About
Why an engineer, not an agency
thesauce.dev is run by one engineer who spent a year inside a business that had been buying leads without knowing what one was worth. The tracking had never been fully set up, and what had been set up had quietly broken. The method on this site is what it took to find that out and fix it.
01The operating principle
An agent driving CLI tools, making disposable tooling, being driven by an engineer.
The engineer sets direction and approves. Research and plan by default; build on an explicit go. The engineer merges, the engineer publishes, the engineer sends the email.
The agent drives the tools directly: your ad accounts, your CRM, your analytics, your website and the database behind it. Thousands of small read-only scripts, each written for one question and thrown away.
Tooling is disposable until it isn’t. A throwaway script becomes a library on its second client. That is how one person covers the ground an agency staffs with five.
Division of labour
- Sets directionEngineer
- Drives the toolsAgent
- Reviews every changeEngineer
- Writes disposable toolingAgent
- Merges, publishes, sendsEngineer
02How we report
Six commitments. Each one has cost us something.
- 01
Findings over deliverables
A forty-page deck is a deliverable. “Your ad platform has never been told about a sale” is a finding. We are paid for the second kind.
- 02
Confidence tiers, in writing
Every number we report is tagged proven, suggestive, untestable or refuted, with its window and method. We keep a register of our own claims that turned out wrong, and we withdraw them in the open.
- 03
Read-only first
Nothing is changed until it has been measured. The teardown and the diagnostic change nothing. Every change afterwards is a dry run, a sample, a go, and a way back.
- 04
The agreed definition governs
What counts as an outcome is written down and agreed before we start. We reconcile our count against whatever you already count and explain every difference, rather than arguing about it. Nobody changes a definition mid-term.
- 05
Seams, not culprits
The vendor before the vendor usually broke the thing, or never finished it. We present where the systems failed to meet, not who to blame.
- 06
Systems are ours, data is yours
The techniques and the module library are reused across clients. Your data, your exports and your credentials never leave your engagement.
03The technical half
The technical half
The measurement needs somewhere to live, and in most established businesses it does not have one. The reference engagement started by moving an ageing website, two brands on one legacy CMS, onto a modern stack with its redirects and search rankings intact, replacing the back office with a staff admin portal, and adding two customer-facing portals. The tracking, the outcome ledger and the reporting were built on that. The same seat covers businesses that already run their own portal or internal software and have nobody holding it.
04In short
How an engagement runs.
- Engagement order
- Teardown, then diagnostic, then build, then retain. Never the other way round.
- Teardown vs diagnostic
- The teardown is free and reads your ad accounts, analytics and website. It does not touch your CRM, any data warehouse, or build anything. The diagnostic is a fixed fee and adds the CRM, the funnel economics and the value of a customer by segment.
- Ads management
- Optional. Your agency can stay.
- Pricing
- Free teardown. Fixed-fee diagnostic. Builds and retainer scoped on the call, anchored to market.
- Reporting
- Monthly, reconciled to your books, with confidence tiers and agreed definitions.
- Who it is for
- Any business, B2B or B2C, where the money is decided in a conversation rather than at a checkout: legal, home services, real estate, B2B services sold on a call, elective medical and the like. Also businesses running their own admin portal or internal software, and agencies who want the work delivered under their brand.
Start
Two ways to start. Both are free.
The teardown is a written read of your measurement stack and your funnel, from your own numbers, in about two weeks. The call is thirty minutes to decide whether a teardown is worth doing.