SparkProxy vs SOAX: What Targeting Precision Costs

Two gates stand in front of the precision question, and both are about your targets rather than either vendor.

Gate one: does your target accept datacenter address space at all? SOAX sells consumer IPs. Its docs are explicit that there are two self-serve network types, "residential (WiFi) and mobile", and the pricing table lists exactly those two on all five plans. The ISP and datacenter product pages both returned HTTP 404 on 9 October 2026. A couple of pages still reference them: the SOCKS5 page carries the title "SOCKS5 Proxies - Available for Residential, Mobile, Datacenter, ISP", and the United States location page still offers "American residential, mobile and US Datacenter proxies available" with an "Explore US Datacenter proxies" link. So the categories exist in SOAX's marketing, but no price is published for them and we won't guess one. SparkProxy sells rotating datacenter IPs: 1M+ premium addresses across 80+ countries including 50,000+ in the US, unlimited bandwidth on 1 Gbps ports, flat plans from $75 a month, and a 24-hour trial with no card. If your target rejects datacenter ranges on the first request regardless of fingerprint, this comparison is over and SOAX is the buy. Sort that first, with residential vs datacenter proxies as the reference.

Gate two: how fine does your targeting need to be? SparkProxy's published documentation describes country-level targeting only: country_code as an ISO 3166-1 alpha-2 value on the Scraping API, country pool selection on the proxy plans. Nothing in our docs describes region, city, ISP, ASN or ZIP parameters, and we won't imply otherwise. SOAX documents all six. If country is enough, you're comparing a meter against a flat plan. If it isn't, you're comparing a product that does the job against one that doesn't, and the rest of this page is what the job costs. The vocabulary, if any of it is new, is in what geo-targeting means in proxies.

SOAX documents six targeting parameters for both residential and mobile traffic, and publishes which plans expose which. Here is the ladder, with the cheapest plan that reaches each rung.

The plan column is quoted from SOAX's pricing comparison table, which reads "Country, Region, City, ISP" for Sandbox, Builder and Team, "As Team + ASN, ZIP" for Scale and "As on Scale" for Enterprise. The developer FAQ says the same thing in plainer words: "Is ZIP-code targeting supported? Yes, on Scale and Enterprise plans only."

Two structural facts matter before you read a single price. SOAX requires at least one targeting parameter in every request, with country-any meaning "the full global pool without geographic filtering", so there is no untargeted default. And targeting itself carries no surcharge: rates vary only by the geo tier of the egress country, and SOAX is precise about which country counts. "The egress IP is the proxy IP that connects to the destination. Its country determines the geo tier. Neither your location nor the destination's location affects the geo tier."

So a narrower filter never raises your per-GB rate. What it raises is the risk that the eligible pool gets too thin to work, and no published number lets you check. SOAX claims 155M+ residential IPs across 195+ countries, 30M+ mobile in the docs against 33M+ in the product page title, a combined "191 million residential and mobile IPs" on the homepage that doesn't reconcile with 155M plus 30M, and separately "12M+ Unique IPs (past 30 days)". The counting window and the deduplication method aren't published. Exactly one per-country figure is: the United States page states "6,843,600 IPs available". There is no per-city or per-ZIP figure anywhere, from SOAX or from us. Treat all of it as vendor claims and measure the rung you need on a trial.

At SOAX the country you exit through is the main input to your bill, which is unusual and worth internalising. Every country sits in one of three geo tiers, and SOAX publishes the membership in full.

• Tier 1: 32 named countries, including the United States, United Kingdom, Germany, France, Japan, Canada, Australia, China, Hong Kong, Korea and Taiwan.
• Tier 2: roughly 60 countries, including India, Brazil, Russia, Turkey, Singapore, Israel and South Africa.
• Tier 3: "all remaining countries", with the Philippines, Kenya, Ecuador, Iceland, Malta and Slovenia named as examples.

The rates, as published on 9 October 2026 in USD excluding tax:

SOAX's own FAQ states the consequence bluntly: "Routing through Tier 1 instead of Tier 3 can cost up to 8x more per GB on the same plan." Your country mix is therefore a pricing variable at SOAX, not a detail. Before modelling anything, take a month of your own logs, total the bytes per exit country, and look each of those countries up in the current tier list. A blended rate somebody else calculated is worthless to you.

