How to automate B2B local lead generation in 2026

Almost every agency we talk to wants the same thing from lead generation: a predictable weekly drip of fresh local businesses to email, with as little manual work as possible. A pipeline they can glance at on Monday and trust is filling itself. We run that exact setup for our own outbound at PostOrbit, and it isn't glamorous. There is no magic agent doing it all. There is a small stack of tools, a few conventions about how we structure queries and CSVs, and a list of things we've learned not to touch.

This is the workflow we actually use, with the failure modes we run into. It assumes you don't want to write code and you don't want to babysit a scraper. It also assumes honesty: automating lead gen doesn't mean lights-out prospecting. You still have to look at your list before you send.

The shape of the workflow

At the top level there are five steps, and they haven't changed much since we started running this in 2024:

1. Fan out parallel Google Maps queries by city and vertical.
2. Enrich the results with emails, with MX validation on every row.
3. Clean the CSV before anything touches a sequencing tool.
4. Pipe into Instantly or Smartlead through Google Sheets or Zapier.
5. Watch what breaks for a week, then adjust the queries.

Almost every problem we've had was in step one or step three. The sexy-sounding parts — AI personalisation, sending sequences — matter less than getting the query shape and the CSV hygiene right.

Step 1: parallel queries by city and vertical

One big query does not scale. If you search "marketing agencies in the UK", Google Maps gives you the top cluster in a few big cities and not much else. What you want instead is a matrix: a list of cities crossed with a list of verticals, each one run as its own search. For our own outbound, we keep a Google Sheet with two columns — city and vertical — and just generate the combinations.

A few things we've learned about query shape:

• Keep the vertical narrow. "Italian restaurants in Manchester" returns a usable list. "Restaurants in Manchester" returns a mess with every kebab shop and a lot of duplicates.
• Use the local spelling. "Zahnarzt Berlin" returns different results to "dentist Berlin". If you're doing EU prospecting, this matters more than people expect.
• Split big cities by neighbourhood. London is not one market. "Web design agencies in Shoreditch" gives you a tighter list than "Web design agencies in London" and you can run the sub-queries in parallel.
• Cap each query at a reasonable size. We usually pull the first 100–150 results. After that the relevance drops off a cliff and you're just spending credits.

Inside LocalLeads you can queue the queries and let them run. Inside Outscraper or Apify you can do the same thing with a batch job. The important part is that you're running many small queries, not one enormous one.

Step 2: email enrichment with MX validation

This is where most workflows quietly poison themselves. Google Maps does not give you emails. You have to derive them from the business website — pattern guessing, SMTP checks, pulling from public pages. The risk is that your list ends up full of plausible-looking addresses that either don't exist or go to a catchall that accepts everything.

In our pipeline every email gets an MX record check before it goes into the CSV. If there's no MX record on the domain, the email gets dropped. If the domain has a well-known catchall pattern, we flag it. Our enrichment is built into LocalLeads so this happens on every Pro search, and we still run a second pass with a verifier like NeverBounce or MillionVerifier before we send anything. Two independent checks catches things one would miss.

Match rates for local SMBs land between 40% and 65% depending on the vertical. Professional services (accountants, solicitors, clinics, agencies) score higher. Trades and retail score lower because their web presence is lighter. We don't chase 100% — we'd rather have a clean 45% match than a dirty 80%.

Step 3: CSV hygiene before sequencing

This is the single highest-leverage step and most people skip it. Spend twenty minutes in a spreadsheet before you touch your sequencing tool. Here's our checklist:

• Drop duplicate domains. Google Maps sometimes lists the same business twice under slightly different names.
• Drop chains. A Starbucks franchisee doesn't decide vendor spend. Filter out any brand you recognise, and any entry with a parent-company URL you've seen more than three times.
• Flag role-based emails. info@, contact@,hello@, office@ go to a separate sheet. They still work, but they convert at roughly a third of the rate of named mailboxes, so we sequence them differently.
• Normalise company names. Strip "Ltd", "LLC", "GmbH", and extra whitespace. If you use{{company}} in your email template, you do not want "Smith Plumbing Ltd." in the subject line.
• Pick a first-name column. If you don't have one, don't personalise with "Hi [first name]" — leave the greeting neutral. Nothing tanks reply rates like "Hi [FIRST_NAME]" in production.

Step 4: pipe into Instantly or Smartlead

Our sequencing tool is Instantly. Smartlead works just as well. The important thing is that the pipe from LocalLeads into the sequencer is not a human dragging a CSV around.

We use two paths. For small weekly lists we export to a Google Sheet and Instantly reads it directly — Instantly has a Sheets connector that pulls new rows on a schedule. For larger batches we run the export through a Zapier zap: LocalLeads Sheets row → filter out flagged rows → create Instantly lead. Either way, the CSV hygiene step above is what makes this safe. If your rows are clean, automating the handoff is boring. If they're dirty, automation just means you bounce faster.

One honest note on CRMs: a lot of posts (including older ones on this blog) talk about HubSpot and Pipedrive integrations like they're live. They aren't — not natively from LocalLeads. Today our Google Sheets export is the real workhorse, and Salesforce export is on our roadmap. HubSpot and Pipedrive users almost always pipe through Zapier or their own Sheets connector, and that works fine.

Step 5: what actually breaks

Three things, in roughly this order of frequency.

Proxy bans and silent throttling

Any scraping pipeline gets throttled eventually. The symptom is that results go quiet — fewer rows per query, gaps in the data — rather than a loud failure. We monitor average row count per query across a rolling week, and if it drops more than 20% we know our upstream is unhappy. Good tools rotate proxies for you. If you're running your own scripts, you'll hit this wall fast.

Role-based emails pretending to be personal

Some enrichers return first.last@domain.com based on pattern guessing when the domain actually only acceptsinfo@. These get accepted by a catchall, never bounce, and never reply either. You only notice because your reply rate for that vertical is mysteriously zero. The fix is the second-pass verifier and treating suspiciously-clean lists with suspicion.

Seasonal dead listings

Google Maps does not instantly remove closed businesses. In some verticals — restaurants, small retail, fitness — up to 10% of results in a given query can be businesses that closed in the last twelve months. There's no clean fix, but we drop any result with zero reviews and a last-review date older than eighteen months. It helps a bit.

Why we don't run fully lights-out

The pitch for automation is usually "you never touch the list again". We don't believe that pitch, and we don't run that way ourselves. Every week someone on our team opens the CSV for ten minutes, spot-checks twenty rows, reads the subject lines out loud, and kills anything weird. It catches a surprising amount: a search that grabbed a wrong city, a vertical that turned out to be half chains, an email template that renders badly when the company name has an ampersand in it.

Ten minutes a week of human attention on top of a decent pipeline is the difference between outbound that keeps working and outbound that quietly corrodes your sending reputation for a quarter before you notice.

A note on what we sell

Obvious disclosure: we build LocalLeads, so our pipeline runs on it. It's €19.99/month for 1,200 credits and €49.99/month for 5,000. Each result costs 1 credit; email enrichment with MX validation is included as an option in every paid plan and costs 1 extra credit — but only when it actually returns a verified address. Failed enrichments are free, and credit packs from €3.75 cover overages. Enrichment was the whole reason we built the tool in the first place, because tacking an enricher on after the fact never gave us clean enough data.

If you want to run a similar setup, you can start at leads.postorbit.io and the free tier lets you preview a real search without a signup. If you already have a workflow you like, ignore us and keep using it — the shape of the pipeline is what matters, not the tool.