How I Investigated Like Holmes, Discovered Like Darwin, and Built Like Edison
Jason Barnard, February 2026 (updated 1 March 2026)
In January 2019, I picked up a microphone at SearchY in Disneyland Paris and started asking questions. Over the next fourteen months, I travelled to 22 conferences on four continents, interviewed over a hundred experts from every subdomain of digital marketing, and discovered that they were all describing different views of the same system. None of them could see it. I could, because I was the only person talking to all of them.
Three things happened during that investigation: I gathered evidence that nobody else thought to gather, recognised a universal pattern that nobody else had named, and built a platform that turns the pattern into measurable, repeatable, scalable results. Those three activities map onto three famous figures, and the mapping isn’t decorative. It explains something real about how this work happened and why it produced what it did.
The investigation followed the Sherlock Holmes method: forensic questioning across every available source
Holmes doesn’t theorise ahead of the data. He observes, interviews, cross-references, and lets the pattern emerge from the evidence rather than imposing a hypothesis onto it. That’s exactly what the podcast tour was.
I paid to travel the world out of my own pocket. I went to conferences, sometimes speaking but more often just attending (many conferences gave me free entry as “press,” which was kind of them). And at those conferences, I was open-minded, curious, and (I feel) courageous. I’d hear someone having an interesting conversation and walk up: “That’s interesting. Can I interview you about that?” Or, better still, I’d already be in the conversation, and when I realised it was something I needed to remember, I’d say “Stop talking!” with a big smile (because it shocked them), and then “Can we record this? It’s brilliant.” And then record.
That’s how I met Nagu Rangan, the first Microsoft engineer to talk to me, at SMX London in 2019. I asked him the obvious question: what are the ranking factors? He politely sidestepped it, then told me something far more useful. Obsessing over individual ranking factors is a futile exercise, he said, since the factors and their weightings are constantly changing. What matters is the system’s three considerations: Relevancy, Quality, and Context. I managed to bully him (gently) into agreeing that everything we do as SEOs should serve Understanding, Deliverability, and Credibility. He was extraordinarily open, and that conversation taught me a huge amount about how to interview engineers from big companies: ask the obvious question, listen to the redirect, and follow where it leads.
At BrightonSEO in April 2019, I sat down with Gary Illyes from Google, and he explained how the system actually selects content. Annotation scores multiply. They don’t average. If your content scores 0.9 across three dimensions but 0.1 on a fourth, the composite is 0.0729. A competitor scoring a consistent 0.7 across all four produces 0.2401. The competitor wins. One weak dimension kills excellent content. Brent Payne, who was also there, distilled it into one sentence: “Better to be a straight C student than 3 As and an F.”
I was the only person in the room who recognised the significance of what had just been explained. Everyone headed to the pub. I went with them but sat in a corner writing everything down while my friends told me to stop being a bore. The barman had a pen but no paper. I took a pile of beer mats, split each one down the middle, and wrote everything I could remember on six beer mats split in two. Twelve half-beer-mats. I didn’t keep them, not yet realising quite how important this insight would become. Those twelve half-beer-mats are the lost origin artefact of what I now call the Multiplicative Destruction Effect (the principle that one weak annotation dimension kills otherwise excellent content, and that consistency across all dimensions matters more than excellence in any single one).
SearchY (Paris), BrightonSEO (Brighton), SMX (London, Munich, Paris), UnGagged (London), PubCon (Las Vegas), YoastCon (Nijmegen), SMS Sydney and a dozen more. Four continents. Aleyda Solis on Progressive Web Apps. Véronique Duong on what Baidu does better than Google. Cindy Krum on fraggle indexing. Dixon Jones on citation flow. Andrea Volpini on knowledge graphs (he had been building that infrastructure since 2013, years before I arrived). Fabrice Canel at Bing explaining that what gets crawled is not what gets indexed, and what gets indexed is not what gets ranked. Bill Slawski introducing me to Danny Goodwin at PubCon who later broke the Darwinism in Search story on Search Engine Journal. Barry Schwartz, who I met in person for the first time at SMX London (he had already invited me to speak).