At SparkProxy the same decision has no price attached on the proxy plans, because bandwidth is unmetered and the country is a pool selector rather than a rate band. On the Scraping API a country-pinned request costs a flat 5 credits more whichever country you pick:

# Plain HTTP fetch (1 credit) pinned to a German exit (+5 credits) = 6 credits
curl -H "X-API-Key: YOUR_API_KEY" \
  "https://scrape.sparkproxy.io/api/v1?url=https://example.com/pricing&render_js=false&country_code=DE&format=md"

# Same job through the rotating gateway: country goes in the username, bytes uncounted
curl -x http://USER-country-de:PASS@gateway.sparkproxy.io:11000 https://example.com/pricing

That is the whole country-level comparison. One vendor charges you differently for a German gigabyte than a Kenyan one. The other charges you the same for both and doesn't count gigabytes at all. Which of those is cheaper depends entirely on your volume, and the plan-fee arithmetic further down settles it.

This is where SOAX earns its reputation, and the pricing is better than most buyers expect. Region and city targeting are on every plan including the free Sandbox, they carry no surcharge, and they work identically on residential and mobile traffic. An engineer can validate city-pinned routing before paying any subscription at all. The only money involved is the $25 minimum top-up, since "Requests stop once your credit balance reaches zero" and Sandbox ships with no credits.

SOAX's docs make the granularity concrete. Their own worked example pins traffic to Comcast subscribers in Los Angeles, and the rule string prefixes the package key on a single gateway:

# Verbatim from developers.soax.com/proxies/residential, 9 October 2026: their own
# Comcast-in-Los-Angeles example, as the credential that prefixes the package key
country-us-city-los_angeles-isp-comcast:YOUR_PACKAGE_KEY

# Assembled by us from the parameter and rotation tables on that same page: adds a region,
# a session and 10-minute rotation. Every rule is documented, this exact string is not.
country-us-region-california-city-los_angeles-isp-comcast-session-a1b2c3-rotate-timed_600

Country codes are lowercase alpha-2, rules are hyphen-separated and order-independent, multi-word values take underscores, and that one gateway on port 1337 carries HTTP, HTTPS and SOCKS5. If you authenticate by IP instead of user and password, those rules move into the subdomain and hit the 63-character DNS label limit, which is why the docs publish a short form for every parameter (c, r, ci, i, a, z, n, s, plus sa for subaccount and oe for onerror). Worth knowing before you generate a few thousand city-pinned endpoints and find half of them unresolvable.

SparkProxy has no equivalent parameter, so city work goes elsewhere. What it's for, and how to test whether a provider's city labels match reality, is in city-targeted proxies. The test matters more than the parameter: a city filter that silently falls back to the country pool looks identical in your logs to one that works.

The isp- parameter is one rung, two products. On residential traffic it selects the access network, isp-comcast style. On mobile traffic the same parameter selects the carrier, which is a different capability wearing the same syntax. SOAX sells a named 3G, 4G and 5G mobile product and prices it at parity with residential: "Both network types share the same credit pool and burn at the same rate per geographic tier. There's no price difference between residential and mobile traffic in the same country." Both are enabled by default on new packages.

Carrier-level targeting with no mobile premium is a real capability gap in SOAX's favour, and it's available from the free Sandbox. SparkProxy's published specification describes no mobile carrier pool and no ISP-level parameter. If your work involves APN-gated configuration, carrier-specific app behaviour, zero-rated content, on-net redirects, or targets that block residential ranges while tolerating carrier NAT, SOAX has a documented product and we don't. Why carrier IPs behave differently from household ones is in carrier-targeted mobile proxies. One caution for any vendor selling this rung: nobody breaks pool claims out by carrier, so you can't tell from a pricing page whether a given MNO has 50 nodes behind it or 50,000. Sandbox exists so you can find that out for $25.

Here the ladder stops being free. asn- and zip- appear only on Scale and Enterprise, so the cheapest route to postcode-level or autonomous-system-level targeting is $1,500 a month plus VAT. There is still no per-request surcharge. The cost is the plan floor.

What that floor actually costs you depends on how much traffic you move, because the plan fee is prepaid usage rather than a separate platform charge. The table below is our arithmetic on SOAX's published rates: Tier 1 exits, USD excluding tax, assuming you top up beyond the included credits at your plan's rate, which SOAX permits on any plan at a $25 minimum.

