Four tools, one system.

This is a teardown, not a results page.

ShotBot has been trading for weeks, not years. Any conversion figure we published would rest on a sample small enough that a careful buyer could do the arithmetic and see it. So we are not going to publish one.

What we can show is the system: what leaked, why the obvious fix makes it worse, and what the replacement does when nobody is watching. Every figure below is either a cost the owner paid or is read out of the running code.

Four tools, and the seams between them.

ShotBot rents a padel ball machine by the hour at a club in Corsham. One machine, one price, one hour at a time. The business is simple. The software running it was not.

Note the shape of that cost. It rose with the number of bookings taken, because the booking plugin charged by transaction. A busy month was billed as a busy month. Software that charges more the better you do is not a partner, it is a tax on your own success.

And the subscription was the cheap part

Running the four tools took about an hour a day before a single booking came in: checking one inbox against another, confirming what had actually been paid for, telling the club who was coming. Every booking on top of that added roughly twenty minutes of its own.

Read those two numbers together and the same thing is true twice over. The software charged more in a good month, and a good month took longer. Success made both bills bigger, which is the wrong way round for any business and a particularly bad way round for one person running it between sessions.

Every seam was a manual step.

We look at a business as four stages: getting attention, turning it into an enquiry, turning that into an order, and turning that into a relationship. Every one of them had a gap.

Every one of those gaps closed the same way: somebody did it by hand. That works until the day it is forgotten, and the day it is forgotten is the day a customer arrives to find the machine already taken.

A fifth tool would have made it worse.

The standard answer is another plugin. Something to connect the booking tool to the email tool, something else to write a customer record, a fourth subscription to tie the first three together.

That fails for a reason worth stating plainly. Each tool owns its own copy of the truth and none of them agrees to be overruled. When they disagree, and they do, there is no authority to settle it, so the tiebreaker is a person checking. You have not removed the manual step. You have hidden it and added a subscription.

So: one database, and one place where a booking is true. The page, the payment, the confirmation, the diary and the venue's calendar all read from it or write to it. Nothing holds a second opinion, so nothing has to be reconciled by hand.

One loop, start to finish.

A customer picks a date and an hour, gives their name, email and phone, and pays by card. Stripe takes the money, and the Stripe account is in the owner's name from day one, so the relationship with the payment processor belongs to the business rather than to us.

The moment payment confirms, four things happen with nobody touching them: the booking is written with the customer attached to it, a confirmation goes out, the owner is notified, and the hour appears in the venue's calendar.

The confirmation does a second job

It teaches. A short setup video sits in the email rather than waiting on the website, because the moment somebody has just paid is the moment they will actually watch it. A customer who arrives already knowing how the machine works has a better session.

The venue sees an hour, not a person

The club needs to know somebody is collecting the machine. It does not need to know who, or their phone number. So the venue subscribes to a read-only calendar feed carrying the hour and nothing else. The alternative was sharing the admin password: one login, between two organisations, on a page listing customer names and numbers.

A test fails the build if a name, email or phone number ever appears in that feed. The rule is not a policy anybody has to remember. It is enforced.

The parts a demo never shows.

Anyone can show you a booking form. The difference is what happens on the bad days.

Two people cannot buy the same hour

Not a check that runs first and hopes. A partial unique index in the database refuses to hold two live bookings against one slot, so two people pressing pay at the same instant is not a race that usually resolves correctly. It is a rule that cannot be broken. A cancelled booking releases its hour, which sounds obvious and was a real bug, found and fixed.

An abandoned checkout gives the hour back

Someone who starts paying and wanders off holds the slot briefly. That hold expires and is swept up, so the hour returns to sale on its own. Completing payment moves the booking beyond expiry, so a paid session can never be released by mistake.

It says so when it breaks

Every booking system claims it works. This one tells you when it did not. A daily check reads payments and bookings separately and looks for the customer who paid and went missing, which is the failure no single screen can report, because each side looks correct on its own.

A subscription, and an hour a day, to nothing.

No subscriptions. Card fees remain, because those belong to the payment provider rather than to the software, and they would have been payable on any of the tools this replaced.

The hour a day went with it. Nobody copies a booking between two screens, writes a confirmation, or tells the club who is coming, because none of those are jobs any more. The twenty minutes a booking used to cost is now nothing, which means the next booking is the first one that has ever been free to take.

One thing did not change, deliberately. Reviews are still asked for by hand, over WhatsApp, by the person who met the customer on court. At this number of sessions a month a personal message converts better than anything automated would, and automating it would be building something to solve a problem the business does not have yet. Knowing what not to build is part of the job.

The trade changes. The shape does not.

ShotBot rents a machine by the hour. A garage sells MOT slots, a studio sells classes, a hire firm sells days. In each case money moves through a process, one thing can only be sold once, and somebody is holding it together with a spreadsheet and a good memory.

All of it transfers: one place where a booking is true, a constraint that makes double-selling impossible, a confirmation that does a second job, and a system that reports its own failures instead of waiting to be caught.

ShotBot is our own business, and building it is how Archline started: the problem came first, and the agency came out of solving it. We would rather tell you that than let you find out. What is described here is running in public, taking real bookings and real payments today.

Start a project