Which proxies do you need for rank tracking, and how many?

Rank trackers send the same kind of request over and over: one keyword, one search engine, one results page. From a single IP, Google notices fast. You get the “unusual traffic” page, a CAPTCHA, or results that stop loading halfway through the run. The usual advice (“just buy residential”) skips what actually decides success: how many searches you push through at once.

A proxy for rank tracking has two jobs. It spreads your searches across many IPs so no single address carries the whole keyword list. And it makes the search come from the country whose rankings you want to see. A third job, city-level local rankings, is where IP-based proxies get expensive and where Storm is honestly not the right fit.

This guide covers which gateway to use, how to set up a desktop rank tracker and a custom script, the thread budget for Google, worked sizing for small and agency keyword lists, and what to do when the numbers you see don’t match a manual search.

Most comparison pages say residential, full stop. The real answer depends on volume and on how you pace requests. Google treats any IP that sends many automated searches with suspicion; a home IP buys you more patience, not immunity.

Rotating datacenter (Main gateway) gives every search a different address from a pool of 700,000+ IPs. Because no IP repeats often, per-IP limits rarely build up, and the flat price means you can check deep result pages without watching a data meter. This is where most desktop trackers and custom scripts should start.

Residential ports give one home IP per port for 5 minutes, shared by every thread on that port. That’s great for trust, but it means repeat searches from the same address inside each window. Ten threads hammering one port send all their searches from one IP for 5 minutes. If you use residential for rank checks, keep one or two search threads per port and add ports rather than threads.

A practical rule: start on rotating proxies. Move to residential only if CAPTCHAs stay high after you’ve cut concurrency to the quarter budget and added pauses.

Throughput depends on how fast Google answers and how long you pause, so treat these as planning numbers you check on your own run. Assume each search thread finishes one query every 10 to 15 seconds including a random pause. That’s roughly 240 to 360 searches per thread per hour.

• Freelancer, 300 keywords, Google USA, page one, daily: 300 searches. The 40-thread rotating plan gives 10 search threads; the run finishes in well under an hour, with 30 threads left for crawling or other tools.
• Small agency, 2,000 keywords, two markets (USA and EU), top 30: 2,000 × 2 × 3 pages = 12,000 searches a day. With 10 threads that’s 3.5 to 5 hours. With the 80-thread plan (20 search threads) it’s around 2 hours. Pick 80 threads if the report must be ready by morning.
• Agency, 10,000 keywords, daily, top 20: 20,000 searches. The 150-thread plan gives about 37 search threads, finishing in 2 to 3 hours. The 200-thread plan (50 search threads) gives headroom for retries.
• Weekly tracking only: divide the daily count by seven and size for the day you run it, not for the week.

Need to test first? The 10-thread package allows 2 or 3 search threads, enough to confirm your tracker works with the gateway, not enough to run a real list. When the list grows, the member area lets you upgrade in one click; the unused part of the current plan is credited.

Two things decide which country’s results you see: the exit IP and the search’s own parameters. Storm gateways give you a USA, EU, USA+EU or Worldwide exit. Within that, you steer the market with the query itself. Google documents gl (the searcher’s country, a two-letter code) and hl (interface language) for its Custom Search API, and rank trackers commonly pass the same names on the regular search URL. Most trackers set them for you when you pick a country.

Match all three signals: an EU gateway, gl=de, hl=de and a German Accept-Language header if you track Germany. Mismatches (a USA IP asking for German results in English) give mixed pages that won’t match what real searchers see.

What Storm can’t do: pick a city or state. Local pack and map rankings change from one city to the next, and trackers that need an exit IP in a particular city need a provider that sells that. If your tracker emulates location through search parameters instead of the IP, a country-level gateway is usually enough.

A position that differs from what you see manually isn’t always a proxy problem. Your own browser is signed in, has search history and sits in your city. The tracker is anonymous and country-level. Check by hand in a private window, signed out, with the same country and language settings.

Also check page depth. If the tracker counts organic results only and you count ads, maps and “People also ask” boxes too, numbers drift by two or three places. And if Google answers with a page asking for JavaScript, a plain HTTP client sees no results at all. In that case switch to a headless browser such as Playwright and budget around 10 threads per open tab.

The same rotating plan can feed a site crawler like Screaming Frog, backlink checks or competitor scraping, as long as the total open connections stay within the plan and Google work stays inside its quarter. Run the rank check overnight and crawls during the day, and a single 80-thread plan can cover both.