In The Vine ran conversion research manually for enterprise clients for years. Same methodology every time. Same bottleneck every time: the research and synthesis phase took weeks before a single test was written. This is the story of what changed.
In The Vine (inthevine.ca) was a CRO and SEO consultancy founded by Ryan Johnson. The client list included major retail brands — Silver Jeans, Warehouse One, Ricki's — and travel operators like Fresh Tracks Canada and Adventures Abroad. Across those engagements, the methodology consistently produced step-change results: tens of millions in added revenue from conversion rate improvements and SEO gains.
The work was grounded in a structured research process: GA4 analysis, heatmap review, survey synthesis, heuristic evaluation, and competitive benchmarking — all feeding into a prioritised hypothesis backlog. The framework was tight. The outputs were reliable. The problem was the timeline.
A full Conversion Research cycle took three to six weeks from kick-off to a prioritised test list. That wasn't a failure of execution. It was the honest cost of doing the work properly when every step required a human analyst to pull data from separate tools, cross-reference it, and write structured findings from scratch.
The consulting methodology was built around a simple idea: most CRO practitioners skip the research phase and jump to testing intuitions. The ones who do research usually do it once, produce a report, and move on. Neither approach compounds.
In The Vine ran research continuously — not just at the start of an engagement. Each test result fed back into the hypothesis backlog. Each GA4 change triggered a re-analysis. The methodology was designed to get sharper over time because the knowledge base grew with every cycle.
But doing that manually set a ceiling on how many clients the agency could serve at that depth, and how fast any single client could move. The synthesis step — reading GA4, heatmaps, surveys, and form data together and turning them into ranked, evidence-backed findings — was where most of the time went. It wasn't something that could be templated or delegated easily. It required judgment about what signals mattered and how they connected.
That was the constraint: the most valuable part of the process was also the most time-intensive part. Every client week started with research that should have been finished before the week began.
The decision to build Growth Roadmaps came from a specific observation: the research methodology wasn't proprietary intuition — it was a repeatable process. The same analytical lenses, applied in the same order, to the same types of data sources, consistently produced useful findings. That meant it could be encoded.
The first phase was mapping every step of the manual research process into a software workflow. GA4 connection and data pull. Heatmap session ingestion. Survey and review synthesis. Heuristic analysis against a structured framework. Buyer Panel evaluation. Scent and message-match analysis for paid traffic. Each lens that a human analyst had been running manually became a module.
The second phase was the synthesis layer — the part that had previously required a senior analyst to sit with six open tabs and write structured findings from scratch. This became the AI layer: a system that reads the outputs from each module, identifies patterns across sources, and produces ranked findings with the supporting evidence attached.
The third phase was closure: wiring the finding directly into a test. A finding that would previously have lived in a Notion doc or a slide deck could now become a running A/B experiment in one click, with the hypothesis already written and the variant editor ready to build the change.
The first complete Conversion Research run on a live client — the same scope of analysis that had previously taken three to six weeks of analyst time — finished in under an hour. The output was a ranked findings list with source evidence attached to each item. GA4 report, heatmap session, survey quote, form drop-off data — every claim linked back to the data it came from.
The velocity change was significant enough to alter the economics of the engagement model entirely. Research that had been a constraint became a repeatable, scheduled process. Instead of one research cycle per engagement, clients could run research continuously — triggered by traffic changes, test results, or a new campaign launch.
The knowledge base component compounded the gains. Every winning experiment added to the team's evidence library. Every test result — win or loss — made the next research run more accurate, because the AI read the accumulated history on every run and surfaced patterns the manual process would have missed.
For agencies running CRO at scale, the implication is direct: the research bottleneck that limits how many clients you can serve at depth, and how fast any single client can move through the testing roadmap, is removable. The methodology doesn't change. The timeline does.
The In The Vine story is the origin of Growth Roadmaps — but the constraint it solved isn't unique to agencies. Any in-house CRO team running research manually faces the same ceiling: the synthesis step takes longer than the testing step, which means the testing roadmap is always thinner than it should be.
Most teams respond to this by skipping or shortening the research phase. They test intuitions instead of findings. They run heuristic reviews instead of multi-source synthesis. The tests still run — they just produce smaller lifts on average, because they're not backed by the evidence density that comes from reading GA4, heatmaps, surveys, and form data together.
Automating the research phase doesn't change the methodology. It removes the time constraint that forces teams to take shortcuts. When research runs in under an hour instead of three weeks, you can afford to do it properly every time — before every test, after every result, and on a continuous schedule rather than as a quarterly project.
That's what Growth Roadmaps is: the In The Vine methodology, without the bottleneck.