Lead generation is four problems that people insist on treating as one.

You have to work out who is worth writing to. You have to find out enough about them to have something to say. You have to establish that the address you are about to use actually exists. And then you have to write something a busy stranger will read.

Each of those is a genuinely different discipline, each has its own failure mode, and — this is the part that costs people quarters — a failure at any stage silently wastes all the work done at the others. A brilliant email to the wrong person is nothing. A perfect list sent to unverified addresses is worse than nothing, because it damages the domain you will need for the next campaign.

This is the map. Each stage links to a longer piece on that stage, because each of them deserves more than a section. What this article is for is the order, and the reasoning behind the order, because almost every outbound programme that fails does so by getting the sequence wrong rather than by getting a step wrong.

The pipeline, and why the order is not negotiable

Four stages: discover, enrich, verify, engage.

People routinely run them in a different order, or skip one, and the results are predictable enough to be worth naming:

  • Engage before you verify and your bounce rate damages the sending domain, which suppresses delivery to the good addresses you did have. The campaign fails for reasons that look like copy and are not.
  • Engage before you enrich and you have nothing specific to say, so you write a template, so nobody replies.
  • Enrich before you discover and you spend money enriching people who were never going to buy from you.
  • Verify before you discover and you have verified a list you should never have built.

The order is not an aesthetic preference. Each stage consumes the output of the one before it, and each one is cheaper than the one after it. Discovery is nearly free; enrichment costs a little; verification costs a little more; and engagement costs the most valuable thing you own, which is your sending reputation. Doing the cheap filtering first is simply how you avoid spending the expensive resource on rubbish.

1. Discover the right people

Discovery is not "get a lot of names". It is deciding who is worth the next three stages, and it is the only stage where a mistake cannot be corrected later.

Start from the company, not the person. Which companies plausibly have the problem you solve? That question has answers you can actually filter on — headcount, sector, technology stack, funding, whether they are hiring for the function you sell into. A company that just posted three SDR roles has decided outbound matters and has not yet built the machinery, and that is a fact you can act on.

Only once you have the company do you look for the person. The routes:

  • Their own website. The team page, the press page, the job listings, and — the most reliable and least-used source on any site — the privacy policy, which must by law carry a real, monitored address. We cover the whole method in how to extract emails from a website.
  • LinkedIn, for working out who. It is the best B2B filtering tool in existence and it will not give you an address. The honest way to use it — and the reason automated scraping is a bad idea rather than merely a naughty one — is in finding an email address from LinkedIn.
  • Newly registered domains. A company that registered its domain three weeks ago has no incumbent vendor, no contract, and no procurement process. It is the most reachable state a buyer is ever in, and almost nobody works it. See newly registered domains as a buying signal.

The tools that do the mechanical part: the Govarova Extension for prospecting while you browse, the Contact Extractor for pulling contact details from sites at volume, and Daily Domains for the fresh-registration feed.

The discovery failure: optimising for list size

Every team that fails at outbound fails here first, and the failure has a signature: someone is proud of the size of the list.

A large list feels like progress because it is visible, countable, and produced quickly. It is also, at any realistic sending capacity, unusable. If you can defensibly send forty cold emails a day from a mailbox, and a sequence is five steps, then a thousand contacts is a hundred and twenty-five working days of capacity — six months. A list of ten thousand is not ten times as ambitious as a list of a thousand. It is a fantasy with a spreadsheet attached.

The constraint is not how many people you can find. It is how many you can write to properly, and that number is small. Which means the correct move at the discovery stage is almost always to cut, not to add — and the discipline of cutting is what makes the specificity in stage four possible at all.

2. Enrich until every record is complete

An address with no context is a row in a spreadsheet. You cannot prioritise it, route it, or write to it — you can only send it a template, and templates do not work.

Enrichment is the stage that turns an identifier into a reason to write. Address or domain in; job title, seniority, company size, sector, technology, funding out. It answers the two questions that determine everything downstream: is this person the buyer, and does this company have my problem.

It matters in both directions, and most teams are much better at one than the other:

  • Outbound enrichment: you chose them, and you need to know enough to say something true and particular.
  • Inbound enrichment: they found you. A signup with no context is a lead nobody can route. Is this a solo founder or a five-hundred-person enterprise? The answer changes who calls them and what is said, and it is available from the email domain alone.

The neglected direction is usually inbound, and it is usually where the cheapest wins are — these are people who already raised their hand.

Govarova Enrichment runs this at volume; the free enrichment tool, the employee count finder and the tech stack detector each answer one qualifying question on their own.

The enrichment failure: data that is true and useless

The trap here is subtle, because the data is correct.

You can enrich a record with twenty fields and still have nothing to write about. Knowing that a company has 240 employees, uses HubSpot, and raised a Series B in 2024 is true, and it does not give you a first sentence. It gives you a filter.

