway is uncomfortable but simple: distribution partners want distribution back. If you have zero owned audience, start building one on day one, even a tiny one. It's the entry ticket to the cheap channels. The same applied to paid podcast introductions. They're a legitimate service, but at our stage that budget was better spent elsewhere.

❌ Pricing and positioning we had to undo

Four changes we made within weeks of launch, each one a mistake we could have avoided:

None of these cost much money. All of them cost a little credibility with the people who saw the "before" version. Check your pricing math against your customer's budget before you publish it.

5. The scorecard

6. Where the budget goes now

In September we wrote a 90-day plan and made one big call: move paid budget away from buying clicks and toward earning links, reviews and placements. Two numbers convinced me.

First, G2's March 2026 survey of 1,076 B2B decision-makers found that 51% of software buyers now start their research with an AI chatbot more often than with Google, up from 29% in April 2025. When those chatbots answer, 45% of buyers say citations from software review sites are the most confidence-inspiring signal. Second, Gartner found that 67% of B2B buyers prefer a rep-free buying experience. They want to research and decide on their own.

Put together: the buyer researches alone, often through an AI assistant, and trusts what reviews and credible third-party sites say about you. A click you bought today disappears tomorrow. A link, a review or a mention on a respected site keeps working, and it's increasingly what AI answers are built from.

In practice that means content and link swaps with people whose audience overlaps ours (yes, like this post), working toward our first reviews on software review sites, attending events on a regular ticket instead of sponsoring them, and newsletter features through partners.

7. What I'd do differently from day one

• ☐ Set up measurement before spending a cent: Search Console, analytics, conversion tracking, internal traffic excluded.
• ☐ Start an owned audience on day one, even if it's 40 people. It's the ticket to the cheap channels.
• ☐ Join two or three communities where your buyers' service providers hang out, and ask for introductions, not attention.
• ☐ Make a first-touch asset that asks for nothing: a sample report, an anonymized example, a teardown.
• ☐ Check outreach law in every market you email before you write the sequence.
• ☐ Run the pricing math against your customer's budget, not against competitor prices.
• ☐ Design for adoption cost if you sell to teams: exports, plain labels, no retraining.
• ☐ Make the self-serve partner program early. Every manual partner setup is time you don't have.
• ☐ Track conversations and customers separately, so a busy calendar doesn't feel like traction.

Final thoughts

A small budget doesn't mainly limit what you can buy. It limits how many mistakes you can afford. Most of ours weren't about money at all. They were about asking strangers for too much, too early, and about skipping basics because we were in a hurry.

The things that worked had one thing in common: someone else's trust carried us further than our money could. A community vouched for us. A piece of content showed something real. An offer let skeptics check for themselves. If you're launching with a few thousand euros, spend your time earning that kind of trust. The money comes second.

We're still early, and I expect to add to the "didn't work" list. If you're launching something similar and want to compare notes, I'm easy to find.

Sources

1. SaaS Capital, 2026 Spending Benchmarks for Private B2B SaaS Companies (1,000+ companies)
2. WordStream, Google Ads Benchmarks 2026 (13,474 US search campaigns, Apr 2025 – Mar 2026)
3. Gartner survey of 646 B2B buyers, via Digital Commerce 360 (Aug–Sep 2025)
4. G2, Half of B2B software buyers now start their research with AI chatbots (1,076 decision-makers, March 2026)