A monthly freelance update about end-to-end analytics, a hard lesson from an old Google Ads account, building products, and the AI stack that actually worked.
Hi! This is my first monthly update, a format I have wanted to launch for a long time. The idea is simple: once a month I want to honestly share what is happening in my freelance work. No filters, no โfake success.โ Client work, tools, mistakes, lessons.
Why? First, it keeps me disciplined. When you know you will have to write a report, you have fewer illusions and more specifics. Second, when I was starting out, I was looking for this kind of content myself: not another โ10 tips for freelancers,โ but the real picture – what you do every day, what works, what does not, which tools matter.
March turned out to be interesting. No new clients, instead I focused on one-off projects, analytics for current clients, and building my own products. I also learned one very important lesson about old Google Ads accounts, and I want to tell that story in detail.
End-to-end analytics for two clients
๐ GA4 + GTM: from click to order
The main focus of March in client work was setting up a full end-to-end analytics chain. That means from the moment a person clicks an ad to the moment they place an order, every step is visible in the data. GA4, Google Tag Manager, Enhanced Conversions, and conversion actions in Google Ads, all connected.
Rozavilla – curtain and blind showroom
Rozavilla is a showroom with two offline locations in Ukraine, where they make curtains, blinds, and roller shades to order. I have been working with them for a few months now, and we are in the second month of active advertising.
The second-month numbers are already very interesting. The average order from ads is about UAH 20,000. The range is from UAH 7,000 to UAH 100,000 per order. In the second month, growth was 6x compared to the first. And this is not even the full picture: orders from the first month are still coming in, delayed payments are normal in this niche. The quarterly results in May will show the full picture, and that is exactly why end-to-end analytics was needed, to see the real ROI.
Tischa – B2B manufacturer
Tischa is a B2B manufacturer of hygiene products. I also set up full end-to-end analytics for them: GA4, GTM, and Enhanced Conversions.
But Tischa is tied to the main story of the month. It deserves its own section, because it is a lesson that applies to anyone working with Google Ads.
Lesson of the month: old Google Ads accounts are a trap
๐ When account history works against you
This is the most valuable lesson from March, and I want to share it in detail, because very few people talk about it even though it can cost real money.
Context: Tischa had a Google Ads account that had not been used for 4 years. Four years. When we launched ads on that old account, everything was extremely slow. Conversions only appeared 5-7 days after launch. The client thought that was normal for the niche. I thought so too at first.
Experiment: at the beginning of March I decided to test the hypothesis. I created a completely new Google Ads account, imported all the settings from the old one, the same campaigns, the same bids, the same site. The only difference was a new clean account versus an old one with four years of โdeadโ history. There were some minor UX changes on the site, but nothing major.
Results for the first 4 days:
- New account: 7 leads, 37 conversions, less than ยฃ200 spent
- Old account: 1 Shopping campaign + 1 RSA, ยฃ200 spent, ZERO conversions. Not even micro-conversions.
Same settings. Same site. Same budget. The only difference was the account. And that difference meant 37 conversions versus zero.
Why this happens: the Google Ads algorithm uses historical account data to optimize. If the account has been sitting idle for years, that old history becomes ballast. The algorithm is based on outdated data that is no longer relevant. It literally limits performance because it โremembersโ past inefficiency.
Practical takeaway: if a client has an old Google Ads account that has not been used for a year or more, seriously consider creating a new one. Do not try to โreviveโ the old one. Start with a clean slate. Yes, it takes extra setup time. But the results speak for themselves.
Bonus: because I am a Google Partner, the client received a ยฃ1,080 promo code. So the first campaigns on the new account were effectively free. Win-win.
What I built this month
๐๏ธ Building products and the website
March was a month of building. I did not take on new clients, and instead focused on one-off audits and on building the things that will work for me long term.
AdAstra website
A fully custom WordPress theme from scratch. Not a child theme, not a page builder, just clean PHP, SCSS, and JavaScript compiled with Vite. It is my portfolio site, and I wanted it to be a case study in itself: fast, beautiful, and built with proper SEO. Service pages, case studies, blog, and contact pages are all custom.
