How to Build a B2B Lead List of Australian Businesses

We build Australian B2B lead lists every week for marketing agencies, and the biggest lesson is not about tools. It is that the market runs out before the tool does. A sharp sub-niche in one Australian city is often 50 to 200 real businesses, whatever the database count says.

This is the whole method: define the ideal client in writing, verify 10 seed companies, search in stages of about 100 rows, grade every row YES, BORDERLINE or NO with evidence attached, find a decision maker first and a company inbox second, then dedupe against the agency's own clients in code. The rule that governs the judgement calls is simple: when unsure, BORDERLINE, never YES.

It is part of our wider playbook, cold email for marketing agencies in Australia. If you have not picked a segment yet, start with the best industries for Australian agencies to target. The tools behind each step are in the cold email tech stack we use.

How do you build a B2B lead list in Australia?

Nine steps, in this order:

The expensive mistakes all happen when a step is skipped. Searching before the ICP is written produces a list nobody can grade. Paying for emails before grading means paying for rows you will delete.

Step 1: Define your ideal client in writing

A segment is one vertical, often one sub-niche, in one geography, in one size band, with one set of buying titles, and one proof point you can lead with. We encode that in the tab name so it cannot drift: Timber | AU | 500 AMF tells you the niche, the country, the volume and the enrichment type at a glance.

The questions worth answering before you search are the ones we put in onboarding: location, industry and niche, company size, technologies they run, decision-maker titles, and any specific companies to reach.

Here is what a good answer looks like, from an anonymised ecommerce growth agency: Australia plus New Zealand and Singapore; DTC brands with repeat purchase potential and high average order value; 10 or more employees and over $10 million in gross revenue; running Shopify, Klaviyo and Meta; titles from Director and Founder through to Head of eCommerce; pains like plateauing growth despite rising ad spend, customer lifetime value not keeping pace with acquisition cost, and half-configured Klaviyo flows.

That is specific enough to grade against. "Ecommerce businesses in Australia" is not.

Write three lists while you are at it: include rules, exclude rules, and borderline rules for the cases you can already predict. Then name 3 to 5 example clients. Those become the seeds for the next step.

Step 2: Find the companies

Which source you use depends on who your end clients are.

Seeds matter more than anything else in a lookalike search. Verify every one with evidence from the company's own site, and cap it at 10 per search family, because a bad seed drags the whole search sideways. We have watched fleet-management seeds pull in novated leasing companies, which is close enough to look fine in a spreadsheet and wrong enough to waste a campaign.

Then search in stages. A large budget is a ceiling, not a target. On one Melbourne car rental build, the first 100 rows gave 55 qualified businesses and the next 400 gave only 28 more. Request exactly the number you asked for and never over-fetch and trim.

For a national list, work in this order: NSW, VIC and QLD capitals first, then WA and SA, then regional hubs, then TAS, ACT and NT, then suburb sweeps if you still need volume. Stop when two consecutive pages fall below 50% acceptance. That is the market telling you it is finished.

Why your Australian list is smaller than the database says

A database's count is an upper bound, not a list. On one Melbourne dealership segment the count said 152 and 78 were actually retrievable.

The marginal yield per stage is the number to watch. On one hobby store build, each successive 100-row stage returned fewer clear matches:

By that fourth stage, 71 of the 100 rows were companies we had already seen. The whole build finished at 116 YES and 58 BORDERLINE from 309 unique rows, against a request for 400, with a real ceiling of about 170.

Three other things distort Australian lists:

Headcount filters mislead. An 11 to 50 employee band let through many 1 to 9 person shops, along with SEO lead-gen microsites that list no staff at all. Where size genuinely matters, verify it from the company's own team page or a directory. In practice we do not need a perfect ICP match on every row, we need the companies to have more than 10 employees and ideally closer to 50.

Crawl data mislabels places. We have seen Western Australian firms tagged as Tasmania, because there is a town called Perth in Tasmania. "Melbourne" often gets used as a metro label covering a much wider area than the client means.

