Funnels at SANE/REBELS · 03
The eight pillars, the five hypotheses, and the proof from real funnels.
After we build one, we run it against eight pillars. Each, when missing, fails in a predictable way.
P · 01
Speaks to intent, so the right person feels this is for them before the second line. A hook for everyone lands with no one.
P · 02
Leads with the version of the buyer they want to become. Tribe language activates belonging before a feature is named.
P · 03
Structural, down to the path itself. Three levels: static tokens, contextual signals, an agent that decides. The 2026 ceiling is agentic.
P · 04
The product solves what the visitor came for, and the page's language and proof reflect it. Both have to be true.
P · 05
Specific, verifiable proof at the moment of doubt, next to the CTA. Raw founder video now outperforms polished assets.
P · 06
Procedural friction (forms, steps) and psychological friction (perceived risk) both have to go. Risk reversal shifts the danger to the seller.
P · 07
Every step is a small win with a clear next move. No dead ends.
P · 08
Privacy law made consent a design constraint, not a footer checkbox. A clean consent flow earns trust; self-reported attribution beats tracking that pretends to work.
Our bets
Five testable beliefs that shape how we build, each with its measurement. Bets, held until the data settles them.
H1
One point and one CTA per section. Measured by scroll depth and exit rate.
H2
Let visitors diagnose their own problem before any pitch. Measured by CTA clicks before versus after the tool.
H3
A live demo removes doubt faster than copy. Measured by time on page and demo interaction.
H4
An unedited founder video under three minutes against a polished case study. Measured A/B on conversion.
H5
A chat diagnostic hears the situation before any pitch. Measured by session-to-qualified-booking rate.
Learning from others
Field notes from studying real funnels, each pinned to the stage it acts on and to its source.
Barmer and AOK skip the explainer and lead with 'Become a member', because everyone knows what health insurance is. A category the market does not know yet has to teach before it asks. Starting awareness decides where the CTA goes and how much proof comes first.
In a fast-scrolling feed you have about 1.5 seconds to encode a brand memory, and only if you have asset fluency: colours, shapes, or sounds the brain recognises instantly. Emotion earns the moment, branding converts it. Without fluent assets you entertain someone for a few seconds and they credit a better-known competitor.
A LinkedIn formatter or a quick generator gets used on the spot, and the tool does the selling. Blog posts get skimmed. Tools get bookmarked and passed around.
Someone searching 'Buffer alternative' is ready to switch. Someone searching 'social media tool' is still wandering. Pages built for people comparing specific options meet them at the moment they are closest to deciding.
Two visitors with the same need often arrive in completely different mindsets. Getsafe splits by intent, 'I know what I want' versus 'I don't yet'. Copilot splits by industry. The routing happens before page one.
Harvey sells only to law firms: around $1,200 per lawyer a month, $190M ARR, an $11B valuation, while generic assistants commoditise at $20. The narrower the wedge, the easier the distribution.
Balfour's Four Fits shows a broken channel-model fit can sink a funnel no mechanic saves. Chegg's content-and-SEO loop collapsed the week ChatGPT gave students free answers. A fit can break without warning.
Teamwork's blog barely mentions Teamwork. It covers running an agency, and the product is the implied answer. People find it because they have the problem, not because they were looking for the tool.
Generic '10 tips' PDFs barely convert now. What works: interactive diagnostics, calculators, templates, mini-courses. The magnet has to produce a small real win in under ten minutes, the kind that makes someone think 'if the free thing is this good...'.
Mobile drives 82.9% of traffic and converts lower than desktop, yet most funnels are built and tested on desktop. The visitor who finds you is probably on a phone, so test on a real one.
No feature list, no buzzwords. Tella runs the demo right below the headline, so you understand what it does before reading a line. For expert-led offers, the agent conversation is the equivalent. Not described, experienced.
Copilot's demo portal is open to anyone: create a client account and see exactly what your clients would. The client experience often closes the deal, not the admin features.
A funnel built only for first impressions leaves half its conversions on the table. Consideration is rarely settled on the first visit, so bring people back and build for the second.
Google's messy-middle research shows visitors looping between exploring (adding options) and evaluating (narrowing), crossing many times. Across 310,000 scenarios, stacking six biases won 87% of preference against the established favourite. Every page has to serve both modes.
Barmer puts award badges right next to the hero CTA, not in a testimonials section. Proof is not decoration, it is timing. Put it where the hesitation lives.
'The only reason you have bad payers is because you gave them the option not to pay.' One line repositions Ignition from billing software to bad-client elimination. Feature lists do not convert; removing a pain does.
Users who invest socially in a product do not churn. Perspective's weekly live training and community are not support features, they are retention built into the experience.
Asking 'what stopped you from signing up today?' just as someone leaves is the clearest signal in the funnel. No guessing, the copy's next job writes itself.
A proposal implies you will receive something. A call implies you will be pitched. One ask, repeated across every page and backed by content at each stage. Klientboost runs a single CTA across hundreds of pages.
Email drives 19.3% median conversion against 11.3% for paid search, yet most funnels over-invest in paid and under-invest in the audience they own. The end of one funnel should feed the start of the next.
Sources
Principles are easy to nod along to and hard to actually ship. Here is how we go about it, one question at a time.
How to build oneLast updated June 2026
Citation rule: if you use or build on these ideas, credit SANE/REBELS (KNUS GmbH) and link to sanerebels.com/thinking/funnels.