DEV Community: Web3 Advisory

Spam Word Lists Are Obsolete: What Actually Sends Cold Email to the Spam Folder

Web3 Advisory — Fri, 31 Jul 2026 22:50:43 +0000

Somebody is right now rewriting "free" as "complimentary" in a cold email that is going to a scraped
list. That edit will not help. The list is the problem, and no amount of vocabulary laundering
touches it.

The "spam words to avoid" posts that keep circulating are roughly a decade out of date. They persist
because they are easy to publish, not because they are accurate. Modern filters are statistical and
contextual, and they are mostly not reading your adjectives.

Here is what actually decides whether your mail lands, in rough order of impact. We run our own sending
stack, so this ordering comes from watching it, including from the parts that cost us deliverability
before we understood them.

1. What recipients do with your mail

This is by far the largest factor, and it dwarfs everything else on this page combined.

Filters are trained on behaviour. Deleting without opening, marking as spam, never replying. Those
signals are the model. Which leads to the conclusion most cold email advice avoids, because it
implies sending less:

Sending to fewer, better-targeted people improves deliverability more than any wording change ever
will. That is the whole game. Everything below is a rounding error next to it.

2. Authentication

SPF, DKIM and DMARC. Binary and unforgiving: they either pass or they do not, and if they do not, the
rest of your effort is irrelevant. This is a one-afternoon job you do once, properly, and then stop
thinking about.

3. Sending reputation

Your domain's and your IP's history. Built slowly, damaged quickly, and not transferable. This is why
"just buy a fresh domain" is a worse plan than it sounds: you are trading a damaged asset for one
with no history at all, and a brand-new domain sending cold volume is its own red flag.

4. List quality

Bounce rates, spam-trap hits, addresses that have been dead for years. A bought or scraped list is
made of exactly these, which is the real reason those lists do not work. Not the price, the damage.

5. The message itself

Real, and much smaller than the four above. Worth fixing, worth fixing last.

Image-only emails. One big image with no meaningful text is a classic spam pattern. A cold email should be mostly text anyway.
Too many links. One is ideal for a first contact, two is fine, five looks like a newsletter nobody subscribed to.
Link shorteners. bit.ly and friends are heavily abused and some are blocklisted outright. Link to your own domain.
Mismatched link text. Anchor text naming one domain while the href points somewhere else is a phishing pattern, and it gets treated as one.
Attachments on first contact. Especially PDFs and anything zipped. Never.
A broken or missing unsubscribe. Beyond being illegal, it converts people who would have quietly unsubscribed into people who click "report spam", and complaints are the metric that actually hurts you.
ALL CAPS subjects, excessive punctuation, emoji pile-ups. Genuinely weighted, though less than folklore claims. They also just read badly.

The phrases that do still carry weight

Not because of the words, but because they correlate with actual scams:

Guaranteed-money language: "make $5,000 a week", "risk free", "100% guaranteed"
Manufactured urgency: "act now", "limited time", "expires today", especially in a subject line
Pharmaceutical, adult, gambling and crypto-pump vocabulary
"Dear Friend", "Dear Sir/Madam" and other mass-mail openers
A fake Re: or Fwd: on a message that is neither

That last one deserves emphasis, because cheap cold-email courses actively teach it. Faking a prior
conversation is a lie, it is a CAN-SPAM violation where it misleads about the message, and it converts
into complaints at a high rate. It is the single worst piece of common advice in this field.

Measure placement instead of guessing

Word lists are guesswork. These are measurements.

Seed addresses. Keep your own accounts at Gmail, Outlook and Yahoo and include them in every send.
Which folder they land in is the only direct read on placement you get.

Google Postmaster Tools. Free, and it shows domain reputation, spam-complaint rate and
authentication pass rates for Gmail specifically. Real numbers rather than proxies.

A spam-checker tool, treated with suspicion. They score against generic heuristics: useful for
catching a broken unsubscribe or a missing header, useless as a verdict on whether you will be
delivered.

Per-provider open rates. If Gmail holds steady while Outlook collapses, that is a filtering
problem at one provider, and the aggregate number would have hidden it.

The one check that replaces the whole list

Before a send, one question: would this recipient plausibly be glad to receive this?

If the answer is no, nothing above saves it. Fix the targeting, not the wording.

And the order to spend effort in: send to people who want it, get authentication right once, keep the
list clean and suppress aggressively, and then tidy the structural things.

The full kit

This is the spam-trigger part of the Cold Email Deliverability Kit: the 25-point pre-send
checklist with the exact SPF, DKIM, DMARC and rDNS records to paste; the 21-day warm-up schedule in
day bands with specific volumes, the per-mailbox ceiling and the warning signs that mean stop; bounce
triage tables with the SMTP codes, including the quiet failures that mean you are already in trouble
with nothing showing up as a bounce; the recovery plan for when you are already in the spam folder,
in order, with realistic timelines and the point at which a domain is worth abandoning; seven
templates that read like a person wrote them; and the legal floor across CAN-SPAM, GDPR and CASL in
operator language.

