I looked at instantly, lemlist, and apollo. then i just built my own outreach system for $0.
spent a week evaluating cold email tools. heres what i found and why i ended up building instead of buying.
the tools i tested:
- instantly ($30-97/mo): great deliverability, but i still had to find prospects, research their websites, and write all the emails myself
- lemlist ($59-99/mo): nice personalization features like custom images, but the "personalization" is still {first_name} {company_name} level
- apollo ($49-119/mo): massive prospect database, but the emails are generic because apollo doesnt know anything about the prospect's actual business
the gap i kept hitting:
every tool assumes you already know what to say to each prospect. they handle the SENDING. but the hard part was never sending — it was knowing what to say that would actually get opened.
so i built something different. a python script that scans each prospect's website before writing the email. it finds real problems — "your site has 47 images without alt text" or "your title tag is 89 characters, google cuts it off at 60."
now every email references something the recipient can verify in 10 seconds. thats not {first_name} personalization. thats "i actually looked at your business" personalization.
the stack (embarrassingly simple):
- python requests + beautifulsoup for website scanning
- smtplib for sending (gmail SMTP with app password)
- json files for prospect lists and sent logs
- cron for daily scheduling
- total infrastructure cost: $0 (runs on existing server)
results after 850+ emails:
- 0.5% reply rate (industry average for personalized cold email is 0.3-1%)
- every single reply was from someone who verified our findings and wanted to talk
- zero "take me off your list" responses
- zero spam complaints
the trade-off is speed. instantly can blast 1000 generic emails/day. my system sends 25 highly personalized ones. but 25 emails that reference real data beat 1000 that start with "i hope this email finds you well."
im now offering this as a done-for-you service at $297/mo. agencies get 500 personalized emails/month without touching a tool, writing a template, or managing a campaign. we handle everything from finding prospects to monitoring replies.
for indie hackers debating which cold email tool to buy — consider whether the tool or the message is your actual bottleneck.
anyone else build their own outreach system instead of buying one?
The build is impressive — and your core insight is right: the tools solve sending, not "what to say."
One flag from someone who lives in the deliverability weeds: the $0 is real on tools but the risk moved to your domain. Self-built senders skip the guardrails (warmup schedules, volume caps, bounce circuit-breakers) that the boring platforms bake in. One enthusiastic afternoon at 200 sends can put a domain in the penalty box for months.
If you keep the DIY route: hard-cap at ~30–50/day/inbox, verify every address before send (scraped/found emails bounce hard), and watch the 2% bounce threshold like a smoke alarm.
The hybrid many end up at: DIY the intelligence layer (your prospect analysis is genuinely differentiated), rent the boring sending infrastructure. Best of both.
Building your own vs. buying is the right call when the tool you need is really a workflow, not a feature. Instantly/Apollo give you sequences; they don't give you context - who this person is relative to your pipeline, what you've already tried, what happened after the last touch.
The problem with tool-first outreach: you end up optimizing sends, not conversations. Solo founders need the opposite - fewer, warmer touches with full context on each one.
What actually works at small scale: a CRM that's native to how you think. Every prospect has a signal note (where you found them, why they're relevant), a stage, a last-touch date, and a next action. Not a sequence - a relationship log.
I've been building this into a Solopreneur OS in Notion: CRM module where your outreach pipeline cross-references your weekly review and your decisions log. When you're wondering why a batch of outreach failed, the context is right there.
/bin/bash is correct. The constraint isn't the tool - it's the system. What does your current CRM/context layer look like alongside the outreach sequences?
"This is a masterclass in 'message-market fit.' You're 100% right—the bottleneck is never the sending, it's the 'why should I care' factor. Scanning alt-text and title tags as a hook is high-intent personalization that actually scales.
I'm currently running a manual, low-volume outreach campaign for a Tokyo-based competition (Tokyo Lore) using the same 'hand-carved' philosophy you described. I’d love to get your eyes on our '1 in 100' model—I have a specific question about how you'd script the hook for this. I'll shoot you a note!"
The "knowing what to say" vs "handling the sending" distinction is the right frame, and it maps to a problem that shows up identically in paid acquisition.
Most ad platforms are the paid equivalent of Instantly — they handle delivery (impressions) and report on it. What they can't tell you is whether you understood your audience before spending. The parallel to your "$97/mo tool, still have to research prospects myself" observation: a Meta Ads campaign can run at 4x ROAS and still be leaving money on the table because the creative is generic and the segmentation is too broad.
Your "verified problem" approach — scanning the site and finding something the recipient can confirm in 10 seconds — is what Custom Audiences and behavioral targeting approximate in paid. Instead of blasting everyone who has a website, you're finding people whose actual situation matches what you can solve.
The 25 personalized vs 1000 generic tradeoff is worth quantifying: if your reply rate is 10x better and send volume is 40x lower, you're ahead on absolute replies and dramatically ahead on conversion quality. Your "zero spam complaints, zero unsubscribes" metric is the paid equivalent of a low CPL with high post-signup LTV.
The hard part you identified — knowing what to say — is the same hard part in paid. Message-market fit can't be outsourced to the platform. The tool just amplifies what you already understand.
This is the right shift. Most tools optimize sending, not thinking. Your edge is in turning real data into context, not tokens into personalization. 25 relevant emails will always beat 1000 forgettable ones.
This is the definition of Systems Sovereignty. Moving away from overpriced subscriptions to a custom-built infrastructure is exactly how you beat the "Manual Trap" in 2026. Most people just buy more tools; architects build better logic. How are you handling the hook-automation within this setup?
This is solid! really shows where the personalization gap actually is. Most tools treat sending like the hard part, but as you found, it's knowing what to say that matters. It's nice how you went straight to scraping real data from sites, that's the kind of personalization that actually gets replies. How do you handle edge cases where the site structure is weird or there's a missing key info? have you thought about scaling without losing that 1:1 authenticity?
The issue of automating outreach while still providing personalized messages is a tough one. It's interesting that you found the existing tools lacking in this regard, and that you decided to build your own solution instead. The point about knowing what to say to each prospect being the hard part, rather than just sending the emails, is a crucial one. This highlights the importance of understanding the prospect's business and needs. By building a custom script, you were able to tailor your approach to your specific needs. Did you find that the time spent building your own solution has paid off in terms of the quality of leads you're getting, and are you planning to continue refining your approach?
honestly yes — the time spent building paid off more than expected. the quality difference is night and day. when someone gets an email saying "your site has 47 images google cant index" they either verify it immediately or ignore it. the ones who verify become real conversations because they already trust the data before we even talk. compared to when i was using generic templates, the replies now are completely different in tone — people respond like they're talking to someone who did homework on their business, not someone blasting a list. still refining constantly though. biggest thing ive learned is that the scanning accuracy matters more than the email copy. if the finding is wrong or trivial, the whole thing falls apart. so most of my time goes into making the scanner smarter, not the emails prettier.