Consent before measurement
Google Analytics loads only after an affirmative choice. Rejection and withdrawal clear the relevant state.
013 Labs — Internal case study
Building a consent-aware operating funnel.
How 013 Labs connected privacy-aware measurement, lead intake, content distribution, gated video production, and digital-product commerce—and made the boundaries visible enough to test.
Evidence boundary: this is our own dogfooding build, not a client engagement. It documents verified controls as of July 24, 2026. It does not claim revenue, conversion lift, cost savings, time savings, or ROI.
19live Gumroad listings
62tests passed across 10 working tools
32content items dedupe-audited
1/dayhard video-render allowance
The operating problem
A simple public funnel hides several systems.
The useful unit of automation was not “the funnel” as one giant system. It was a sequence of bounded handoffs, each with a visible decision and evidence trail.
Validation before storage
Form data is checked for origin, content type, size, field quality, and bot signals before Airtable receives it.
Context before follow-up
Landing page, referrer, and campaign parameters travel with the lead so follow-up starts from evidence.
Originality before publishing
Sources, normalized copy, claims, angles, and media assets are checked against scheduled and published history.
Allowance before rendering
Same-day manifests and credit limits can stop a paid video run before another asset is generated.
Verification before sale
Release archives, licenses, prices, artwork, listing routes, and tool tests are reconciled before catalog publication.
Controls in practice
What the system is verified to do.
Measurement + leads
Consent-aware conversion measurement
The site gates analytics and conversion events behind consent, without sending form PII to Google Analytics.
- Localized accept, reject, persistence, and withdrawal controls
- Optional newsletter interest remains separate from checklist delivery
- Lead-magnet download remains available if intake fails
- A controlled production submission created one Airtable row and one founder notification
Content operations
A queue that preserves failure history
A duplicate publish exposed a retry weakness. The deleted post remains in history, and the control model changed.
- A format change no longer resets uniqueness
- Content IDs, source URLs, copy, claims, angles, and media are checked
- Thirty-two scheduled and published items passed the token gate; a manual semantic audit then found and replaced seven future theme-level repeats
- An unapproved video cannot block the other four daily X slots
Video production
A hard daily spend gate
The video workflow audits every same-date Israel manifest before audio or rendering and stops if paid use already exists.
- One DMoney audio and one HeyGen render maximum per day
- Sixteen-credit ceiling and 100-credit remaining-balance floor
- No automatic re-render, asset deletion, avatar substitution, or public posting
- The July 23 render remains in review with no publication record
- The July 24 render used 15 credits, left a recorded balance of 924, blocked duplicate generation, and published only after human approval with AI disclosure
Digital products
A catalog tied to release evidence
The live catalog contains the original guides plus eight working tools and two operating suites. Two more verified tools and their three-product evidence suite are packaged behind the provider's daily product-creation limit.
- Twenty-five Personal, Consultant, and suite archives verified
- Sixty-two automated tests across ten working products
- Checksums, file hygiene, required files, licenses, prices, and artwork validated
- All 19 Gumroad product pages and the production store routes reconciled
Evidence, not theater
What this proves—and what it does not.
| Verified control | Outcome not established |
|---|---|
| Analytics: the consent layer can prevent loading and event dispatch until a visitor accepts. | No claim that the implementation increased trust or conversion. |
| Lead path: a controlled production request reached Airtable and triggered one notification. | No claim of real inquiry volume, close rate, or attributable revenue. |
| Content: approval state, public URLs, sources, and dedupe checks are operational and auditable. | Views and clicks are not proof of a repeatable acquisition channel. |
| Video: spend, balance, originality, QA, and human-review gates produce observable stop conditions. | No claim that the published July 24 video will generate leads or outperform text. |
| Commerce: 19 listings are live; 25 archives and 62 tests across 10 working tools passed the documented local checks. | No claim of purchases, product revenue, marketplace discovery, or future platform-review outcomes. |
The practical lesson
Put evidence at each edge.
Consent before measurement. Validation before storage. Source context before follow-up. Originality before paid generation. Approval before publication. Verification before sale. When a check fails, the system can stop at the affected boundary without pretending the rest of the process succeeded.
Apply the same approach
Start with one costly workflow.
If work is moving between forms, spreadsheets, inboxes, content tools, and manual approvals, we can map the boundaries, build a controlled pilot, and define the evidence required before broader automation.