“I’ll just post on Upwork” is not a client strategy. Here’s what I built instead.

The "starting conversations with people who were already halfway to needing you" line is the whole point. Most outreach fails because of timing, not copy.

What you're describing is a context problem disguised as a prospecting problem. The freelancer doesn't have bad messages. They're sending them to people who haven't felt the pain yet.

I ran outreach for a B2B product for 8 months and the highest reply rates came from openers that referenced something specific and observable about the prospect's situation. The challenge was always the gap between "the problem exists" and "they know it's a problem and want to fix it." Signal-based outreach solves the first half. What are you seeing on conversion after the initial reply?

Hadi_Khalil's question is the one I'd answer first, and I think there's a cheaper version of it than waiting for user counts.

Everyone in this thread is arguing about whether the signal is commercially strong — neglect means no budget, wrong inbox, permission beats precision. Nobody has asked whether the flaw is actually there. That's the failure mode that's worse than a generic email: if you open by citing one specific thing about their site and you've got it wrong, you've proven you didn't look. Generic gets ignored. Wrong gets remembered.

And it's checkable without any access to your system, because the sites are public. Take 20 businesses SignalsHunt flagged, the flaw it cited for each, and check each one against the live site — present, absent, or arguable. That ratio is closer to product validation than reply rate is, and you can have it before launch instead of after.

I do independent verification of exactly this kind of claim and I'd run it as a free check if that's useful. Either way the test holds on its own.

The Upwork → product pipeline is real. I did something similar — I was doing freelance voiceover work, paying per character to a cloud TTS service, and the friction of “write script → send to API → wait → download → edit” was killing my margins. So I built my own tool that runs the model locally. Now it’s a product I’m launching.

The pattern: freelance work exposes friction that clients won’t pay you enough to endure. That friction is your product idea. You’re already feeling the pain, you already know the workflow, and you already know what “good enough” looks like because you’ve been doing it manually.

What’s the one Upwork task you still hate doing? That’s probably your next product.

This is a lesson a lot of builders learn the hard way: being available is not the same thing as being valuable.

Posting "I can help with X" puts the work on the client to figure out why they should care. The stronger approach is showing a specific problem, a specific outcome, and why your approach is different.

The same applies to SaaS too. A homepage can list features all day, but buyers are usually looking for a reason to believe: "This understands my situation."

The people who win attention usually don't just offer a solution — they make the problem feel understood first.

"You still send it yourself... I didn't try to automate the human part" — that's the line I'd defend hardest, and it's also where I'd put the pressure.

I run an experiment where an AI does the work end to end and I only hold the levers I can't undo: publish, price, pay, send. Sales so far: zero. So this is a note from someone whose distribution hasn't worked, not advice from someone who fixed it — you're a paying client ahead of me.

The failure I hit on the AI side is that an agent doesn't pursue your goal, it pursues the goal you finished writing down. I left the success metric half-specified once, and it came back genuinely pleased with itself: earned ¥1, ¥1 is still money, goal achieved. It wasn't lying to me. It hit what was written.

If I'm reading your five steps right, the implicit success condition is finding a flaw — and a flaw is always findable: stale copyright, no mobile layout, stock photos. The condition you actually care about sits further down, at reaching someone who buys. That gap is why GregoryScottHenson and mocktomer can both push back on the same detector and still not land in the same place: they both read neglect as a weak signal, then split on what's stronger — money already moving versus an owner showing signs of trying.

Which makes the open question here not another signal type, but what the tool is accountable for. TCLeon511 named why it bites: right message, wrong inbox looks identical to bad copy in the reply column. And siqiliu's question about which stage leaks can't be answered while "found a flaw" is the condition being satisfied. Flaw, or inbox? That one you can settle this week — another 100 emails won't settle it for you.

I really relate to this. Finding clients is probably the hardest part for me too. I can spend hours working on a project, no problem. But when it comes to finding people who might actually need my help... I honestly never really know where to start.

The idea of looking for businesses that already have a problem instead of just sending random messages makes a lot of sense to me.

