IsabellaOzols-Garland

LinkedIn GitHub

AI-native marketer. Three passports — Australian, UK, Latvian. Ask me about my family’s hidden treasure lore. Addicted to Zapier workflows. A lover of television and reading, and a person with a lot of opinions — because opinions, expansive research and experience are the path to building good taste.

fig. 01 — image to come

Website migration & design: Vortexa.com

Moving an energy-analytics site off a CMS it had outgrown — and redesigning the pages that carried the most weight while they moved.

Vortexa sells clarity about where the world’s energy is moving. The website did not read that way. Content had accumulated across two CMS platforms and several redesigns, product pages disagreed with each other about what the product was called, and the marketing team could not ship a page without an engineer.

I ran the migration as a content problem first and a platform problem second. Before anything moved, I inventoried every live URL, tagged each one by traffic, backlinks and sales use, and made a call in the open: keep, merge, rewrite, or retire. Roughly a third of the pages were retired outright. That decision — made before design, with search and sales in the room — is what kept the migration from becoming a redesign of everything at once.

The page system that came out of it is small on purpose: a product page, a solution page, a resource page, a proof page. Each one is a set of blocks the marketing team can reorder without asking anyone. I wrote the block library with the sales narrative in mind rather than the layout — a claim block, an evidence block, a data-visual block, a next-step block — so that a page’s argument survives being rearranged.

Design work concentrated on the pages doing commercial work. Vortexa’s own product visuals are dense and good, so the pages got out of their way: one idea per screen, generous space around the data, type that lets a shipping analyst read three paragraphs without effort. Where the old site reached for stock illustration, the new one uses the product.

Redirects were the unglamorous half of the project and the half that protected the result. Every retired and merged URL was mapped by hand, checked against search data, and monitored after launch, alongside metadata, structured data and internal links rebuilt against the new information architecture rather than copied across.

What I would keep from this project: deciding what not to migrate is the highest-leverage decision available on a site this old. What I would change: I would have brought sales into the URL inventory a fortnight earlier — they knew which pages they actually sent to prospects, and that list did not match the analytics.

role — marketing & design · scope — migration, IA, page system

The calendar

TV Calendar: Keeping on top of Substack

Using Claude Code to create a calendar of TV shows to watch, and generate a schedule I can stick to.

I like TV! For a while I wrote a Substack about TV, and I've recently brought that back.

With one issue. See, I have specific filters I use for deciding what shows to post about. They're usually American, and they have to have aired on one of the major streaming platforms (Excluding Netflix). They must also be in the English language, and I only want to watch the series premiere - not returning series.

Simple, right? Trakt.tv was my go to - spend £16/year on a subscription for a calendar view that synced to my Gcal, everything is great.

Except that Trakt.tv sold its soul to the devil, and I'd be better off smearing mud on my face than using their website anymore. So I look elsewhere - Simkl sounds promising, next-episode.net, maybe. Except that they're both awful, and both need subscriptions. I will not settle!!

So I had a think. Where did Trakt get its data from? Well, turns out it's from TMDB, and TMDB has an API. Cue Claude Code.

The writing was never the bottleneck — the calendar was. A show’s finale is the one moment people want to read about it, and I kept finding out about that moment two days late, from someone else’s post.

So I built the boring version of the fix. Air dates for everything I’m following flow from a TV metadata source into a single calendar, and each episode arrives as an event with the show, the episode number and a link back to my notes. A second automation walks ahead of that calendar and creates a draft in my writing app a few days before anything worth covering airs, pre-filled with the show name, where I left off, and the questions I had after the last episode.

The decision that made it work was refusing to build a dashboard. I do not need a view of my content pipeline; I need a draft sitting in the place I already open every morning, with the first sentence already impossible to avoid. Everything else in the system — tags, statuses, backlogs — got deleted after the first month because I never once looked at it.

The rig is deliberately unclever about failure. If the metadata source is wrong about a date, the event is wrong and I fix it in the calendar like any other appointment; nothing cascades. If I ignore a draft, it stays a draft and quietly stops nagging. Automation that can only ever suggest is much easier to live with than automation that assumes.

What it changed: I now publish against the television schedule instead of behind it, and the week’s writing decisions get made on Sunday rather than every evening. What I would build next: a view of the season ahead, so I can choose to skip a show early rather than discovering in week six that I have nothing left to say about it.

role — built it myself · stack — zapier, calendar, substack

fig. 03 — image to come

Reaching new demographics: Muscle Car UK

Taking a British enthusiast brand to an audience that had never been spoken to directly — without losing the people already there.

Muscle Car UK had a devoted audience and a narrow one: men who already knew what they were looking at. Everything the brand published assumed that knowledge — model codes in headlines, no explanation of why any of it mattered. Growth had flattened, and the obvious read was that the market was full. It wasn’t; the door was just hard to find.

I started with the comments and the inbox rather than a persona workshop. Two groups kept appearing who nobody was writing for: people who had inherited a car and were suddenly responsible for it, and people who liked the culture — the design, the sound, the era — without owning anything. Both were being quietly filtered out by the shorthand.

The strategy was not to make the brand more beginner-friendly, which would have annoyed the existing audience for nothing. It was to add entry points beside the existing content. Explainers written as stories rather than glossaries. Ownership pieces about cost, storage and paperwork — the questions people are embarrassed to ask in a forum. And a photography-led strand that treats these cars as objects, which needs no prior knowledge to enjoy.

Channel choices followed the audience rather than the brand’s habits. The enthusiast core lives in long-form and email, so that stayed exactly as it was. The new audiences were found in short vertical video and in interest-led search, so those got their own formats, their own captions and their own tone — the same voice, fewer assumptions.

I measured it on two axes on purpose: whether the new work brought in people who had never engaged before, and whether the existing audience’s engagement moved at all. The second number mattered as much as the first, because the failure mode of a repositioning like this is winning strangers and losing regulars.

What I would keep: reading the inbox before writing the strategy. What I would change: I underestimated how much editorial guidance the new strands needed — the tone held only once the rules for it were written down rather than demonstrated.

role — audience strategy & content · scope — positioning, formats, channels

fig. 04 — image to come

Say hello

Marketing work, automation problems, or an argument about a television finale. A paragraph is plenty.

Message me on LinkedIn GitHub Back to top

london · 3 case studies