Preflight.

Most funnels fail somewhere nobody walked before launch. Not the offer, not the targeting, the plumbing: a booking link that opens to the wrong calendar, a follow-up SMS that fires before the email does, a decision branch nobody built the no-show path for. Nobody catches it because nobody experiences the funnel as a prospect would before spending is live. Reviewing a funnel today means opening five different tools, an ad manager, a page builder, a CRM, an email platform, a texting app, and mentally stitching them into a sequence. Nobody actually does that. So the first real walkthrough happens after the money's out the door, when a lead comments that the booking page 404s or the SMS never sent.

The failure mode isn't bad strategy. It's an unwalked path. A funnel is a sequence of screens and decisions, and the only way to know if the sequence holds together is to move through it in order, seeing exactly what the prospect sees at each step, including the branches: what happens if they book, what happens if they ghost, what happens if they show up and no-show. Nobody builds a tool for that because it sits between two disciplines: it needs the visual fidelity of a design tool and the logic of a flowchart. Preflight is built to close that gap.

Preflight does three jobs and the product is organized around exactly those three, no more: simulate the funnel, pitch it to a client or stakeholder, and audit it against real benchmarks. Simulate means build the sequence as nodes and walk it in play mode like a prospect would, including every branch. Pitch means the same canvas, in front of a client, replaces a slide deck: click through the actual ad, actual landing page, actual booking flow, with commenting turned on so they can react node by node. Audit means run the finished sequence, or a single piece of it, against a benchmark knowledge base and get a scored report with specific fixes, not a vague grade. Three jobs, one canvas, no context-switching between tools to do any of them.

The core data model is a node graph: `funnels.graph` is a JSON blob of nodes and edges, each node carrying its own `config`, optional `metrics`, and optional `variants`. Edges carry timing labels ('Day 1', 'after 3 days no-show') so the sequence reads as a real campaign cadence, not just a flowchart. Decision nodes are the one node type that fans out more than a single edge, one branch per outcome (booked / no-show / ghosted), and every other node type enforces a strict one-outgoing-edge rule. That asymmetry is deliberate: it's what makes branching legible on the canvas instead of turning into an unreadable web, and it's the rule the drag-to-create interaction and the AI graph generator both have to respect.

React Flow (`@xyflow/react`) is the canvas engine, but the differentiator isn't the graph library, it's the mockups. Every node type renders as a full-fidelity, platform-authentic mockup: five ad variants across Facebook feed, Instagram feed and story, LinkedIn, and Google; a browser-framed landing page (live URL via iframe or an uploaded screenshot); a Calendly-style booking screen; a Gmail-style email; an iMessage-style SMS thread; a phone-call or DM action screen themed per platform. This matters because Preflight has to work as a pitch tool in front of a non-technical stakeholder. A generic flowchart with labeled boxes doesn't land in a client meeting. A mockup that looks exactly like the Facebook ad they'll actually see does. Faithful platform chrome is what turns 'here's our plan' into 'here's what your customer sees,' and that's the sentence that closes the pitch.

The product was originally named Funnel Simulator, which described the mechanism but not the reason anyone would use it. On 2026-07-03 it was rebranded to Preflight, with the tagline 'Run your funnel before you run your ads.' The rename reframes the product around an aviation metaphor that's already fluent to anyone who's flown: you don't take off without a checklist, you don't spend on ads without walking the sequence first. The visual system followed the metaphor directly, a checkmark-to-ascent logo mark meant to read as 'cleared for launch,' a coral-and-navy palette chosen for confidence over cuteness, and a literal preflight-checklist device used across the marketing site. Positioning language shifted from 'build and preview a funnel' to 'clear it before you spend,' which is a small wording change that changes who the product is for: not just funnel builders, but the person who has to sign off on ad spend.

Every non-trivial feature after the rebrand was scoped against those three jobs. Page builder blocks and an A/B variant system serve simulate (make the mockup match reality). Comment threads on public share links serve pitch (a client reacts on the node, not in a separate feedback doc). The analyzer, benchmark citations, and PageSpeed-backed URL audits serve audit (a scored report instead of a gut check). Anything that didn't clearly serve one of the three didn't ship.

Preflight is live at preflight.neilb.app (also reachable at preflight.busqueneil.com and funnels.busqueneil.com, all serving the same deployment) with real multi-tenant accounts, transactional email, and a public marketing site separate from the authenticated app. It's in active use two ways: as a client-facing pitch tool, walking a prospect through their exact funnel with real platform mockups instead of a slide deck, and as Neil's own planning canvas for scoping funnel work before it's built, catching sequencing problems on a node graph instead of in a live campaign.

The audit engine runs against a curated benchmark knowledge base (Unbounce, WordStream, Klaviyo, plus copy-quality heuristics) and, for landing pages, against live Google PageSpeed Insights data, so a report cites a real number instead of a generic score. That's the piece that turns the tool from a mockup builder into something with teeth: it doesn't just show what the funnel looks like, it tells you specifically where it's weak against real conversion benchmarks and lets you apply the fix on the spot.

1. A tool that has to double as a pitch instrument needs pixel-faithful mockups, not schematic placeholders, fidelity is the difference between a plan and a demo.
2. One asymmetric rule (decision nodes branch, everything else is linear) kept a graph-based UI legible at scale instead of turning into an unreadable web.
3. The bug that broke mobile page width lived in a single missing `min-w-0`, viewport-specific QA at real breakpoints catches what desktop testing never will.
4. Retrofitting multi-tenancy is a database-default problem more than an application-code problem: get the schema and RLS right and the app layer barely changes.
5. A rename isn't just an asset swap, "Preflight" changed who the pitch was for, from funnel builders to the person who signs off on spend.
6. Debounced autosave needs a hard max-wait as well as a quiet-period trigger, continuous interaction (like dragging) can prevent the quiet period from ever occurring.

Preflight is a marketing tool built by someone who runs marketing funnels for a living, which is why the product logic mirrors how a real campaign actually breaks: sequencing, branching, and the gap between what a deck promises and what a prospect sees. It shows I can take a product from a raw mechanism to a positioned brand, ship the unglamorous infrastructure underneath it (multi-tenant auth, RLS, benchmark-grounded scoring), and keep shipping without breaking what's already live.