ReventaData

Fair point, thanks. The 11 slugs are the Actor's own allowlist, taken from the category filter links on gumtree.com when it was built in August. Each one is sent as Gumtree's search_category URL parameter. It's a short list of common categories, not Gumtree's full tree, and only cars was benchmarked for this post. The allowlist exists so a typo fails the run instead of maybe returning an unfiltered search that looks like success. I've added this to the post.

Thanks for the careful read, you are right on both points. The post now installs with pip install "apify-client~=3.0", because a plain pip install apify-client leaves an existing v2 in place and the code relies on the v3 typed run object. I also added a second sample row from a saved bol.com page that has a reference price, so the referencePrice, referencePriceLabel, referencePriceCrossedOut and discountPercent fields are shown filled in.

Great catch, @OnlineProxy! You're completely right on both points: Price formatting: For unpriced listings (price: null), evaluating ad.get('price'):, raises a TypeError in Python. A defensive formatter like f"R$ {price:,.0f}".replace(",", ".") if price is not None else "Sob consulta" avoids any crash and respects Brazilian currency formatting. Run status verification: Checking if run.get("status") != "SUCCEEDED": before touching the dataset prevents reading empty/partial results if the Actor timed out or failed. I've updated the script in the guide to incorporate defensive null checks across price, city/state, and run status. Appreciate the sharp peer review!