Read the bottom three rows first. Above roughly 682 GB of Tier 1 traffic a month, Scale is the cheapest plan you can buy anyway, so ASN and ZIP targeting arrive free. Below that, they are the most expensive feature in this entire comparison, and at small volumes they are absurdly expensive per gigabyte moved. A team pulling 50 GB a month for postcode-level price checks is paying $26 a gigabyte for a product whose list rate is $3.00.

The break-even moves with your geo tier, which is the second-order effect almost nobody models. Cheap traffic means the plan fee dominates for longer. Team's Tier 3 rate is $0.90/GB, so it takes 1,667 GB before a Team plan plus top-ups reaches Scale's $1,500 floor, against 682 GB on Tier 1. Collect in low-cost geographies and the ASN door stays shut nearly a terabyte longer.

SparkProxy sells no product at this rung, so there is no column for us in that table. What ASN-level selection is for, and when subnet diversity is the thing you actually wanted, is in ASN-targeted proxies.

The insight that reframes this whole comparison comes out of SOAX's own documentation.

The docs state that "SOAX measures usage in credits. 1 credit = $1", hedged in the pricing FAQ as "one credit usually represents about $1 of usage. This is a pricing reference, not a cash value." Now line the fees up against the included credits: Builder $200 includes 200, Team $500 includes 500, Scale $1,500 includes 1,500, Enterprise $3,000 includes 3,000. On every paid plan the included credits equal the fee in dollars.

So the monthly fee is not a platform fee sitting on top of usage. It is the usage, prepaid. SOAX's account-billing docs confirm the arithmetic in gigabytes, listing Builder at "$3.00/GB ~66.7 GB", Team at "$2.20/GB ~227 GB" and Scale at "$1.50/GB ~1.0 TB". The plan tier's only real job is to lower your per-GB rate, which collapses the decision to two numbers: how many gigabytes, through which tier?

The Builder, Team and Scale Tier 1 figures are SOAX's own; the rest is division. The Enterprise row holds only at the entry band, because Enterprise Monthly runs three volume bands per tier, reset each billing period. SOAX's worked example charges Tier 1 "the first 5 TB at 0.85 credits/GB, the next 10 TB at 0.75 credits/GB, and the remaining 5 TB at 0.50 credits/GB", and Tier 3 "the first 5 TB at 0.35 credits/GB, the next 10 TB at 0.30 credits/GB, and the remaining 1 TB at 0.25 credits/GB", with a lower band that "does not change the rate charged for earlier traffic". That example gives bands only for Tiers 1 and 3, though the pricing page's own plan data carries Tier 2 at 0.55, 0.45 and 0.35 credits/GB on the same three-band structure, so the schedule is published even where the prose skips it. Enterprise Annual rates are agreement-set: "Your agreement lists the rates." Annual is also the only plan you can't buy self-service, on a 12-month commitment. One wrinkle if you're modelling Enterprise: the pricing page says from $0.85, $0.55 and $0.35 per GB across the tiers while the account-billing docs say "from $0.50/GB", "$0.35/GB" and "$0.25/GB". The band schedule reconciles them, since $0.85 is the Tier 1 entry band and $0.50 the third. Quote the pricing page and ask which figure your agreement uses.

Now put SparkProxy next to that. There is no per-GB rate to compare, because the plans are unmetered and priced by parallelism instead: 100 concurrent threads on Starter at $75, 250 on Core at $140, 500 on Boost at $240, 1,000 on Plus at $440, every tier with unlimited bandwidth on a 1 Gbps port. So the comparison is metered against unmetered, not cheap against expensive. $200 on Builder buys about 67 GB of US traffic and the 68th gigabyte costs another $3.00. A $75 pack has no gigabyte number to exhaust, and none to forecast either. What that is and isn't worth is argued both ways in are unlimited bandwidth proxies worth it.

Two of SOAX's published terms decide whether metering is a price or a risk for you, and both are easy to miss on a pricing page.

