What has gone wrong?

There are four build artefacts and they are not interchangeable.

• Apple silicon Mac — the arm64 disk image. Apple menu → About This Mac: if the Chip line says “Apple M-something”, this is yours.
• Intel Mac — the disk image whose name ends -x64. If About This Mac shows a Processor line naming Intel, take this one.
• Windows — the setup .exe is the ordinary route: it asks where to put LeadCo, makes the shortcuts, and needs no administrator rights. A portable .exe also exists for a locked-down machine or a USB stick — it installs nothing and runs from wherever you put it. Both are the same build, and the Windows build ships as a beta.

The wrong Mac build will not open, and macOS does not say why in so many words. If a download refuses to launch at all, check this first. The download page lists all four.

This is expected, and it is not a problem with your download or your licence. The installers are not yet signed with a paid developer certificate, so both operating systems hold them on first launch. Every unsigned application gets the same treatment.

• macOS. Drag LeadCo out of the disk image first — do not run it from the mounted window. Then right-click the app (or hold Control and click) and choose Open. macOS shows a warning with an Open button; press it. You do this once; a normal double-click works from then on. Depending on your macOS version the wording is “Apple could not verify LeadCo is free of malware”, “cannot be opened because the developer cannot be verified” or “is damaged and can’t be opened”. None of the three is true of the file you downloaded.
• Windows. SmartScreen shows “Windows protected your PC”. The control that gets past it is More info, then Run anyway.

If you ever see either warning for software you did not deliberately download, do not do this. It exists for a reason.

If the app opens and closes again immediately, that is a crash rather than a block. Note anything on screen and send us your operating system and its version — the version is the part that identifies it.

Straight answer: the Windows build is produced and inspected, and nobody has yet watched it start on Windows hardware. It is built on a Mac and checked as an archive. Everything in it is the same code that runs on macOS, and the configuration is complete — but “should work” is not “has been seen to work”, and we are not going to print the second when we mean the first.

Practically: try it, and if it does not start on your machine, tell us what it did. A refund request inside the 14-day window is accepted for exactly this — see the Refund Policy. macOS is the platform this build has actually been run on.

In your user profile’s application-data folder — never next to the application:

• macOS — ~/Library/Application Support/LeadCo. In the Finder, choose Go → Go to Folder… and paste the path.
• Windows — %APPDATA%\LeadCo. Paste the path into the File Explorer address bar.

That folder holds your settings, proxy credentials and run history. Nothing in it is uploaded anywhere by the application, and it is deliberately kept apart from the program itself: move LeadCo, reinstall it or replace it with a new version and the folder is untouched.

The separation cuts the other way too. Deleting the application does not delete your credentials or history — to remove LeadCo from a machine completely, delete the application and then the folder above. Anything you exported is an ordinary file wherever you saved it, and stays yours either way.

Yes. LeadCo reads the map through your own proxy account, and the Start button stays switched off until one is connected. No proxy ships with the application and none is included in any licence — you buy the traffic directly from a provider at their published rate, and they bill you for it.

The reason is capacity, not concealment. Sustained volume from a single connection gets rate-limited and the run stalls part-way through; a rotating pool spreads the requests across many addresses so a long run finishes.

Any provider offering rotating residential proxies over HTTP works. DataImpulse is simply the one the in-app instructions are written for. What the application needs from any of them is a host, a port, a username and a password.

Work through it in this order, testing after each change.

• Prefer the connection line over the four fields. Settings → Connect proxy takes a single pasted string in the form host:port:username:password and splits it itself, which removes the two most common mistakes at once.
• If you are filling the fields by hand: host and port belong in separate boxes. A host still carrying :port on the end will not resolve.
• The username and password are the proxy’s own credentials from your provider’s dashboard — not the email and password you sign in to DataImpulse with. Mixing those two up is the most common cause of all.
• A stray space picked up by a copy will fail exactly like a wrong password. So will an exhausted plan: check the account still has traffic on it.
• If your provider restricts access to allow-listed IP addresses, add the address of the computer running the application.
• Office networks, VPNs and security software block outbound proxy ports often enough to be worth ruling out. Try the same credentials on another network.

The test answers one question — can the application reach the internet through your proxy — and answers it in seconds, against the exact area you are about to search. Use it after every change rather than starting a run to find out.