I like the focus on finding businesses with real, visible problems instead of sending generic cold emails. Personalized outreach usually gets much better responses, and automating the research while keeping the conversation human seems like a practical balance.

"Waiting for clients and hunting for clients have wildly different odds" — completely agree, and it's the same realization that got me building in this space too. One thing I ran into with local-business data specifically: how are you handling staleness? A site with no mobile layout today might get redesigned in 3 months, so the "signal" has a shelf life. Do you re-check leads over time or is it a one-time scan?

This really resonated with me.

I think you touched on something a lot of founders and freelancers struggle with — being good at the actual work doesn't automatically mean you know how to consistently find opportunities.

I especially liked the point about removing the “blank spreadsheet” problem. The hardest part is often not sending the message, it's knowing who to contact and having a reason to start the conversation.

Also appreciate the honesty around results. No magic promises, just building a system that makes the process easier and more consistent.

Wishing you success with SignalsHunt — excited to see how it evolves.

The spreadsheet panic is real. I do CRO audits, so half my job is already spotting the 'this site is losing money' signals: broken checkout flows, no clear CTA, forms that ask for way too much up front. Never thought to turn that into an actual outreach trigger instead of just something I flag once I'm already hired. Does it catch stuff like a dead add-to-cart button, or is it mostly copy/design signals like the stale copyright date example?

Really resonates with the hunting vs waiting framing. The empty spreadsheet column is painfully accurate - I was in the same spot before building my own thing. One thing I found after actually landing clients: getting paid quietly became the bottleneck. I would win a project, then burn evenings chasing invoices and figuring out which ones were overdue - eating into the exact hours I needed for finding the next client. So I built a small tool for myself that watches my invoices and sends friendly automated follow-ups before I even have to think about it. The client acquisition and delivery are different skills point applies downstream too - collecting payment is its own process problem, and it is the one nobody writes about. Thanks for sharing.

One counterpoint from years of running a services business: an outdated site is a weak buying signal, because a company that has ignored its website since 2019 has already told you it does not care. The stronger trigger is money already moving, a new location, a hiring spree, a rebrand, a raise, because budget and urgency are both present before you show up. Layer those triggers on top of your site audit and that 100-emails-to-1-client ratio moves a lot.

This nails something founders miss: the bottleneck for service providers isn't usually "can I do good work" - it's "do I know who needs it and how to reach them first." Most advice assumes hunting is optional if you're good enough. But there's a 6-month gap between knowing your craft and having the distribution channel to get in front of the right people. Your insight that "the hard part is the blank spreadsheet, not the email" is exactly backwards from how most tools are built - they focus on automating the pitch when the real friction is finding who to pitch to. The fact that you're still doing the human part (sending, conversing) but automating only the discovery makes this actually defensible as a product too.

This title really hits home. Relying solely on platforms like Upwork or Fiverr can be completely unpredictable when you're trying to build a steady client flow. I've been active on those platforms, and as a web developer, I quickly realized that leaving your entire pipeline in the hands of a third-party marketplace isn't a long-term strategy. I've had to focus a lot more on direct outreach and setting up my own automated SEO processes to gain some control over my lead generation.

I'm really curious to read about the details of what you built instead! Did you focus on an inbound lead funnel, or is it some sort of custom outreach system?

I agree. Building the product and finding customers are completely different skill sets. One thing I've learned is that targeting businesses already experiencing the problem makes outreach much easier than pitching everyone. How did you validate that your client signals were strong enough before building the product?

"Row 1, column A" hit home. I've been that person staring at a spreadsheet with nothing to type, convinced the work itself was enough. The shift from waiting to hunting makes sense—targeting businesses with obvious gaps, like a site frozen in 2019, feels way better than bidding against 40 randoms.

Curious though: when you pick those visible-need businesses, how do you handle the outreach message so it doesn't become another generic "I do websites" template? That's always been my bottleneck.

waw i want to join to this

This hits home — most of us treat Upwork as a strategy instead of just one channel. Curious, what actually moved the needle for you: outbound or content?