Credits expire. "Credits added on Sandbox or a monthly plan expire 60 days after they are added. Credits added on an annual plan expire 12 months after they are added." Each block keeps its own fixed expiry, and the earliest-expiring block is spent first. The terms of use are unambiguous about what happens next. Section 6.3.2: "Upon expiry of the applicable Validity Period, any unused Credits within the relevant grant are automatically forfeited." Section 9.2.1: credits "are non-refundable, expire in accordance with the Agreement, and do not constitute money, stored value, or a refundable balance, in each case subject always to mandatory applicable law." Read that closing qualifier before you lean on it. Section 9 is headed "Consumer Withdrawal Rights (Individuals / B2C)" and 9.1.1 applies it "solely to the extent required by such law", adding that otherwise "Plan Fees, Top-Up Credits payments, and Credits are non-refundable." A company buying a monthly plan should read the forfeiture as written. No rollover provision is stated, and credits cannot be withdrawn or transferred to another account.

The meter stops at zero. "Requests stop once your credit balance reaches zero", and postpaid billing is "off by default for every plan, including Enterprise". Turning it on requires an Enterprise plan plus completed business verification plus a written agreement on rates and payment terms, and SOAX is explicit that "Being on the Enterprise plan or completing identity verification doesn't enable it on its own." Plan fees stay prepaid regardless.

Stack those two and you get the shape of workload metering punishes. A bursty or seasonal crawler pays twice: once for prepaid credits it never spends inside 60 days, and once in a mid-job stoppage when a heavier-than-forecast month drains the balance at 02:00. So the decision rule isn't volume alone, it's lumpiness. Large and smooth favours metering, because the per-GB curve is genuinely good and the credits get used. Lumpy or unforecastable makes metering the risk rather than the price, which is the workload shape a flat unmetered pack exists to absorb. If you've never put a number on your own variance, how to forecast proxy bandwidth needs is the arithmetic.

Credit to SOAX for publishing all of it plainly, including the parts that cost them sales. Their position on negotiating is worth quoting rather than summarising: "Our self-service prices are public, and you can get started without navigating a sales maze. We do not inflate prices just to unveil a dramatic discount after a few calls. On Enterprise Monthly, lower rates apply automatically as your usage moves through the volume bands." Cancellation is equally clear: it stops the next renewal, the plan runs to the end of the period, the account moves to Sandbox, unused time isn't refunded, unexpired credits stay usable at Sandbox rates, and closing the account forfeits them.

Precision is either a requirement or a line item. Before you pick a plan, drop every job you run into one of these two piles.

Jobs where the fine rungs are the whole product. Local SEO geo-grid tracking, where local pack results shift across a few miles and a city centroid isn't a grid point, which is why the methodology in proxies for local SEO geo-grid tracking is built around coordinates rather than countries. Ad verification by DMA or metro, because geo-fenced campaigns, frequency caps and creative rotation are bought at metro level and a national exit verifies nothing, as covered in how brands use proxies for ad verification. Carrier testing, where APN-gated config, zero-rating and on-net redirects only reproduce from the carrier's own egress. Postcode-level retail work, where price, stock and delivery promise are all functions of the shopper's ZIP. ASN pinning for routing, peering or CDN-edge verification. For any of these, a country-level exit isn't a cheaper substitute. It's a wrong answer that returns a 200.

Jobs that pay for precision and never use it. National catalogue crawls. SERP collection for non-local queries. Sitemap, feed and filings harvesting. Price monitoring where the price is national. Uptime and synthetic monitoring. Localisation QA at language or country level. Training-data collection. Each of these needs at most one country parameter, and several need country-any.

Then do the division, because this is where the $1,500 door gets expensive quietly. If postcode-level work is 6 GB of a 446 GB month, Scale is being bought for 1.3% of your traffic. Three honest responses: accept it, re-test whether city- is actually sufficient (it's free from Sandbox upward, and for a lot of DMA-level work it is), or split the workload so the 420 GB of country-level bulk runs unmetered while a small metered account covers the 26 GB that needs a finer rung. The third is what most mature teams land on. It isn't disloyalty to either vendor, just a refusal to price bulk traffic at the rate your hardest 1% requires.

A comparison where the vendor writing it wins every axis isn't worth reading. These are the places SOAX is the better buy, and several are capability gaps rather than preferences.