A run that starts fine and fails later is usually the address it is going out through rather than the application. A single residential IP serves a limited number of map requests in a session before the far end starts refusing them. There is no fixed figure — it varies with the address, the area and how hard you run — so treat it as a ceiling that exists rather than a number to plan against.

What helps, in the order worth trying:

• Use a rotating plan, so a long run is spread across the pool instead of loaded onto one connection.
• Split a large area into several smaller runs, and leave a gap between them.
• Reduce the radius. Area grows with the square of the radius, so doubling it is roughly four times the ground to cover.

If instead everything fails from the very first request, it is the connection and not the volume — go back to the proxy test.

Two things arm it, and the application will not begin a run without both.

• A proxy. The chip in the top-right of the window reads Not configured until one is connected. Settings → Connect proxy.
• A place. Type a city into Where, or click the map to drop the pin yourself.

Nothing else is required. The default is All local businesses with the filters at their defaults, so you do not have to pick a niche before your first run.

The radius sets the circle around your pin: everything inside it is in scope, everything outside is not. The slider runs from 0.1 km to 5.0 km in steps of 0.1, starting at 1.0 km, and the box beside it takes the same figure typed. The unit is kilometres — there is no miles option.

Five kilometres is the ceiling because it is the largest circle one run can honestly finish. A wider setting existed once and it made every large run report itself as partial, which customers reasonably read as a broken product. To cover a city, run several pins rather than one wide one.

A larger radius is more area, more businesses to look through and more requests — and requests are proxy traffic you are paying your own provider for. In a dense centre a small radius on the right pin usually returns more useful businesses than a large radius on the wrong one.

You pick exact map categories, not free text. The picker holds 1,102 of them in fifteen groups and opens on a shortlist of the common ones; typing searches the whole catalogue. The default mode is All local businesses, which selects nothing and takes everything the map lists in your circle.

The detail that matters is that one business is filed under several categories at once. A sushi place can be listed as Restaurant, Sushi restaurant and Seafood restaurant simultaneously; a bar with a kitchen may not be filed as a restaurant at all.

So a single category returns only the businesses filed under that one. Selecting several related categories covers the same street far more completely. If a niche looks thin, add its neighbouring categories before you widen the radius.

The drawer has three segmented controls and two switches, and no presets.

• Website — No site at all, Social only, or Any. These three are mutually exclusive: you choose one. There is deliberately no combined “no site of their own” option that merges the first two, so the two counts are never added together.
• Phone — Has a phone, No phone, or Any.
• Rating — Any, 4.0+ or 4.5+. A rating filter is a filter on the businesses that have a rating; a brand-new listing with none is not excluded by it.
• Advanced — two switches, both on by default: Independent businesses only and Hide public and institutional. Turn them off if you want chains, or town halls and schools, in the list.

Business status is not a control. Permanently and temporarily closed listings are never collected, and opening hours are not consulted at all — a restaurant that opens at six is still a result at eleven in the morning. A live count above the drawer’s footer says how many businesses the current filters keep, and a “Why were businesses filtered?” disclosure breaks down what each one removed.

Separate the two possible failures first: reset the filters and run the area again. If businesses appear, the search was fine and your filters removed everything. If nothing appears even at the defaults, it is the search itself.

• Categories. You are selecting exact map categories. If the businesses in that street are filed under a neighbouring one, they are not in scope. Add the related categories, or switch to All local businesses.
• The pin. A small radius dropped one street off centre can miss the whole cluster you were aiming at.
• Filters. No site at all, Has a phone and 4.5+ together can legitimately leave an empty list in a small area. Remove them one at a time.

LeadCo reads a public mapping platform that we do not run, and that platform decides how much automated traffic it will serve. It can slow requests down or stop returning results. This happens, it is outside our control, and no tool can honestly promise it will not — this one does not.

The right response is to ask for less, not to push harder. What tends to help:

• Smaller areas and fewer categories in a single run.
• A proxy plan with enough capacity for the volume you are asking for, so a run is not funnelled through one rate-limited connection.
• Space runs out over a session instead of sweeping a whole city in one pass.
• Stop when it starts. A source that has begun refusing requests is telling you it has had enough for now. Re-running immediately is both less likely to work and the wrong thing to do — leave it, and come back later with a smaller run.

Separately, every run has a ten-minute ceiling built into the application. A run that ends at ten minutes has not failed; it has reached the wall, and everything it collected up to that point is kept and exportable.

