The cohort report was filtering for Upwork-referral desktop sessions. One row came back. Thirteen sessions from a single visitor. In the middle of that data: a seven-minute read of /work/, cover-to-cover. Two course pages after it. Several articles. The visitor had read the room and come back a dozen more times.
No hire request came.
The prior post, Seven minutes on /work/ and no hire button: the problem wasn’t the copy, established that the structural gap was a missing conversion mechanism, not the quality of the text. This post documents what was actually shipped to close it. Four pieces, one day.
The hire strip
The last element on /work/ had been a case study that transitioned into the standard footer. No point of decision. No CTA. A visitor who had spent seven minutes evaluating the work arrived at the bottom of the page with no obvious next step.
The hire strip closes that. Ink-inversion panel, a deliberate visual break from the rest of the page. Blood-coloured left rule as the entry mark. A volt availability pulse: a small live indicator that the slot exists and is open. Then the hire CTA.
The retail framing is deliberate. I have a hospitality and retail background, and the shape of this failure was immediately familiar. The /work/ page functions as a sales floor. A visitor who spent seven minutes there was not browsing passively. They were evaluating. The problem was not an absence of product. It was the absence of a till.
A prospect who has made a decision and encounters no mechanism to act on it does nothing. They leave.
The counter-argument: a motivated buyer will find the /contact/ page themselves. The thirteen-session visit disproves that theory. This visitor was motivated. They returned to the site a dozen times. They still didn’t convert. “Motivated buyers find their own path” is a comfortable assumption. The data does not support it.
Out of scope for this iteration: auto-close on hire, follow-up sequences, conversion tracking below click level.
The What’s New band
Thirteen sessions means thirteen separate visits. Between those visits, the site changed: new pages shipped, /work/ entries updated, articles published. None of that was visible on the homepage. A returning visitor landed on the same hero they’d seen before with no signal that anything had changed since their last session.
The What’s New band sits directly below the hero. Blood top as the brand divider. Three most recent launches, each with a title and a destination link. The intent is simple: a returning visitor sees immediately whether anything is different before they have to navigate to find out.
This addresses a different failure. The hire strip catches decided visitors who need somewhere to act. The band catches the undecided ones still looking for a reason to keep going, and a prospect doing due diligence across thirteen sessions deserves a front door that acknowledges they’ve been before.
The band is manually maintained: no build-time feed, no CMS query. The trade-off is accepted. At current ship cadence, the update cost per launch is negligible.
The /contact/ timelines note
The /contact/ page opened with a paragraph explaining what I don’t take on. Minimum runway. Out-of-scope project types. The implication, read without goodwill, was that the reader might not qualify.
That copy was written as a quality signal. It reads as a bouncer.
The replacement covers the same ground on timeline requirements and project scope, but frames them as part of the offer rather than a pre-qualification filter. The information is identical. The direction it points is different.
Fixing this required reading the page as the thirteen-session visitor rather than as the person who wrote it. When you have spent seven minutes evaluating someone’s work and arrived at their contact page ready to ask about a project, the first paragraph should not open by listing conditions under which you will be turned away.
The fix took twenty minutes. It should have been obvious in the first draft.
The /security/ page
The security-by-design documentation had existed internally since the first pentest sprint: threat models, code-audit logs, a pentest report. None of it visible to a prospect evaluating whether to trust the work. An evaluating visitor reading the site carefully would find no signal that security was treated as a discipline, only that work shipped.
The /security/ page surfaces the philosophy publicly: what security-by-design means as a practice, what the process produces, and how that shapes decisions. No raw audit logs, no source code. Methodology and commitment, stated plainly.
The direct conversion angle is weaker here than for the hire strip. The intent is positioning. A prospect doing serious due diligence, the kind that runs to thirteen sessions, will find this page and have a question answered they may not have known to ask.
The source material was already written and audited. The sprint was surfacing it, not producing it.
The common thread
Four separate fixes, shipped on the same day. The connection between them: each was a point where a motivated, returning visitor could stall without any visible path forward.
Each gap maps to one of the fixes:
- No front door for fresh content → the What’s New band.
- No exit from /work/ for someone ready to act → the hire strip.
- /contact/ opening as a filter → copy reframed.
- Invisible security work → the /security/ page.
None of this required new positioning, copy strategy, or a rethink of what the site was for. It required reading the site as the visitor was reading it, finding where the path went quiet, and filling the gaps.
The GA4 data named the visitor. The visitor named the gaps.
What the data says next
Post-ship conversion data is not available yet; this shipped on the same day it was written. What was clear before shipping: the traffic wasn’t the problem.
One visitor, thirteen sessions, seven minutes of sustained engagement with /work/. That is high-intent traffic. The funnel wasn’t leaking because the wrong people were arriving. It was leaking because the right people arrived and found no till.
The hire strip is now there. The front door is labelled. The timelines note no longer opens with a conditional. The security work now has a surface.
What comes next is incoming data.
