Not a portfolio and not a list of skills. Every account here carries its problem, the change I made and the number that came out.
Each row is one account. The problem is where it started, the change is my work, the result is a number and the period it covers.
| Project | Problem | Change | Result |
|---|---|---|---|
| +30 %Outdoor equipment store | Campaign structure did not match the range, and the feed lacked the attributes people search by. | Audit, structural rebuild, then product feed optimization. | Growth above 30 %, cost of sale unchanged. Two more stores followed after the first year; we are in the third season. |
| 3×Store after eighteen months elsewhere | Budget poured into a structure that could not spend it well. | Rebuilding the structure before touching the budget. | Three times the orders at a ROAS above 5. |
| 7.58Store, 14 months in | Budget split evenly, regardless of where a click pays for itself. | Tactical campaigns where the margin carries the cost of a click. | ROAS 7.58 with orders tripled over the year. |
| 78 → 141New store, starting from zero | No history and no data for the algorithm to learn from. | Building the account, the tracking and the campaigns from scratch. | 78 orders in month one, 141 in month two at a ROAS of 4.27. |
| 3,366Catalog with thousands of products | Feed titles and attributes did not match how people search, and editing the store meant waiting on a developer. | A supplemental source for Merchant Center, with no change to the store. | 3,366 products rewritten in a day. More accounts run the same setup. |
| 3,000International marketplace | Hundreds of thousands of items across more than ten markets. Manual management makes no sense at that scale. | Building campaigns through the Google Ads API instead of clicking through the interface. | 3,000 granular PLA campaigns across 12 countries in a week. After a structural rebuild and feed work, revenue from US PLA campaigns grew by roughly 40 %. |
| 3 of 5Five inherited accounts | The Conversions metric mixed purchases with store visits and add-to-cart. | Sorting the conversion actions out before any optimization. | Three of the five were optimizing toward a number that did not mean revenue. Not any more. |
Four client accounts, measured against a control group of products from the same catalog. Each study links to the full report, which is in Czech.
The +719 % is conversion value, not clicks. Clicks went from 480 to 1,582, gross; net of the control group trend, +260 %. The enriched products grew well above the rest of the catalog.
Open the full report (in Czech)I am not reporting revenue yet. The conversion sample is small; I will add it once there is enough to defend.
Open the full report (in Czech)Roughly five times the traffic, but with a mixed signal on quality. The breakdown is in the report.
Open the full report (in Czech)The average across all products was only +8 %. Splitting it by subcategory is what showed where it landed and where it did not.
Open the full report (in Czech)„This is our third season working with Karel. We started with an audit and changes to campaign structure, then moved on to optimizing the product feed. The result was growth of over 30 % at an unchanged cost of sale. After the first year we handed him our two other stores as well.“
Petr Křivánek, OwnerHappySport / FreeskierBesides my own clients, two large brands.
Consumer electronics, accountable for revenue and return. A large catalog and margins where a bad decision shows within a week.
Today I lead Shopping campaigns and product data across more than ten markets. Hundreds of thousands of items and thousands of granular campaigns built through the API.
Online stores with budgets from $10k to several hundred thousand a month. Fashion, electronics, automotive and supplements.
Most clients have a confidentiality clause. So I quote the numbers anonymized, without naming the store.
For the feed enrichment case studies I measure against a control group of products from the same account and subtract the seasonal trend. Where the data will not carry that, I say so next to the number instead of claiming credit for it.
I reply within 24 hours. The call takes 30 minutes.
On the call I will tell you what I would fix first and whether it is worth doing at all.