The distinction worth holding: enrichment data is for deciding, not for writing. Use it to choose who is worth the effort. Then, for the people who survive that filter, go and look at them — read the profile, read the job listing, look at the signup form — because the thing you can actually open an email with is almost never a field in a database. It is an observation, and observations require a human to make them.

Teams that miss this end up with beautifully enriched records and emails that open with "I saw you use HubSpot" — which tells the reader you ran a query, not that you paid attention.

3. Verify before you send

This is the stage that people skip, and it is the one that punishes them hardest.

A dead address does not merely fail to arrive. Bounce rate is one of the cleanest signals a mailbox provider has for "is this sender writing to people who asked to hear from them", because legitimate senders have low bounce rates and scraped lists do not. So a high bounce rate does not cost you the dead addresses — you never had those. It costs you the live ones, by degrading inbox placement for the entire list.

If 20% of your list is dead, you do not reach 80% of your audience. You reach rather less, and you cannot see how much less, because a message filtered to spam is reported to you as "delivered".

The full argument, including spam traps, catch-all domains, and how to recover a domain you have already damaged, is in email verification and your sender reputation. The short version:

  1. Check the syntax. Free, instant, removes the typos.
  2. Check the domain has an MX record. A domain with no mail server cannot receive mail, and this single check removes a large share of a stale list.
  3. Check the mailbox with the mail server.
  4. Classify what is left — catch-all, role, disposable — and treat each differently.
  5. Segment by confidence and send the confirmed addresses first.

Our free email verifier does this for one address; the verification API does it at volume and at signup, which is the cheaper place to solve it.

The verification failure: trusting a tool that never says "I don't know"

A catch-all domain accepts mail for every address, real or not. Ask whether anything@bigcompany.com exists and the server says yes. Which means the standard verification method cannot work there — and catch-alls are disproportionately common at exactly the large companies you most want to reach.

An honest tool reports this as unknown — catch-all. Many tools report it as valid, because it makes the accuracy figure look better. When a vendor advertises 98% accuracy, this is almost always where the number comes from.

The consequence for you is that a derived address at a catch-all domain has been checked by nobody, and you are about to send to it believing it has been verified. Our catch-all detector says so plainly, and the only defensible thing to do with such an address is to treat it as the unknown it is: a small, separate, closely-watched send, never mixed into the main campaign.

A verifier that never says "I don't know" is not more capable. It is less truthful, and the bill arrives on your domain rather than theirs.

4. Engage with sequences that feel personal

Now, and only now, you write.

The first email carries the whole burden. It needs one specific, true observation about their business — something that could not have been sent to anyone else — and one clear, small ask. Three sentences is plenty. The subject line and the first line of the body are a single thought, read together in a list on a phone, and the deciding question in the reader's mind is not "is this interesting" but is this for me, and will it cost me anything to find out. That is the argument of cold email subject lines that actually get opened.

Then the follow-ups, which are where most of the replies actually come from — and which almost everyone ruins by sending the same email again with "just bumping this" on top. The rule that fixes it: every step must carry information the last one did not. If you have nothing new to say, the sequence is over, and stopping is a legitimate option that nobody takes. The full structure — five steps, widening intervals, and the reroute email that nobody sends — is in the follow-up sequence that books meetings.

Govarova Sequences runs the cadence, and the subject line generator and follow-up generator take the mechanical load off the parts that are mechanical.

The engagement failure: winning the open and losing the reply

The most common self-inflicted wound in outbound is optimising for the wrong metric.

Open rate is easy to measure, satisfying to watch, and systematically wrong — Apple Mail Privacy Protection pre-fetches images and reports opens that never happened, and corporate gateways do the same. Worse, the things that reliably raise it are the things that destroy your reply rate: fake "Re:" prefixes, curiosity gaps, manufactured urgency. Each one buys an open by misleading the reader, and the reader finds out within one second of opening.

The number that actually tells you something is replies per open. A subject line that raises opens and lowers that ratio has made your campaign worse while making your dashboard look better, which is the most dangerous combination available.

And watch the complaint rate obsessively. Google's sender guidelines put the line at 0.3% — three complaints per thousand emails. A sequence that irritates one person in three hundred will eventually degrade delivery for everyone else on your list, including the people who would have said yes.

Where the leverage actually is

If you are deciding where to spend the next month, the honest ranking is not the one most teams pick.

  1. Cut the list. Free, immediate, and it makes every subsequent stage better. Half the size, twice the research.
  2. Verify what remains. Cheap, and it is the difference between a campaign that arrives and one that does not. If you have never cleaned a list, the first clean is usually the single largest improvement available to you.
  3. Fix authentication. SPF, DKIM and DMARC. Without them, nothing you send is being judged on its merits.
  4. Research the first email properly. One genuinely specific opener buys you four ordinary follow-ups. Four extraordinary follow-ups cannot rescue a generic opener.
  5. Only then worry about copy, cadence, and testing.

