Search the guides

Filtering & analytics

Why a VPN label should not close the door

VPNs, shared networks and hosting addresses describe connections, not buying intent. Use network labels without excluding an entire class of customers.

A network label is not a description of a person

A customer cannot open your page from a work laptop because the filter labels the connection as a VPN. The person may simply be using their employer’s network. Before blocking that category, identify the harmful action you want to stop. The route a request takes does not tell you whether its owner wants to buy.

A VPN sends a connection through another endpoint. A proxy is an intermediary between the visitor and the destination. A hosting label usually describes an address associated with server infrastructure. These labels can overlap, and the database can be wrong. Treat them as clues to check, not a ready-made customer verdict.

An ASN is a number identifying a network used for routing internet traffic. It helps group addresses by network operator; it is not a person’s identifier. Blocking a large operator because one address misbehaved can exclude unrelated customers and services.

Many legitimate people share an exit

Several people can appear to your site under one address. This happens with office gateways, shared mobile-network exits and some privacy services. A rule that treats an address as one person can mistake a group’s ordinary activity for one person making excessive requests.

Apple’s guidance explains that Private Relay replaces the original address and may share exit addresses among users. It recommends adjusting IP-based checks so legitimate users are not harmed. That does not mean every privacy service behaves the same way. It does show why a privacy label alone is insufficient.

Event flowAn address describes a route
  1. PeopleSeveral unrelated visitors
  2. Shared exitOffice, carrier or privacy service
  3. Your siteOne visible network address
  4. AssessmentBehaviour and context still matter

One visible address can represent several independent people.

Imagine twenty colleagues opening a webinar invitation after a company announcement. That is a hypothetical example of a burst from a shared exit. Compare their separate normal registrations with repeated identical invalid submissions. The behaviour and its consequences matter more than the fact that both groups share an address.

Choose the action according to the risk

Protect the action that causes the harm. Reading a public explanation, sending an enquiry and making a payment do not carry the same risk. Where possible, keep the explanation accessible and apply the necessary control to the affected action.

Keep service eligibility separate from bot detection. A visitor outside the delivery area may be a genuine customer whom you cannot serve. Explain that limitation and record it as a location decision. Calling the visitor a bot would make both the message and the reporting misleading.

  • Use network labels as one signal among several.
  • Prefer endpoint-specific limits over a sitewide network ban.
  • Keep useful verified automation and necessary service callbacks working.
  • Record the reason, duration and owner of each broad exception.
  • Review accepted outcomes and support complaints after changes.
Decision pathWhat should happen to this request?
A network label appears

Look at the action and its context.

  • Only an unfamiliar route

    Do not infer abuse from the label.

  • Outside the service area

    Explain the eligibility limitation.

  • Repeated harmful actions

    Protect the affected endpoint.

Choose the action from the observed problem, not the network label alone.

Test the users the rule might exclude

Test the routes your audience commonly uses: home broadband, mobile data, office connections and relevant privacy settings. Repeat the same useful action on each. You are looking for a reproducible difference, not trying to sample every network in the world.

Check the refusal screen too. Give a legitimate visitor a clear next step and a non-sensitive event reference where available. Support should be able to find the request without asking for a password or complete browsing history. Disabling privacy protection should not be the automatic answer to every case.

Give exceptions a lifecycle

When you allow an exception, write down its purpose, scope, owner and review date. A short-term fix for one verified service should not become a permanent exemption for an entire network. Remove it when the underlying problem has been corrected.

Verify useful automated services using their documented methods. A request’s claim to be a monitor or trusted crawler is not proof. Match the strength of verification to the action: access to a public guide and access to an administrative function need different treatment.

Consider a hypothetical monitor that stops working after a network rule changes. Allowing all hosting traffic might restore it, but that is broader than the established need. Identify the service and required endpoint, then narrow the lasting exception. Record emergency recovery separately from the final rule.

Revisit old restrictions when the audience changes. A rule made for a small consumer campaign may later affect business buyers on office gateways. Keeping its original reason makes it easier to see when the assumption no longer fits.

Measure the cost of exclusion

Compare accepted enquiries and completed purchases in the affected group before and after a change. Account for changes in the campaign mix and confirmation delays. A few successful tests show that a path works; they do not establish the long-term effect on conversion.

When a buyer is wrongly restricted, locate the failed assumption. Was the network label incorrect, the rule too broad or the counter treating a shared address as one person? Update the relevant part instead of adding a vague permanent exception.

Use network information to explain a request, not to judge the person behind it. Restrict observed abuse, keep a usable recovery route and retain enough evidence to revisit mistakes. The useful result is a safer action that legitimate customers can still complete.

Questions answered

Should all hosting addresses be blocked?

Not automatically. Service callbacks, monitoring and some privacy routes can use server infrastructure. Decide which endpoint and behaviour need protection.

Does a VPN mean the visitor’s location is fake?

The visible exit can differ from the original connection, but the reason may be ordinary work or privacy. Evaluate actual service eligibility separately from automation.

ADSTRAFFIC / ENAll guides →