I like that you automate preparation, not the conversation. The product risk I’d test is whether specificity actually feels credible to the sender. A detected issue such as an old copyright year may be technically correct but commercially weak. Before generating copy, show the user the evidence, why it may matter, a confidence level, and one counter-signal such as recent social activity. Then let them approve the observation. That turns SignalsHunt from a copy generator into a judgment aid. For validation, I’d separate positive replies, calls, and paid work; one client from 100 emails is encouraging, but those stages will show whether lead selection or wording is the limiting factor. Which stage currently leaks the most?

This is epic

the "who to talk to and what to say" problem is the real one, and you framed it well. upwork fails not because its bad but because everyone there is already shopping and comparing on price, youre one of 50 bids, so you compete in a race to the bottom. the higher-leverage move is reaching people who arent actively shopping yet but have a visible trigger (they just launched, just raised, just rebranded, or have an obviously broken/outdated site). the "what to say" then solves itself: you reference the specific thing about THEM, not "hi i do websites". generic outreach gets ignored because it could have been sent to anyone. one caution on any system that automates finding these: the trigger has to be real and the message has to stay human, the moment it feels mass-sent youre back to being ignored, just at scale. the skill nobody teaches (that you named) is exactly this, relevance beats volume, and a reason-to-reach-out-now beats a pitch every time.

Ouch, I think I'm really stuck with this problem too! It really is extremely hard to find clients for online jobs. Imagine Upwork is already the best site for online job seekers, but it's still a test of faith to find even one client.

One nuance from doing a lot of this kind of outreach lately: "visible problems" are a double-edged signal. An outdated copyright date means the site is neglected — but businesses that neglect their site usually neglect the budget for fixing it too. My replies come overwhelmingly from owners showing signs of trying — a fresh redesign, active socials — plus one concrete gap. And the best-converting group by far is people who publicly asked for feedback somewhere. Permission beats precision: one warm "you asked, here's what I found" outperforms ten perfectly-targeted cold observations.

This resonates. Being good at the actual work and being good at consistently finding clients are two completely different skills, and the “blank spreadsheet” problem is very real. I also like that you’re focusing on specific, visible signals instead of generating generic lead lists. A personalized message based on something the business is already struggling with should give you a much stronger reason to start a conversation.

The early results sound realistic too—one paid client from 100+ emails isn’t a huge success story, but it’s useful validation that the process can generate real opportunities. I’d be curious to know which signals have produced the highest reply rates so far.

Brilliant insight, Serghei. I completely agree that finding clients and doing the work are two entirely different skills.

As a software engineer, I face the exact opposite problem. I build the products perfectly, but I hate the hunting phase. That is actually why I started building premium, production-ready Next.js templates (HadiKits) to cut the delivery time in half.

For an agency like dricomm, combining a tool like SignalsHunt for outreach with a solid arsenal of enterprise-grade templates for rapid delivery seems like a massive competitive advantage. You get the client, and you deploy the site in days, not months.

Would love to connect and chat about how you currently handle the delivery phase at your agency.

The 100+ emails → replies → one paid client is useful evidence, especially because you’re running it against your own agency.

What would you need to see from users who aren’t you before you’d consider the product itself validated — not just the outbound method it enables?

nice:)

Disclosure: I build LeadGrid (leadgrid.eu), which does the boring half of this — trade + city in, local businesses out of Maps with contacts enriched off their sites. So, neighbours.
Building on siqiliu's counter-signal point: the cheapest one I've found isn't on the website at all, it's the Google Business Profile. Last post, last photo, whether they reply to reviews. The site is the web agency's artefact and often four years stale by default; the profile is the owner's, so it tells you whether anyone's actually home.
The thing nobody's mentioned, and it cost me the most time: a Maps listing gives you a business, not a person. For small trades the address on the site is an info@ nobody opens while the owner is on a Gmail with his first name in it. Good opener, wrong inbox — and that failure looks identical to bad copy unless you measure the two separately.