• Real mobile carrier IPs at residential parity. A named 3G, 4G and 5G product, 30M+ IPs per the docs and 33M+ per the product page title, carrier selection via isp-, no price premium over residential in the same country.
• ZIP and ASN targeting. Documented zip- and asn- parameters on Scale and Enterprise. For postcode-level price checks, local-pack verification or pinning traffic to one autonomous system, nothing we sell substitutes.
• ISP and carrier targeting on the free plan. isp- works on Sandbox, Builder and Team too, so ISP-pinned routing can be validated before any subscription.
• Country breadth, with the list published. 195+ geos, split into 32 named Tier 1 countries, roughly 60 Tier 2 and all remaining in Tier 3, each enumerable through the locations API. Against our 80+ countries that is broader, and you can check membership before you buy.
• Cheap bulk traffic through low-cost markets. Tier 3 at $0.90/GB on Team, $0.60 on Scale, 0.25 credits/GB in the Enterprise third band, with the band schedule published so finance can model a multi-terabyte crawl before signing anything.
• Per-request rotation control. rotate-requests_N up to 1,000,000, rotate-timed_N from 10 seconds to a week, bind-node for hard pinning that returns 503 BOUND_NODE_FAILED rather than silently swapping the exit, onerror-retry_N to a maximum of 10, onerror-fail, and prefer-lookalike to bias replacement toward the same region and ISP. Work that needs rotation counted in requests, or an exit that must never change mid-flow without telling you, is better served by those rules than by a fixed per-request or 5-minute rotation.
• Multi-team governance. Up to 10 proxy packages and 10 seats, per-package RPS, connection and traffic caps, Owner and Member role separation, usage tracked per package. An agency capping and attributing proxy spend per client gets that structure instead of building it.
• Programmatic provisioning. A REST management API with an OpenAPI spec covering packages, connection-string generation, location enumeration down to ISP, ASN and ZIP, credit blocks and usage analytics, with scoped API keys. Limits are 5 requests per second per key and 500 per minute per IP, separate from proxy traffic limits.
• Procurement paperwork. A published DPA, KYC and KYB policy, CSR policy, privacy policy, and a contractual SLA with named service credits: 10, 25 and 50 percent of monthly plan fees on Standard, 15, 35 and 75 percent on Premium, capped at 50 and 100 percent. Premium commits 99.9% monthly availability and the pricing table assigns it to Enterprise; Standard is a reasonable-efforts tier with no percentage target, and Sandbox organisations accrue no service credits. SOAX Ltd is a UK entity in Uxbridge, London.
• UDP over SOCKS5. Niche, but documented: "Residential and Mobile proxies support UDP. UDP works over SOCKS5 only." Our published protocol list is HTTP, HTTPS and SOCKS5, with no UDP.
• Open-ended evaluation. Sandbox has no subscription fee, no time limit, full residential and mobile access, and country, region, city and ISP targeting for a $25 minimum top-up. For a long, intermittent proof of concept, a free account with no clock beats any fixed-length trial window, and the Sandbox rates are published so the test cost is known up front.

On sourcing and compliance, use SOAX's own words rather than anyone's characterisation. Their trust page calls the network "ethically sourced", "built on transparency and consent, with each connection sourced from users who willingly share their bandwidth", and the residential page says the pool is "built on real IP addresses voluntarily shared by individuals". They state that "SOAX and our partners comply with all relevant data protection regulations, including the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA)", that they monitor the network "24/7 to identify and mitigate any issues, attacks, or abuse", and that they allow "only pre-approved, ethical use cases". On certifications they're candid: "We're in the process of acquiring the SOC 2 and ISO 27001 certifications", which is to say not certified as of this reading. Personal KYC runs through Sumsub in about five minutes, and SOAX states it never sees the documents, "only the final verification status".

Stop comparing vendors and classify your own jobs. Twenty minutes with your target list beats a week of reading pricing pages.

Step one: group your jobs by the finest rung each genuinely needs, and weight by traffic.

# precision-audit.py  Which rung does your traffic actually need, and how much of it?
from collections import Counter

RUNGS = ["country", "region", "city", "isp", "asn", "zip"]
JOBS = [
    # (job, finest rung it genuinely needs, monthly GB, geo tier)
    ("national catalogue crawl",       "country", 420, 1),
    ("DMA-level ad verification",      "city",     18, 1),
    ("local pack geo-grid, 40 ZIPs",   "zip",       6, 1),
    ("carrier config check, 3 MNOs",   "isp",       2, 1),
]

gb = Counter()
for _, rung, vol, _tier in JOBS:
    gb[rung] += vol

finest = max(gb, key=RUNGS.index)
print(f"finest rung needed: {finest}")
print(f"GB that need it:    {gb[finest]} of {sum(gb.values())}")
print(f"share:              {gb[finest] / sum(gb.values()):.1%}")

On that example list the answer is zip, on 1.3% of the bytes. That single percentage is the whole purchase decision, and almost nobody computes it before shopping.