The investigation earned a reputation inside Google itself. John Mueller, Google’s Senior Search Analyst, started sending problems my way on Twitter: “This seems like a rabbit-hole for @jasonmbarnard. I feel bad sending so many curious things so I won’t add him directly.” Holmes gets consulted because the evidence trail proves he knows the territory.
Every expert served one small aspect of one (sometimes two) gates of a pipeline I couldn’t yet see clearly. Holmes doesn’t know what the crime is until the evidence tells him. Neither did I.
The synthesis followed the Darwin method: recognising one universal system where everyone else saw separate disciplines
Darwin travelled the world collecting specimens. Finches in the Galapagos, beetles in Patagonia, barnacles in Plymouth. Each expert he spoke to (and each organism he catalogued) was describing its own local truth. Darwin’s contribution was not the data. It was the recognition that the data pointed toward a single underlying mechanism: natural selection.
My finches were the subdomain experts. Technical SEOs described crawling, rendering, and site architecture. Content strategists saw relevance, depth, and audience matching. Link builders tracked authority signals and citation patterns. Structured data specialists worked on entity resolution and knowledge graph integration. NLP researchers mapped semantic relationships and topic modelling. And the engineers at Google, Bing, and the AI platforms described the machinery that connects them all, in fragments, from their particular vantage points.
None of these people were wrong. They were describing different views of the same pipeline, which I now model as ten confidence gates: Discovered, Selected, Crawled, Rendered, Indexed, Annotated, Recruited, Grounded, Displayed, Won. Content enters at Gate 1 and either survives to Gate 10 or loses confidence somewhere along the way. The confidence at each gate feeds the next. The relationship is multiplicative, not additive. That’s Cascading Confidence: the principle that a brand’s end-to-end algorithmic confidence is the product of its transfer coefficients at every gate of the pipeline.
Scale Gary’s four-dimension example to ten gates, and the arithmetic gets brutal: 0.9 confidence at every gate compounds down to 0.35 end-to-end. Score 0.9 at nine gates but 0.1 at a single weak one, and the pipeline collapses to 0.04. Most brands aren’t scoring 0.9 at every gate. They’re scoring well at two or three gates they actively manage and haemorrhaging confidence at the seven they’ve never heard of.
Darwin called his mechanism natural selection. I called mine Darwinism in Search, because it’s the same principle applied to a different ecosystem. The fittest content survives, but fitness isn’t about strength on one dimension: it’s about maintaining sufficient confidence across all dimensions, at every gate, through the full pipeline. The Multiplicative Destruction Effect (the Gary Illyes beer-mat insight) is natural selection operating at the annotation layer. Cascading Confidence is natural selection operating across the pipeline as a whole. Same principle, different scales.
The Brand SERP was the diagnostic window that made this visible. From 2012 onwards I had been looking at what Google shows when someone searches a brand name, and sensing it was the key to everything. I didn’t have enough knowledge of technical SEO, content marketing, UX, social media, links, or NLP to build a framework. But the Brand SERP was already a view into the entire system, and I was using it as such. Knowledge Panels caught my eye because they were initially just another rich element on the Brand SERP, but they led me into entities and knowledge graphs (I was one of the first, with Andrea having been there for years before me). Then featured snippets (the Answer) caught my attention because they were the first sign a brand was about to get a Knowledge Panel. Each discovery extended the map outward and upward.
Darwin unified Kepler, Galileo, and Hooke. I unified the hundred experts I interviewed across four continents. The principle is the same: different people describing different aspects of one system, visible only to someone willing to listen to all of them. The differentiator was not intelligence. It was open-minded curiosity and the willingness to travel the world, pay for it out of my own pocket, and ask strangers if I could record their conversations.
The construction followed the Edison method: turning a universal theory into infrastructure that works at scale
Darwin observed, Holmes investigated, and Thomas Edison built things that worked and then filed patents on them.
