Check your real rankings. Not your personalized results.

What daily rank tracking costs

A search results page is heavy, around 120 KB gzipped once you keep the markup and drop the images. Volume follows from your keyword list:

This is one of the few jobs where datacenter is not a cheaper option, it is not an option. Search engines challenge hosting ranges within a handful of queries. Eighteen dollars a month buys the whole programme on residential, so the money question here is smaller than the correctness question.

A new address per query, and locale that matches the market

Leave the session id off entirely. Every query goes out through a different address, which is exactly the pattern search engines do not flag:

Python: rotate per query, pin the location

def serp_proxy(country, city=None):
    login = f"login_c_{country}"
    if city:
        login += f"_city_{city}"
    return f"http://{login}:password@proxy.sotaproxy.com:10000"

for kw in keywords:
    for market in markets:
        proxy = serp_proxy(market.country, market.city)
        r = requests.get(SERP_URL, params={"q": kw, "hl": market.lang, "gl": market.country},
                         proxies={"http": proxy, "https": proxy}, timeout=30)
        time.sleep(random.uniform(2, 5))

• Set the language and country parameters as well as the proxy country. The address decides which index you reach, the parameters decide which version of it you see.
• Local packs move by city, so pin _city_ when you track anything with a map result. Country level tracking of a local business tells you nothing useful.
• Randomise the gap between queries. A perfectly even two seconds is a machine signature all by itself.
• Track from the same locations every day. Changing the sampling geography changes the numbers, and you will read it as a ranking move.

Why rank data drifts for no reason