Step two: price the rung you need, not the rung you'd like. This reproduces the tables above from SOAX's published rates, so you can swap in your own volume and tier.

# soax-floor.py  What do the last two rungs of the ladder cost at your volume?
# Rates read from soax.com/pricing on 2026-10-09. USD, ex-VAT.
# Model: the plan fee is prepaid credits at 1 credit = $1, topped up at the plan rate.
BASIC = {"country", "region", "city", "isp"}
FINE = {"asn", "zip"}
PLANS = [
    # name,      fee,  (tier1, tier2, tier3),  rungs exposed
    ("Sandbox",     0, (5.00, 4.00, 2.00), BASIC),
    ("Builder",   200, (3.00, 2.25, 1.20), BASIC),
    ("Team",      500, (2.20, 1.65, 0.90), BASIC),
    ("Scale",    1500, (1.50, 1.15, 0.60), BASIC | FINE),
]

def cheapest(gb, tier, plans):
    return min((max(fee, gb * rates[tier - 1]), name) for name, fee, rates, _ in plans)

coarse = [p for p in PLANS if not FINE & p[3]]
fine = [p for p in PLANS if FINE <= p[3]]

for tier in (1, 3):
    print(f"--- Tier {tier} exits ---")
    for gb in (25, 50, 100, 250, 500, 682, 1000, 1667):
        c_cost, c_name = cheapest(gb, tier, coarse)
        f_cost, f_name = cheapest(gb, tier, fine)
        print(f"{gb:>5} GB   up to ISP: {c_name:<8} ${c_cost:>7,.0f}"
              f"   with ASN+ZIP: {f_name:<6} ${f_cost:>7,.0f}"
              f"   delta ${f_cost - c_cost:>+8,.0f}")

Run it unchanged and the Tier 1 block reproduces the table above, while the Tier 3 block shows the door staying shut until 1,667 GB. Swap in your own volumes and your own tier mix.

Step three: check the things price pages don't cover. Four of these have bitten real projects. SOAX restricts some destinations by default: "Some domains (financial services, government, certain news sites) and ports (including SMTP 25, 465, 587) are restricted by default", with access beyond the defaults limited to Enterprise customers holding a signed Order Form and completed verification, and every request going through use-case review. Blocked requests return 403 with X-SOAX-Error: DESTINATION_BLOCKED. Sessions expire after 60 seconds of inactivity, session IDs are letters, digits and underscores to a maximum of 32 characters, and rules lock on a session ID's first request, so reusing an ID with different rules returns 409 SESSION_PARAMS_MISMATCH. Concurrency is published only as "4,000 per proxy gateway on the Sandbox plan", with concurrent sessions and RPS both listed as package-dependent and a customer-level ceiling across all packages, so a concurrency commitment on a paid plan is something you ask for in writing; the marketing pages claiming "Unlimited concurrent sessions" aren't the document to rely on. And traffic counts in both directions, "including partial responses, retries, and interrupted connections", with one GB defined as one billion bytes.

Step four: measure, don't read. Run 1,000 real URLs, including your hardest page type and your true country mix, through a SOAX Sandbox account at your finest required rung and a SparkProxy trial at country level. Same code, same headers, only the endpoint changing. Record successful extractions, total bytes and median latency. Divide bytes by usable records to get your real consumption unit, scale it to a month, then price that month on the ladder above. Compare success rates last, using your own numbers rather than the published 99.95% and 0.55s, or our 99%+ and 99.9% uptime. Nobody audits those figures, ours included.

Step five: decide the split, not the winner. If more than a few percent of your bytes need ASN or ZIP, buy Scale and stop optimising. If it's 1%, keep the fine-rung work metered and move the bulk to something unmetered. If nothing on your list needs better than country, the ladder is irrelevant to you and this is a volume question instead. Broader options for that case sit in SOAX alternatives, and if you're weighing SOAX against another consumer-IP network rather than against us, SOAX vs Oxylabs runs that pairing.