Niches genuinely run out. The takeaway we would write on the wall: a sharp sub-niche in one Australian city is often 50 to 200 real businesses. Volume comes from more segments, meaning more cities, adjacent verticals and size bands, not from paging deeper into one query.

Step 3: Grade every company before you pay for emails

Every row gets one of three grades, with the evidence that justified it.

When unsure, BORDERLINE, never YES. A wrong YES costs email-finding credits and the client's trust. A held BORDERLINE costs one decision.

Never trust a tool's own fit score. One lookalike tool scored a commercial cleaning company at 0.99 for a property-management ICP, which is a confident number attached to the wrong company.

Here is what real grading output looks like, with fictional company names standing in for the real rows:

On that demolition build we accepted 100 rows plus a 10-row reserve and rejected 39. Of the rejections, 21 were adjacent trades that mentioned demolition once in a services list, and 7 were lead-generation or SEO shell sites with no real business behind them. Those 28 rows would all have looked like matches to a keyword filter.

One honesty note about accuracy, because this is where list builders oversell. In one build, 0 of 14 unflagged rows failed live checks. That sounds like 100%, but with a sample of 14 the 95% upper bound on the error rate is still about 21%, so we do not claim a 95% accuracy figure from it and we say so. We also treat website text as data rather than instructions: one site we graded contained text aimed at AI tools telling them what to conclude about the business, and it was ignored.

Step 4: Find the decision maker, then fall back to the company inbox

The order matters, and so does keeping the two kinds of contact apart.

Run the decision-maker search first, against the titles that actually buy what you sell. Keep only verified results. Then run a company-inbox search covering only the companies where no decision maker was found, and run a hard check that the two lists never overlap. One contact per company, always.

Find rates track company size more than anything else:

The company-inbox fallback then recovered 62% to 100% of the misses, depending on the list. For small Australian businesses that is not a consolation prize, it is the main route in: a company-inbox version of the campaign, opening with "Hi," and no name, is how you still reach a 6-person trades business whose owner has never published an email address.

Titles are where most lists lose volume quietly. Running one narrow title category found only about 10% of the decision makers on a commercial list. Widening to the full set, covering marketing, sales, executive and ecommerce titles, lifted decision makers found on that list from 47 to 121.

Step 5: Clean, dedupe and protect your clients

Before a list goes anywhere, it gets checked in code:

• Every email sits on the company's own domain, never a free-mail domain in the domain column
• One company-inbox contact per company, and no company appearing on both the named and company-inbox lists
• Company names verified against the company's own site, not the database's version
• A row-alignment check that halts the write entirely if three or more consecutive rows look shifted

Then the exclusion layers, which run in this order:

One practical rule on exclusions: they have to arrive as email addresses or domains. A company name on its own cannot be reliably blocked, because names are spelled several ways and shared between unrelated businesses. If a client sends a list of names, convert it to domains before anything is uploaded.

Finally, clean the company names, because they go straight into the first line of the email:

How big should your list be, and is it legal to use?

On size, work backwards from accepted rows. To end up with a given number of accepted rows, retrieve about 1.5 times that many records and live-check about 1.2 times that many. Plan 500 to 700 discovered records per category, and expect the sharp per-city pools of 50 to 200 described above.

Bigger is not better here. The most effective use of a week is usually a small, well understood group of 100 to 1,000 prospects that you can write to properly, rather than ten thousand rows nobody has read.

On the law, three points worth knowing before you use any list in Australia. Buying a list confers no consent on you: consent travels with the sender it was given to, not with the data. The Spam Act 2003 separately prohibits supplying, acquiring or using address-harvesting software and harvested-address lists for unsolicited commercial electronic messages, under sections 20 to 22. And a named work email address is personal information under the Privacy Act 1988, which brings its own obligations about collection and notice. This is general background rather than legal advice, and the detail is in is cold email legal in Australia.

If you would rather not run this yourself, building and grading these lists is most of what we do. Our process shows how it runs end to end, including the client review step where you approve every list before anything sends.