How is this different from buying your product or hiring another agency?

The product is self-serve software. The studio is done-for-you: we deploy the systems, run weekly optimization, and own pipeline outcomes. Most agencies bill a flat retainer for reports. We take a lower base and earn the rest on sales-accepted meetings or incremental revenue.

How fast until we see something?

Most engagements ship first instrumentation and chat within days, not quarters. Early signals (intent alerts, handoffs) typically appear in the first one to two weeks; revenue metrics compound over the first 6–12 weeks.

Do we need engineering on our side?

No. We implement tracking, chat, enrichment, and integrations. You need a stakeholder who can approve access and a sales or growth owner who will act on Slack handoffs.

What data do you access? Are you GDPR-ready?

We work with first-party site data and the tools you authorize (CRM, Slack, ad platforms). Processing is scoped to conversion systems you approve. Privacy and terms are linked in the footer; we can walk through DPA needs on the call.

What's included in the discovery call?

A 30-minute live funnel review, a rough leak estimate with priority fixes, and a clear yes/no on fit. No hard sell — if we are not the right partner, we say so.

Can we start with one service?

Yes. Launch often starts with conversion intelligence and AI chat. We expand into paid ops or enrichment once the core system is producing signal.

What if it doesn't move the needle in 90 days?

Then we earn far less, by design. The performance half of our fee only bills on sales-accepted meetings or incremental revenue, so a flat quarter costs you the base and nothing more. Baselines are signed before work starts, monthly reviews show what moved, and after the initial 90 days you are month-to-month.

How does the performance fee work — can it spiral?

No. Each plan includes a monthly performance cap agreed in your contract before work starts. Base plus cap is your absolute ceiling for the month, and you know it before you sign.

How do you define a sales-accepted meeting or incremental revenue?

A sales-accepted meeting is booked, attended, and marked accepted by your own rep in your CRM — you hold the record, not us. Incremental revenue is measured against a trailing 90-day baseline we measure and sign with you before work starts. If we cannot agree that a meeting or sale is attributable, we do not bill it.

Is the Revenue Leak Audit fee credited or refundable?

Credited in full against your first month's base if you start a retainer within 30 days. It is not refundable on its own, because the deliverables — leak map, tracking report, named-account sample, and 90-day plan — are yours to keep and act on with or without us.

What happens if we stop — do we keep what you built?

Yes. You keep the data, tracking configuration, chat training, intent models, and playbooks, and any tool accounts we set up in your name stay with you. There is nothing to hold hostage.

BLUEPRINT · CRL-INTENT-01

The Anonymous Visitor Engine

Know which company is reading your pricing page while they are still on it.

The complete build, given away. The real scoring model with its actual weights, all ten steps, and the six mistakes that cost us the most. This is the system we run for clients, written down so you can follow it.

v1.0 · August 2026 · Written for anyone with b2b traffic and a sales team

01What it does
02The architecture
03Instrument and identifyemail unlock
04Weight, reveal, and extractemail unlock
05Score, gate, and learnemail unlock
06The scoring model, with the real weightsemail unlock
07Thresholds, floors, and the gateemail unlock
08Calibration, backtest, and the feedback loopemail unlock
09What it looks like in the wildemail unlock
10Gotchas we hitemail unlock
11What the production version addsemail unlock
12Build it yourselfemail unlock

What it does

Anonymous traffic has one failure mode that costs more than bad copy ever will. A company that could buy from you reads your pricing page three times on a Tuesday, does not fill in a form, and nobody on your team ever finds out.

This is the system that closes that window. It watches every session on your site, works out which company is behind it, scores what they actually did rather than how long they stayed, and hands the ones that matter to a human in Slack while the visit is still warm.

Sessions that look busy but mean nothing get scored low and left alone. Alerts are capped at one per visitor per day, because the fastest way to kill this system is to make sales mute the channel in week two.

Everything below is what we actually run. The weights are the real defaults, the thresholds are the ones in production, and the mistakes are ours.

The architecture

Four stages, and a loop that feeds the first stage from the last. Events arrive from the page and from your server. Identity resolves the visitor across sessions and, where possible, to a company. Scoring turns raw behaviour into one number per session and one per visitor. A gate decides whether a human hears about it.

Instrument

First-party script plus server-side events. Page views, clicks, scroll, dwell, section focus, text selection, chat messages.

↓

Identify

Stable visitor ID, optional fingerprint merge, then company reveal from IP with aggressive exclusions.

↓

Score

Page catalog weights each URL. Features become factors, factors become a capped session score, sessions blend into a visitor score.

↓

Gate and hand off

Above threshold and not alerted today, send to Slack with the top factors and the last ten pages. Thumbs up or down feeds the weights.

The order matters. Most teams build stage one and stage four and skip the middle, which is how you end up with a Slack channel full of people reading your careers page. The value is entirely in stages two and three.

What each stage writes down

Stage	Persists	Used for
Instrument	Raw event rows per session	Feature extraction, replay, debugging
Identify	Visitor record, company match, cache entry	Cross-session continuity, firmographic points
Score	Session score, factor breakdown, visitor score	The gate, dashboards, explaining the number
Gate	Alert event with factor keys	Dedupe, and attaching feedback days later

Read the rest, and take the PDF

10 more sections including the full scoring model with its real weights, the thresholds we alert on, and the six mistakes that cost us the most. Email only.

One email with your resource. No spam — unsubscribe anytime.

No gate on the useful part: everything we publish here is what we run in production, and you are welcome to build it yourself.