proxynomad.com / use-casesfive jobs · three questions
What people route through us.
Five jobs, described the way they actually run: which proxy type, which session pattern, and the mistakes that show up as invoices. If your job is not here, it is probably still one of these wearing a different name.
Question 1 of 3
Does your target block or fingerprint datacenter IP ranges?
Question 2 of 3
Do you need to hold one stable identity, like a login, for days?
Question 3 of 3
Is this high volume, or precision work against the hardest defenses?
A starting point, not a verdict. When the cheaper type holds your success rate, it is the right one, and we will say so.
Collection that survives contact with the target.
The pattern that works
Tier the retry. Every URL hits the datacenter pool first, and only responses that come back blocked retry through residential. Most targets never trigger the second tier, so the blended rate lands near $0.60/GB while the success rate looks residential. One credential set covers both gateways: the tier switch is a port number in your retry logic.
The mistake that costs
Concurrency without pacing. A thousand workers through one city pool reads as an event, and targets respond to events. Ramp up, spread targeting as wide as the job allows, and treat a rising block rate as a signal to slow down before it becomes a reason to move up a price tier.
The price a real shopper sees.
The pattern that works
Arrive as the local customer. A residential exit in the right city gets what the local customer gets, which is the entire dataset. Use short sticky sessions per store visit: land, load the product, read the price as one identity, rotate. Carts and checkouts are where dynamic pricing concentrates: if your pipeline needs basket-level prices, hold the session for the whole basket and never reuse it across stores. And collect at retail-shaped hours in the target's timezone.
The mistake that costs
Reading the generic price as the market. Retailers price by geography, and increasingly by who seems to be asking: a datacenter exit gets the generic price or the block page. National averages hide exactly the spread you are being paid to find, and a snapshot taken when no human shops is a snapshot of the wrong regime.
one query per exit
Rankings from where the searcher stands.
The pattern that works
Pin the exit to the searcher. Residential exits pinned per city give you the results page your client's customer sees in Turin or Denver. Rotate per query: search engines key personalization and rate logic to session continuity you do not want.
The mistake that costs
Buying more pool than the job needs. Rank tracking is only as honest as its exit: a datacenter exit gets the datacenter-flavored page served to obvious automation, not what the searcher sees. This is the rare job where the good proxy type is also the cheap line on the invoice, so spend the savings on query depth, not on a bigger pool.
per check
residential · city + asn
See the campaign the audience sees.
The pattern that works
Arrive as the audience. Verification means arriving as the buyer's segment arrives: residential exits in the campaign's cities, and ASN targeting earns its keep here. Fresh identity per check, always. Screenshot and log what actually rendered, placement, creative, landing URL: the disputes this work exists to settle are settled by evidence, not by a dashboard export.
The mistake that costs
Auditing yourself. You cannot audit a geo-targeted campaign from an office IP: the ad server has already decided you are not the audience. And sticky sessions accumulate frequency caps and retargeting state, so by the third impression you are measuring your own contamination.
low and steady
Firmographics at a sustainable pace.
The pattern that works
Low and steady, tiered. Enrichment traffic is many small reads across many public sources: registries, directories, company sites, news. The economics favor the tiered pattern, datacenter for the long tail that does not resist, residential for the handful of sources that do, and the volumes are small enough that proxy spend stays a rounding error next to the value of a clean record. Cache aggressively so you never fetch a page twice, and keep provenance on every field: enriched data gets challenged, and the record that knows where it came from is the one that survives review.
The mistake that costs
Flooding a source that would have tolerated you for years. What matters here is rate; the IP class is secondary. Public sources tolerate polite, spread-out collection for years and shut down floods in a week. Pace per source.
What may be collected at all is covered in our acceptable use policy; it draws real lines and we hold customers to them.
Your job is not exotic. Its tuning is.
Send the target and the shape of the job; we will suggest the setup we would run ourselves, at the price we would pay.