Reservation and booking automation

The first job is to write a booking down in one place, in a shape the rest of the system can act on: who, which service, which slot, which staff member or location, and how to reach them. The confirmation, the reminder, the reschedule link and the staff day view all read from that one record. If the record is incomplete, or lives in two places, no amount of messaging automation fixes it.

That is also why the phone does not have to go away. A booking taken at the counter or on a call gets entered into the same record as an online one, which is what makes the calendar worth trusting. A system that only knows about its own online bookings is the one staff quietly route around, and then you are paying for software and still keeping a notebook.

Most of the build time goes into rules that are obvious to your staff and invisible to a developer until somebody writes them down. On the Domino Ra'anana ordering site, which took about two months and has been in daily use for more than five years, the rules that mattered were not visual at all.

Sending one message is easy. Sending the right message to the right person at the right minute, every day, without a queue somebody has to babysit, is the part worth designing. Mincha Time does this in production: a Firebase Cloud Function runs once a minute and looks for a Firestore document at the bucket for the current time, keyed hour_minute, so 13:47 is the document 13_47. If that document is not there, the run does nothing and costs almost nothing.

Two more choices keep it cheap. Users are grouped by rounded latitude and longitude, so the Hebcal zmanim API is called once per location per day instead of once per user. And each run writes tomorrow's bucket for that group after it finishes today's send, so the schedule advances itself one day at a time rather than precomputing a calendar that might need invalidating.

Booking reminders need the same discipline plus the opt-outs. Mincha Time checks two independent flags before every send, a permanent disable and a same-day snooze. A booking reminder needs the equivalent: somebody who has already confirmed, already cancelled, or asked to be left alone should not get the day-before nudge.

This matters more than it sounds and it is cheap to get right at the start. Mincha Time sends notification copy in six languages — Hebrew, English, Russian, Spanish, French and Yiddish — chosen from a per-user language field and stored as plain lookup objects rather than through an i18n library. The mechanism is not clever: know the language when the person signs up, then send every outbound message through the same lookup.

For a booking flow aimed at customers in Israel that usually means Hebrew and English, which also means right-to-left layout in the customer-facing screens and in anything you send them. Retrofitting that later touches every template and every screen at once, so it belongs in the first version even if only one language is switched on at launch.

The fastest way to never launch is to try to cover every service, location and exception at once. A first version usually covers one location, one group of services, one reminder pattern and one payment path. Deposits, waitlists, recurring appointments, multi-staff calendars and no-show reporting are all easier to add once you can see which of them your bookings actually need.

Leaving something out is not the same as ignoring it. Domino's checkout calculates pricing and delivery rules in one place, which is why adding a zone or a new deal is a small change instead of a rewrite. The parts still worth improving are written down in that case study, including moving more checkout validation into shared functions that can be tested without rendering the whole checkout page.

The honest anchor is the Domino build: about two months for a public menu with search and category filtering, deals that carry their own selectable options, a cart, delivery and pickup checkout, card payment through Cardcom, a separate cash path, customer and admin emails with full order details, the POS handoff, and the admin screens behind all of it.

A booking flow that stops before payments and before an external system handoff is a smaller job than that. Adding either one moves it back toward that shape. For contrast, Mincha Time's first version, including the minute-resolution reminder engine described above, was about a month, because it had one calculation, one reminder and no payment step. Scope is what moves the date.