A win-back engine that filled the fall schedule
The same company had years of past customers and a manual scheduling crunch every fall. I built the system that finds who to text, drafts it, reads the replies and turns them into booked jobs.
- Headline figure
- 41 of 60 fall bookings on the books by late September that came from the texts
- Built with
- Next.js, PocketBase, Field-service API, CRM SMS
- Client
- Irrigation company, kept anonymous
Every fall this company has a short, brutal window where hundreds of customers need the same service. Booking it was manual: one office person scheduling about 400 services in 30 days for two techs. Meanwhile hundreds of past customers sat in the system and had never been asked back.
Measured on the project. Nothing rounded up, nothing projected.
What I built
An operations console that mirrors the field-service system into its own database and adds what the office was missing:
- Win-back campaigns that pick who to text from the company's own history and draft the message
- A Follow-Up Desk that puts every reply on one list, sorted by what the customer actually asked
- Zone clustering and route planning, so a tech is booked where they already are
- A Booking Advisor that shows which days a tech is already close by
- A Schedule Builder that creates the job in the field-service calendar with a tech and an arrival window
- Nightly review requests after completed jobs
One rule runs through every customer conversation: the office presses send, never the AI. The system drafts, sorts and recommends. A person decides.
Mockups with sample data. The way the system works is real.
- Sample data
01 / 03Win-back campaigns pick who to text from the company history and draft the message. The office approves every send.
- Sample data
02 / 03The Follow-Up Desk puts every reply on one list, sorted by what the customer asked.
- Sample data
03 / 03The Schedule Builder books the yes where a tech already is, with an arrival window.
What happened
By late September, 41 of the 60 fall bookings on the books came from the texts. By late September, 208 customers had been texted and 127 had replied, a 61% reply rate. The Schedule Builder then created 59 jobs straight into the calendar, each with a tech and a window.
Most service businesses are sitting on work they already earned. The customers are in the system. Nobody called them.
Built with
- Next.js
- PocketBase
- Field-service API
- CRM SMS
- Leaflet
- Census geocoder
- Vercel Cron
The client is kept anonymous. Every number on this page was measured on the project. Photos are stock and screens are mockups with sample data.