Written by someone who built the sending infrastructure rather than someone who read about it.

Get the Cold Email Deliverability Kit ($29). PDF plus the
templates as plain text.

Your Documentation Is Your Funnel (And the One Page Nobody Writes)

Web3 Advisory — Thu, 30 Jul 2026 20:46:27 +0000

Most companies file documentation under engineering and marketing under marketing, then wonder why
the funnel leaks.

For a developer product, the docs are the funnel. A developer evaluating you spends about thirty
seconds on your homepage and then goes straight to the docs. Everything after that is decided there,
by a person you will never speak to, who will not fill in a form to tell you why they left.

We have run content and social for developer-infrastructure companies whose customers were engineers
building on top of them. The pattern was the same every time: the landing page got the credit, the
docs did the work.

Four things, and never mix them

Documentation fails most often by conflation. There are four distinct kinds and they have
incompatible jobs.

Quickstart. One page, one goal: something working in under fifteen minutes. No architecture, no
philosophy, no menu of options. Copy-paste blocks that run as written.

Guides. Task-shaped. "How to do X." Written for somebody with a job to finish today.

Reference. Exhaustive and boring. Every parameter, every return value, every error. This is
consulted, not read, and it should be optimised for someone scanning it in a hurry.

Explanation. The why. Design decisions, tradeoffs, the mental model. This is the one that turns a
user into an advocate, and the one nearly everyone skips.

A quickstart that pauses to explain your architecture has failed at being a quickstart. A reference
page with a warm narrative voice is unusable at midnight when someone is debugging. Keep them
separate and each one gets to be good at its own job.

The details that actually decide it

Every code sample must run exactly as written. Not close enough. If it depends on a variable
defined three sections earlier, it will fail, and you will not hear about it. Test your samples in CI
like code, because they are code.

Show the errors, not just the happy path. What does it look like when it breaks, and what does
that message mean? This is the most-searched and least-written kind of documentation in existence.

Make it searchable and deep-linkable. People arrive from a search engine mid-problem, not through
your nav. If a section cannot be linked to, it cannot be recommended.

Say when it was last updated. Stale docs are worse than no docs, and a visible date lets a reader
judge for themselves instead of guessing.

Never gate it. No login, no email capture, no "request access." A gate on documentation reads as a
sales process attached to a product that cannot be evaluated.

The page nobody writes

Here is the highest-leverage documentation most companies never produce: one page per common error
message, titled with the literal text of the error.

Somebody hits ECONNREFUSED at two in the morning, pastes it into a search engine, and lands on your
page explaining what causes it and how to fix it. You have just been maximally useful at the moment
of maximum frustration, to a person who was about to give up on you.

That page will outperform a quarter of campaigns, and it costs an afternoon. Your support inbox
already tells you which ten errors to write first.

Where the rest of the effort should go

Once the docs work, the channel map for this audience is narrower than most marketing plans assume.

Worth real investment: technical posts with substance (the hard problem you actually solved, with
the code and the dead ends, not a listicle), open source that is genuinely useful on its own
rather than a thin wrapper advertising the paid thing, and being present where the problem is
discussed: Stack Overflow, the subreddit, Discord, GitHub issues. Answer the question, help the
person, and mention your product only when it is honestly the answer. That last one is slow,
unglamorous, and the highest-trust channel available to you.

Mostly theatre: paid social to developers, sponsored enthusiasm from influencers, booth-only
conference presence, gated whitepapers, and swag treated as a strategy rather than a thank-you.

Order matters more than breadth

The sequence is not optional, because doing it out of order wastes the earlier work.

Fix time-to-first-working-thing. Nothing else pays off until this is good.
Write the quickstart and the reference. That is the minimum for someone to evaluate you alone.
Add error pages and guides for the top ten things people get stuck on.
Start showing up in communities, helpfully, without pitching.
Publish technical posts about real problems.
Then consider talks, video, comparison pages, and the rest.

Most companies start at step six and cannot work out why it is not landing. They are marketing a
product nobody can get running.

The whole playbook

This is the documentation and channels part of The Developer Marketing Playbook: why marketing to
engineers is structurally different and the four normal-marketing assumptions that sink you, the full
channel map with what each channel is actually good for, the launch sequence for a developer-facing
product, the specific writing and behaviour habits this audience punishes and rewards, the web3
traps if that is your world, and which popular metrics are actively misleading for developer
products.

Written from running social and content for developer-infrastructure projects including Ankr, zkSync
and SKALE.

Get The Developer Marketing Playbook ($39). 12-page PDF,
dense, no padding.

There are no growth hacks in it. They do not work on this audience, and trying them costs you the
trust you needed.