Kalicube Pro is the platform I built to make the theory operational. It processes 25 billion data points across 73 million brand profiles, tracking how brands are represented across the Algorithmic Trinity (Knowledge Graphs, Large Language Models, and Search Engines). It automates The Kalicube Process™ (the methodology that trains AI assistants to understand, trust, and recommend brands accurately). It takes the Darwin observation and the Holmes evidence and turns them into measurable, repeatable, scalable results. Sixteen patents filed with INPI (the French patent office), covering the UCD Analysis System, the Constitutional Sandwich for prompt assembly, AI Citation Tracking, Cascading Confidence optimisation, and more. Because Newton never patented gravity, and look where that got him.
And this isn’t the first time I’ve built infrastructure at scale. In 1999, Macromedia invited me to UCON as one of the leading Flash developers in the world. I had been using Flash for six months. The other invitees included Dreamworks and the BBC. I built uptoten.com from scratch on Flash, essentially engineering a browser before browsers could do what I needed (the web in 1999 couldn’t deliver the interactive children’s entertainment I had designed, so I built the delivery system myself). By 2005 I was leveraging CDN to serve a billion pageviews from a single server, at a time when most companies serving that volume required server farms. In 2008 I won a Gold Davey Award for the EdTech platform. The instinct has always been the same: see the system, then build the infrastructure that makes it work at scale.
Edison held 1,093 patents. His most famous line is “genius is one percent inspiration and ninety-nine percent perspiration.” For me, the ratio is inverted. Ideas arrive constantly. They always have. The real work, the thing that took years and nearly broke me, was synthesis: taking a hundred expert perspectives from every subdomain of digital marketing and recognising that they all describe different views of the same system, then building the platform that proves it.
Up until the last year, I was on my own. Trying to bring all the ideas together was a huge struggle. I’d generally manage it on a plane, bus, or train (I love those long-haul trips because I can’t work properly, bad or no internet). So I’d sit with my eyes closed and think. That’s when all the ideas came. And when they came, I’d write them down desperately fast so I wouldn’t forget, then sit with my eyes closed and think again, open my eyes, write some more. Rinse and repeat. On one trip I went around the world twice (once the “wrong way round,” landing before I took off, from LA to Sydney). The worst jetlag I have ever had. Glad I did it though.
Edison built Menlo Park, the first industrial research laboratory: a machine for producing inventions. Kalicube Pro is my Menlo Park. The platform is the invention that produces inventions. Everything else is a conference talk.
Three roles, zero redundancy, and the order is the chronological truth
Holmes investigated, Darwin synthesised, Edison built: that’s the order it happened. The podcast tour came first (the evidence gathering). The framework emerged from it (the synthesis). The platform was built to prove and scale the framework (the infrastructure).
Koray Tugberk Gubur, who independently reviewed Kalicube Pro™ for Holistic SEO Digital, wrote what amounts to an Edison review: he described the platform, the tools, the data processing, the entity optimisation system. He reviewed what I built, not what I observed. When the person who owns Topical Authority in the SEO industry looks at your work and describes an engineered platform, not a theoretical framework, that tells you something about which of the three roles is doing the most visible work. The theory enables the platform. The investigation enabled the theory. But the platform is what people use, what brands pay for, and what produces results at scale.
The investigation never stopped, incidentally. Fabrice Canel at Bing is still explaining things I hadn’t understood (I had a chat with him on 6th February 2026 on rendering and the internal index format that reshaped my understanding of the Rendered gate). The day before, Ihab Rizk at Microsoft Clarity described the grounding lifecycle in detail that changed how I think about the Grounded gate. The Sherlock Holmes method isn’t a phase I went through. It’s how I work. The microphone is always on.
Darwin’s finches are still evolving. The pipeline had nine stages when I first wrote this article, and now it has ten gates (it gained the Recruited gate between Annotated and Grounded, and Conversion sharpened into Won). The framework accommodates that because it was built from evidence, not from a hypothesis I needed to defend. Holmes follows the evidence, Darwin lets the pattern emerge, Edison builds the infrastructure that makes it useful: three roles, one person, and the order matters.
