Evidence system
Most agency case studies are assembled after the outcome is known. By then, the awkward parts are easy to leave out.
PROOF starts earlier.
This is where we publish DankRank’s experiments, launch baselines, build records and updates while the evidence is still developing. That includes zeroes, failed tests, limitations, reversals and ideas that mattered less than we expected.
I don’t want this to become a wall of victory stories. If we only publish the work that makes us look smart, it isn’t much of an evidence system.
Index
BUILD
The site itself is part of the operating record.
The build record covers the decisions made before code, the rules Claude Code had to follow, what changed during implementation and the problems we caught when the site reached a real browser on mobile, tablet and desktop.
BASELINE
DankRank begins without an established search history. That gives us a clean starting point before there’s a flattering period to choose from.
The launch baseline records AI citations, mentions, recommendations, organic visibility, referral evidence and the zeroes that come with a new domain.
EXPERIMENT
Experiment 001 is a public 90-day test of whether a new domain can begin earning citations for non-branded prompts in ChatGPT.
The protocol is published before Day 0 so the prompts, definitions and timing can’t be rearranged after we see the result.
01
If a citation appears after we change a page, that is an observation.
It may support an idea, but it doesn’t prove the page change caused the citation. Search never moves one variable at a time. Crawling, retrieval, rankings, source selection, prompt context, model behaviour, competitors and the rest of the web can all change while a test is running.
PROOF should make that uncertainty visible. Every experiment needs to show what changed, what we watched, when we watched it and what else may have influenced the result.
02
Not every experiment produces something worth celebrating. A new site might launch without earning a single AI citation. We could run a test for weeks and see no meaningful change, or discover that the results move around too much to support a confident conclusion.
That doesn’t make the work useless.
Zero gives us an honest baseline. If nothing improves, the idea may have been wrong. If the evidence is inconsistent, we may need to fix how we’re measuring the test before changing the page again.
I’d rather publish that than polish an inconclusive experiment into a success story.
an honest baseline
03
A ranking graph isn’t the only way to judge how an agency works.
Look at how this site was built. The architecture, keyword ownership, content QA, performance decisions, Claude Code controls and post-launch change process reveal quite a bit about how we approach search work.
A build record has limits. It can’t prove we’ll produce the same outcome for a client, and we won’t present it that way. What it can do is make our decisions visible, including the trade-offs, corrections and parts that didn’t go according to plan.
That gives you something more useful than another polished claim about our process. You can inspect the process yourself.
04
Founder-led work by Vee Popat, originally delivered through ColaDigital before DankRank launched. The original engagement and results belong to ColaDigital. The strategy reflects the search practice Vee now brings to DankRank.
05
I don’t want to blur those two things.
Methods contains the reasoning behind our work. It explains why we treat search as one connected ecosystem, why measurement comes before optimization and where human judgment still needs to take control.
PROOF keeps that reasoning honest. It shows what happened when we applied an idea, including the occasions when the evidence didn’t support what we expected.
If repeated evidence changes my view, the Method should change with it.
06
One experiment can tell us what happened in a particular situation. It can’t tell us that the same thing is happening across an industry.
That requires a larger question, more observations and a method someone else can examine. Research is where we bring together broader prompt sets, multiple engines, different brands or repeated tests to see whether a pattern actually holds.
We’ll publish how the work was conducted alongside what we found. We won’t take one result from DankRank, stretch it beyond what it proves and call it an industry benchmark.
07
Start with the free Visibility Baseline.
We’ll review the current footprint, compare one meaningful competitor and show you the three things we’d investigate first.
Free. Founder-reviewed. Delivered within 2 business days. No call required.