Article template system
I built a system with 9 article template types for the blog, from how-to guides to case studies. Each template has its own structure, schema markup, and a place for a PDF lead magnet. That is the foundation for scalable content marketing. Instead of building each article from scratch, I now have a ready structure.
Freelance OS CRM
My own CRM for freelancers. It is a tool I am building for myself, but I plan to open it to other freelancers too. It solves a problem I feel every day: tracking clients, projects, deadlines, and finances in one place, without noise. More about Freelance OS. Migration to the SPA V2 architecture is still in progress.
Agent Office MVP
A local dashboard for visualizing AI agents. When you work with several AI agents at once, which I do every day, you need an interface where you can see who is doing what. The MVP is ready, and I use it for myself.
One-off audits
A few one-off audit projects: website diagnostics, ad account reviews, and recommendations. No long-term commitments, just helping other businesses understand where they stand.
AI stack of the month
๐ค Claude + Gemini as the core workflow
My AI stack finally settled in March. The combination is Claude plus Gemini, with each tool doing what it does best:
Claude is the main workhorse. Code, content strategy, project management, data analysis, long documents. Claude keeps context better, gives more structured answers, and hallucinates less on complex tasks. This whole site, including the theme, articles, and automations, was built with Claude.
Gemini is for image generation, the 3D cosmic style you see on the site, and fast research. Gemini 2.5 Flash is very good at generating bright, stylized visuals, something Claude does not do and DALL-E does differently.
Key insight: specialized AI tools always beat the โone tool for everythingโ approach. Find each toolโs strengths and combine them. The real time savings for me came from doing less rework. The first answer is more often the one I can actually use.
Plans for April
April is the month of scaling. March was about the foundation and the analytics, April is about making the projects produce visible results.
- Launching the case studies section on the site. 15 case studies with a new template, content, and PDF lead magnets. This is the biggest chunk of work, but also the most important for conversion on the site.
- Quarterly Rozavilla report in May. Pull together the full picture: three months of advertising plus the effect of end-to-end analytics on optimization. That will be the first full case study with numbers.
- Scaling Tischa on the new account. Increase spend and campaigns on the new account, which performed far better than the old one.
- Q2 content plan: 11 articles. The content plan for April and May is already ready. Four series: AI for business, Google Ads guides, startup practice, and case studies. Every article will include a PDF, schema markup, and GEO optimization.
Ambitious? Yes. Realistic? With the right AI stack and project management system, absolutely. We will see in a month.
March recap
March 2026 was a month of analytics and building. I set up end-to-end analytics for two clients, learned a valuable lesson about old Google Ads accounts, and built infrastructure for my own business. No new clients, and that is fine. Sometimes you need to pause and build the things that will work long term.
If I had to choose one takeaway from the month: more clients is not always better. Sometimes better analytics and better tools are worth more than another project.
Thank you for reading. If you have questions or want to discuss something from this update, write to me on Telegram. The next update will be at the beginning of May.
Read also: My AI stack as a marketer ยท AI for business: 2026 data ยท Freelance OS ยท About me ยท Cases
March centered on GA4, GTM, Enhanced Conversions, and conversion tracking for Rozavilla and Tischa, so decisions could be tied to real orders instead of guesswork.
A custom WordPress site, a reusable article template system, Freelance OS CRM, Agent Office MVP, and a few one-off audits turned March into a month of infrastructure instead of chasing more retainers.
Claude handled code, strategy, and long-form work, while Gemini covered visuals and fast research. The real gain came from less rework and a stronger first usable answer.
A clean account outperformed a four-year-idle account with the same campaigns, site, and budget. Historical ballast is sometimes a real performance drag, not an asset.
March reinforced that stronger analytics, sharper tools, and better operating systems can create more leverage than forcing extra client volume too early.
Comments
Discuss on Telegram