Almost every team starts at five and works backwards, which is why almost every team rewrites subject lines that were never the problem.

What "AI-powered" should actually mean here

The phrase is doing a lot of work in a lot of marketing, including ours, so it is worth being precise about what it can and cannot do in this pipeline.

AI is genuinely good at the mechanical parts: deriving an email pattern from examples, classifying a company from its website, clustering a list into segments, drafting a first pass at copy you will then rewrite, and summarising a hundred customer comments into the four things people actually said.

It is not good — and this is the load-bearing limitation — at deciding who is worth writing to. That decision depends on knowing what your product is genuinely good for and who it genuinely helps, which is a fact about your business that no model has access to. Teams that hand this decision to a tool get a large, plausible, useless list.

Nor can it manufacture the specific observation that makes a cold email land. A model can produce a fluent paragraph about a company. It cannot notice that their signup form accepts disposable addresses, because noticing that requires going and typing one in.

The right division of labour: the machine does the filtering, the derivation, the verification and the first draft. You do the deciding and the noticing. Any tool that promises to take the second half off your hands is promising to make your outbound generic, and generic outbound has a response rate approaching zero.

The other half of the funnel: the leads who find you

Everything so far is outbound. It is the harder half, and it is also the half that gets all the attention, which is a mistake — because the people who found you already have a demonstrated interest, and they are usually served worse than the strangers.

Three things worth as much as any outbound work:

  • Rank for what people are actually searching. And when you do rank, earn the click — most sites are ignoring the half of the equation that is not position. Our piece on meta descriptions that get clicked covers the mechanics, including the fact that Google rewrites most of them and what to do about it. AI SEO Writer handles the drafting and the schema together.
  • Capture the ones who are hesitating. A visitor reading your pricing page for the third time is more qualified than anyone on your outbound list. A chatbot that asks a question at that moment, and social proof at the point of decision, both operate on people who have already selected themselves.
  • Ask your happy customers to say so. Reviews are the highest-leverage marketing asset most businesses never deliberately build — see how to get more Google reviews, including the three rules most businesses are quietly breaking. Review Management runs the asking.

And measure sentiment in a way that survives contact with statistics — which mostly means not believing your NPS score, for the reasons in what NPS actually tells you.

The stage nobody plans: what happens after the reply

Every article about lead generation stops at the reply, as though the reply were the finish line. It is not. It is the moment the whole exercise finally becomes a conversation with a human being, and it is where a surprising share of hard-won pipeline is destroyed.

The three replies you will actually receive, and what each is really telling you:

  • “Not the right person.” The most useful negative there is. Ask, once, who is. Most people will tell you, because it costs them nothing and it gets you out of their inbox. This single question converts a dead end into a warm internal referral more often than any clever sequence.
  • “Not now.” This is not a no; it is a timing statement. The correct response is to ask when, and then actually wait until then. The number of teams that receive “not this quarter” and follow up in three weeks is remarkable, and it reliably converts a warm future buyer into a permanent no.
  • “Tell me more.” The one everyone wants and the one most frequently squandered — usually by immediately sending a calendar link and a deck. Answer the question they asked. Nothing else. They have not agreed to a sales process; they have agreed to one more email.

The instinct on receiving any positive signal is to accelerate: push for the meeting, send the collateral, move them to the next stage. More warm replies die here than at any other point in the funnel. The reply is the beginning of something, not the completion of a funnel step, and treating it as the latter is how a genuinely interested prospect discovers that the personal email was a machine after all.

There is an operational half to this, too, and it is unglamorous: a reply must stop the sequence, immediately. Sending step four to someone who replied to step two is the single most damaging routine error in outbound, it happens constantly, and it undoes every bit of care taken at the four stages above. Whatever runs your sending has to guarantee that as a property of the system rather than as something everyone remembers to check.

The uncomfortable summary

Everything in this guide is downstream of a decision no tool can make for you: who is this actually good for?

Get that right and the rest is plumbing — findable, enrichable, verifiable, writable, and largely automatable. Get it wrong and every subsequent stage amplifies the error. You will find more of the wrong people, know more about the wrong people, verify the addresses of the wrong people, and write to them more persistently, and the complaint rate will eventually tell you what you should have known at the start.

That is why the order matters, and it is why the first stage is the one worth being slow about. The machinery exists to make the mechanical parts cheap. It does not exist to save you from having to think about who your product genuinely helps — and the teams that expect it to are the ones with large lists, clean data, verified addresses, immaculate authentication, and no replies.

Start narrow. Verify everything. Write like a person who looked. The rest is tooling, and the tooling is the easy part.

Every stage in this playbook is one platform in Govarova. See how the pieces price together on the pricing page, or start free.