Saffron Leads

The data already lived there, so the source of truth needed no migration. It stays human readable, so the business survives the app breaking. And it hands over hosting and authentication for free. The tradeoff is accepted openly: reads take three to five seconds, and it caps out in the tens of thousands of rows. At this team's size and lead volume, those limits are irrelevant.

Spreadsheet rows are positions, and positions move as leads arrive and sorting changes. Writing by position means a status eventually lands on the wrong customer, which in a sales tool is a revenue bug. Every save recomputes a composite fingerprint from timestamp, email, phone and name, and writes only to exact matches. Position becomes irrelevant.

One word, two completely different conversations: brands who want boba drinks manufactured, and cafes who want to buy popping boba as an ingredient. Different pricing, different follow up cadence. They became separate categories, and leads from both sites were unified by intent. Buyer intent is the right grouping axis, not lead source.

When bookmarking was requested, the easy path was another status. I pushed back: a status answers "where is this deal," a star answers "do I care right now," and mixing them corrupts both. A starred lead keeps its stage and its date order, and gains a gold border plus its own filter. Orthogonal concerns stay orthogonal.

This field holds quoted prices. A dropped connection during an auto save can destroy a price note silently, and nobody would know until the customer quoted it back. Explicit save is the right call when the field holds money.