ABOUT ME
A performance marketer who builds.
I'm Marcel Jendrusch, a senior performance marketing manager who got tired of waiting for tools that think like a media buyer. So I built one. For 3.5 years I've owned paid growth across Meta, TikTok, Google and Apple Search Ads, with up to €150k in monthly budget and over a million app downloads to show for it. Adscalr is the AI ad-intelligence platform I designed, coded and shipped solo in three months.
HOW I WORK
Ads and funnel are one system.
Most teams optimize the ad and the landing page in separate silos. I don't. My background is mobile subscription growth, where CAC, purchase rate, payback and retention decide whether you scale, and the same funnel logic drives e-commerce through intent, conversion and LTV. Every decision (hook, audience, budget, page) gets measured against payback. That full-funnel view is exactly what I encoded into Adscalr: a closed loop that analyzes, generates, measures and learns.
THE BUILD
What three months solo actually taught.
Anyone can wire an API to a language model. The hard part was judgment: knowing where rigor pays and where it is theater. These are the calls I made, and why.
Daily, not real-time, on purpose
Competitor ad libraries do not change by the hour, so I refresh them daily and spend the saved compute on depth. A real-time pipeline would have cost more and taught nothing new.
Every score earns its confidence
Format-specific shrinkage means a UGC clip needs six times the impressions of a static before it ranks, because UGC is volatile. A lucky twelve-impression test never crowns itself a winner.
Significance gates before any pattern
Pattern mining only fires past two hard gates, a minimum sample and a minimum effect size, plus a false-discovery correction. Most AI ad tools walk straight into the multiple-comparisons trap. This one does not.
Compute spent where nuance pays
Cheap models read what is in the frame, mid-tier models handle language and personas, the most expensive model writes only the final strategic report. The cost curve was a design constraint, not an afterthought.
The model never gets to lie
Whenever the AI puts a number on something, a score or a ranking, the code recomputes it independently instead of trusting the model. And any metric stays hidden until the data behind it is solid enough to show honestly.
FEATURED PROJECT
Adscalr, built solo, in three months.
A complete AI toolkit for performance marketing, built on Claude with custom agents I designed myself. Not a wrapper, a closed-loop learning system. Solo in three months meant the hard part was judgment, not code: I refresh competitor data daily instead of in real time because ad libraries do not change by the hour, and I shrink each score by its format's own volatility, so a static earns trust faster than a noisier UGC clip and a lucky twelve-impression test never crowns itself a winner.
- Automated competitor analysis across Meta, TikTok & Google, spotting winning angles, pain points and creative patterns
- Data-driven creative concepts, UGC scripts and storyboards grounded in proven market mechanics
- ML-assisted budget allocation, fatigue detection and statistical validation (Bayesian shrinkage)
- 11 Guardian Skills that auto-review every output for statistical accuracy, strategic coherence and cost
- Full dashboard: competitor intelligence, ad-performance tracking and a creative workshop
EXPERIENCE
Where I've done it.
Owned up to €150k monthly ad budget across Meta, TikTok, Google and Apple Search Ads. Drove 1M+ downloads on iOS and Android and lifted store-to-install conversion from 18% to 35% through systematic creative testing. Budget and scaling decisions ran on CPI, ROAS and payback. Built AI workflows and Claude agents to accelerate creative iteration and market intelligence.
Managed paid acquisition for Legacy Academy across Meta, TikTok and Google on a growing budget, and established the systematic creative and audience tests that founded the product's early user-acquisition structure. Part of the 3–4 person marketing team behind ~300k Instagram and 600k TikTok followers and 600M+ organic views.
Built and moderated the community channels (Discord, Telegram, social) from day one, and set the communication guidelines and brand voice for the community.
Sole operational ownership of an early-stage product: day-to-day operations, brand communication and the workflows that kept both running.
Created SEO-optimized content and grew organic visibility for digital platforms.
Ran the online shop end to end: catalog and product copy, on-page SEO and campaign support for a building-supplies retailer.
Digital strategy, SEO and content for clients, plus my own YouTube and Twitch channels. Ads were part of it from the start: Google AdWords campaigns for my own affiliate projects.
SKILLS & TOOLS
The toolkit.
Meta Ads · Google Ads · TikTok Ads · Apple Search Ads
GA4 · GTM · Conversion API · Search Console · Attribution (Adjust)
Claude (daily) · custom Claude agents · AI workflows · prompt engineering
ROAS / CPA / payback optimization · creative strategy · funnel analysis · A/B testing · budget allocation
Dashboard development · HTML/CSS · SEO · Shopify · email automation
German (native) · English (fluent)
EDUCATION
Background.
Focus: Marketing & Business Psychology
Economics & Administration
Informatikkaufmann
Hiring someone who takes an idea from zero to a working product?
That's exactly what I do. Let's talk.