How Many Proxies Do You Need? Count Identities, Then Measure Throughput
By the ProxyV4 Engineering Team · Updated:
The number of proxies you need comes from one of two questions. For account and client work, count how many separate identities need their own stable IP; that number is your proxy count. For data collection, divide the request rate you need by the rate one IP can sustain on that site within its limits, a rate you measure rather than look up. No site publishes a safe per-IP number, so the honest method is to test one IP, then size the order from what you measured.
How many proxies do I need? Start with the job
Proxy count is not a property of the proxy; it is a property of your work. Separation jobs (client accounts, team workspaces, test environments per market) need one IP per thing you are separating. Throughput jobs (price monitoring, rank tracking, collecting public pages) need enough IPs to carry the request rate without exceeding what each site tolerates. Many teams have both, and they size each part separately.
A useful first step for either is to buy one IP, run the real workflow on it for a few days, and write down what you learned. That turns every number below from a guess into a measurement.
Separation jobs: proxies per account or workspace
List the identities that should look like they are used from separate places because, in reality, they are: each client of an agency, each business unit, each regional store, each test market. Each of those gets one static IP, kept for as long as the identity exists.
Several accounts that belong to the same client or business can reasonably share that client's IP, because they would share an office network in real life. Accounts for unrelated clients should not share. Platforms publish no ratio of accounts per IP, and a proxy does not change their rules on multiple accounts, so follow each platform's terms and use its business or manager tools where they exist.
A quick count
- Clients or business units you manage: one IP each, in the country where that client's accounts are normally used.
- Team members who must work from a fixed address (for example, a partner system expects a known source IP): one each, or one shared if they work as one team.
- Markets you test or monitor manually: one per country.
Throughput jobs: proxies for scraping volume
Many sites limit requests per IP, and spreading requests over several IPs spreads that load. That only works within the site's rules: a pool does not entitle you to more than a site allows overall, and it does not change a site's terms on automated access. Respect robots.txt, rate limits and each site's terms; when a site offers an API, use it. Within those limits, the arithmetic is simple.
The formula
Required rate = total requests ÷ the time window you have. Proxies needed = required rate ÷ the per-IP rate you measured, rounded up, plus whatever headroom you want for retries and slow responses. Every input is yours: the job size, the deadline, the measured rate and the headroom.
A proxy count calculator in Python
The function below does the arithmetic. The values in the example call are placeholders to show the units, not recommendations; replace them with your own job size and your measured per-IP rate.
import math
def proxies_needed(total_requests, window_hours, per_ip_per_min, headroom=1.0):
"""Return the proxy count for a job; per_ip_per_min is what you measured."""
required_per_min = total_requests / (window_hours * 60)
return math.ceil(required_per_min / per_ip_per_min * headroom)
# placeholder inputs: replace with your own numbers
print(proxies_needed(total_requests=50_000, window_hours=10, per_ip_per_min=8, headroom=1.25))Measuring the per-IP rate
Run one IP against the target at a modest pace and raise it gradually over a few sessions while you watch status codes. HTTP 429 or a Retry-After header means you are over the site's limit: honor Retry-After, back off, and treat the last pace with no 429s as the ceiling. Size the pool to stay well below that ceiling, not at it, because limits can change and differ by page type.
Concurrency is a separate question
One IP can carry many simultaneous connections; the proxy does not require one IP per thread. What limits concurrency is the target: a site that tolerates a certain request rate per IP does not care whether those requests come from one connection or ten. So set concurrency from your measured per-IP rate and your response times, and size the pool from the formula above. If your client library opens far more connections than the rate needs, lower the connection count rather than buying more IPs.
How ProxyV4 fits the count you arrived at
ProxyV4 sells static IPs by the piece, which suits both kinds of sizing: you can order exactly the number you counted, starting with one. Private datacenter IPv4 is available in the United States, Vietnam, Japan, South Korea, Singapore, the United Kingdom and Hong Kong; static ISP in the United States and Vietnam. Each IP has one tenant for the term. Traffic on private IPv4 and static ISP is unmetered, so a larger crawl changes how many IPs you need, not a bandwidth bill.
Pricing is per IP per month in USD, with 1, 3 or 12 month terms or a custom number of months, and the per-IP price falls as the order grows; see live prices. Renewing before expiry keeps the same IPs, and you can renew all of them or only a selection, so a pool that shrinks after a project does not cost you the IPs you keep. Every proxy supports HTTP(S) and SOCKS5 with username and password login.
Where it does not fit: if your design depends on a new IP for every request from a large rotating pool, ProxyV4 is the wrong tool; rotating proxies (Vietnam) are waitlist only, and there are no mobile proxies. IPv6 proxies are a separate product that only works on IPv6-ready sites. A proxy that fails for a technical reason, such as not connecting or showing the wrong location, is credited to your balance under the refund policy. For background on the rotating vs static trade-off, see static vs rotating proxies.
Keep the pool healthy after you scale
Check every new IP before it goes into production: paste the line into the proxy checker or run curl --max-time 15 -x http://USER:PASS@IP:PORT https://proxyv4.com/api/my-ip. In your code, track errors per proxy so you can see if one IP behaves differently, back off on 429 instead of switching IPs to keep going, and keep timeouts on every request. Our web scraping proxy guide covers the client side in more detail.
Revisit the count when the job changes. A new client means one more IP; a smaller crawl means renewing only the IPs you still use.
Frequently asked questions
How many proxies do I need per account?
Count separate identities, not accounts: one static IP per client, business unit or market, shared by accounts that belong to it. Platforms publish no safe ratio, so follow each platform's own rules on multiple accounts.
Is there a standard number of requests per minute per proxy?
No. Each site sets its own limits and rarely publishes them. Measure with one IP, watch for 429 responses and Retry-After headers, and stay well below the point where limits start.
Do I need one proxy per thread?
No. One IP can carry many connections. Size the pool from the request rate the target tolerates per IP, not from your thread count.
Will a bigger crawl cost more in bandwidth?
Not on ProxyV4 private IPv4 or static ISP: traffic is unmetered. You pay per IP per month, so only the number of IPs changes the price.
Can I start with one proxy and add more later?
Yes. You can order a single IP, test it, then order more; the per-IP price falls as the order grows.
Start with one IP, scale with bulk tiers
Order one private IPv4 or static ISP proxy, measure your real workload on it, then add exactly the number of IPs you need. Traffic is unmetered and the per-IP price falls as the order grows.
See live prices