What you collect and how you use it is your decision, and any platform’s terms are between you and that platform. Nothing on this site is a statement about what is or is not permitted — see the Acceptable Use Policy.

A run’s length is set by how many requests it has to make and how long each one takes. Four things move it:

• Area. Ground covered grows with the square of the radius.
• Categories. Each category selected is more of the area to sweep.
• Proxy latency. Residential traffic is routed through someone else’s connection and is slower than a direct request. That is the cost of not going out from one address.
• Pacing. Requests are spaced deliberately. Going faster raises the chance of being refused, which costs more time than it saves.

There is also no cache, by design: every press of Start makes live requests, so re-running the same area is exactly as expensive as running it the first time. Business listings change, and a cached answer would quietly stop being true.

No throughput figure is published anywhere on this site, because there is no measured one to publish yet. When there is, it will be stated with the conditions it was measured under.

Two formats, and no others. There is no XLSX, no JSON, no API, no CRM connector and no webhook.

• CSV — 36 columns, one row per business, written UTF-8 with a byte-order mark so a spreadsheet opens it with accents intact. Scope is All businesses by default, with Filtered and Selected as explicit choices, and the field separator can be a comma or a semicolon.
• PDF — a formatted call sheet with a fixed three-column layout: business name, phone, and the link to the listing. Column choice and separator deliberately do not apply to it, because its layout is designed rather than generated.

There is also Copy phones, which puts the phone numbers of the selected rows on your clipboard, one per line.

The 36 CSV columns cover identity, contact, category, rating, opening hours, coordinates, the map link, and a provenance block recording which search produced the row. There is no email address column and no owner name — the map does not carry them, so neither do we.

Exporting is never licence-gated. Everything you have already collected can be written out whatever state your licence is in.

Check the file first: open the folder the application wrote to and confirm it is there and is not zero bytes.

• Every field crammed into one column is a separator setting in the program opening it, not a broken export. European Excel expects a semicolon — choose it under Choose columns…, or re-import and set the delimiter explicitly.
• Accented characters showing as mojibake is a character-encoding setting in the same import dialog. The file itself is UTF-8 with a byte-order mark.
• If nothing was written at all, the destination is the usual cause: a folder the application cannot write to, or a file still open in a spreadsheet and therefore locked. Close it, or export somewhere else, and try again.

Saved runs are kept for seven days, and then the most recent 200 of whatever survives that. Both limits are applied automatically, age first, and an open run is never removed.

It is deliberate rather than a storage saving. Business listings go out of date, and a month-old list of “businesses with no website” quietly stops being true — the moment it does, it costs you calls. Re-run the search instead of trusting an old one.

Anything you exported is a file on your disk and is untouched by this. If you want a run to outlive the week, export it.

Your key is shown on the confirmation screen immediately after payment, with a button to copy it. That screen is the delivery: copy the key and keep it somewhere you will find it again.

If you closed the page before copying it, write to us with your order reference and we will send it to you. The order reference is on the receipt from Creem, the seller of record on your statement.

That is correct for the build you install today, and nothing is missing from your order. Version 5.0.0 ships with activation switched off: the licence panel in Settings is hidden rather than shown inert, because a panel about a system that cannot be used is worse than no panel. Nothing in the application is locked behind it, so you can install and run straight away.

When a version ships with activation, the Licence panel appears in Settings and you paste the key you already have. Version 5.0.0 has no in-app update channel, so that version arrives as a download from this site rather than as an update inside the application. Keys are in the form LCO5-XXXX-XXXX-XXXX-XXXX.

Until then: keep your key, and do not worry that you have not been asked for it.

By downloading it. Version 5.0.0 has no in-app update channel — it will not check for a new build and will not notify you of one. Replace the application with the new download; your settings, proxy credentials and run history live apart from the program — ~/Library/Application Support/LeadCo on macOS, %APPDATA%\LeadCo on Windows — so the swap cannot touch them.

If you would like to be told when a version is published, write to us and we will record the address against your order.

LeadCo purchases appear on a statement as CREEM.IO*, because Creem is the Merchant of Record and the seller on the transaction. Receipts, card changes and cancellation all live in their customer portal — sign in with the email address you paid with; there is no password.

Cancelling stops the next charge and never cuts short the term you have already paid for. A refund returns money already taken, and requests are accepted within 14 days of the original payment — see the Refund Policy. They are different things and you can ask for either.