Questions to Ask EASM Vendors (With Model Answers)

Questions to Ask EASM Vendors (With Model Answers)

A vendor call goes well when you can tell, within about ten minutes, whether the person on the other end is describing a method or a dashboard. The questions to ask EASM vendors are not hard to find; most buyer guides hand you a list and leave. What they do not give you is a way to grade the answers, and that is where shortlists get decided by slide quality rather than substance. So each question below comes with what a credible reply actually contains, the evasions that should worry you, and the way to put the claim to the test on your own domain before you sign anything.

Write down your objective before you write any questions

Evaluations go wrong early, usually because nobody wrote down what the tool is for. Three objectives are common, and they pull in different directions.

  • Find the unknown estate. You suspect shadow infrastructure: campaign microsites, an old agency’s hosting, a subsidiary nobody has audited. Here, discovery breadth and attribution accuracy matter more than anything else.
  • Keep a known estate monitored. Your register is broadly right and you want to know within hours when something changes. Now refresh cadence and change alerting carry the weight.
  • Assess an acquisition before completion. You have a company name, a few domains and no cooperation from the target’s IT team. Speed, evidence quality and the ability to explain findings to non-technical readers matter most.

Then decide who acts on findings. A two-person IT team drowning in 300 medium-severity items will stop opening the emails by week three; a 20-strong SOC with a ticketing pipeline can absorb volume and will care more about API fidelity than about a tidy dashboard. Say so in the call.

Send each vendor a one-paragraph problem statement in advance. Something as plain as: “We are a 400-person organisation, we believe we run around 60 external hosts, we have made two acquisitions since 2022, and we want to know what resolves under our brand that we did not build.”That produces a demo aimed at your problem. A generic feature matrix produces a demo aimed at the matrix.

Keep a single sheet per vendor with your objective at the top and a row for each question. Record the answer, not a score out of ten. You are looking for named sources, numbers and honest caveats, and those do not compress into a star rating without losing the thing you were testing for.

Discovery and attribution: can you find us from one seed domain?

The question: start from one registered domain, with no asset list supplied by us, and show the estate live in the call.

Grade the answer by the sources named, not by the adjectives used. A strong reply walks through certificate transparency logs, passive DNS and authoritative DNS records, domain registration and WHOIS data, and IP block or ASN ownership, and explains how those feed each other: a certificate issued for a hostname nobody documented leads to a resolution attempt, which leads to an IP, which leads to a netblock, which suggests more hostnames. Each source finds different things. Certificate logs catch names that were never published anywhere else; ASN ownership catches hosts whose names you will never guess. A vendor who cannot tell you which source produced a given asset is showing you a dashboard, not a method. We have set out the mechanics in more detail in our piece on how EASM discovers the assets you forgot.

Then listen for confidence language. Honest discovery returns candidates, not certainties, because a hostname resolving under your brand may belong to a supplier, a marketing agency or a project that died in 2019. Ask three follow-ups: what evidence is displayed per asset, how an asset is promoted from candidate to confirmed, and how you reject a wrong attribution so that it does not quietly reappear the following week. The last one catches more products than you would expect.

Red flags: “proprietary intelligence”with no sources named, a single confidence-free list, and no mechanism to dispute or remove an asset. If every asset is simply present, with no provenance and no grading, the vendor has made an inference and presented it as a fact.

Data freshness: how fast do you see change, source by source?

“Continuous”is a marketing word until someone attaches an interval to it. Ask for cadence per source. A certificate entry can appear in public logs within minutes of issuance. A port and service sweep across a few hundred hosts may run weekly, fortnightly or monthly depending on how the vendor manages its own infrastructure costs. A product can so be genuinely continuous for hostnames and badly stale for services, and both statements can appear on the same page of the brochure.

The follow-up that separates the field: if we publish a new subdomain this afternoon, when does it appear in your inventory, and when does its service detail get filled in? Two different numbers. Make them say both.

There is a concrete yardstick here. Cyber Essentials, the UK scheme overseen by NCSC and delivered by IASME, expects high and critical severity patches to be applied within 14 days of release. A product whose service detail refreshes every 30 days cannot support that deadline on its own, however often it refreshes the hostname list. Say that out loud in the call and watch the response.

Probe change detection item by item: new host, new open port, new certificate issued, a changed service banner, an asset that has gone quiet. Ask which of those can trigger a notification, and to whom. Then check timestamps in the demo tenant. If every “last seen”value in a supposedly continuous product falls on the same Sunday night, you have learned something the roadmap slide was not going to tell you. Our breakdown of how often external assets should be scanned by tier is useful preparation for this conversation.

Prioritisation and noise: how do you decide what we look at first?

Ask how a finding is ranked. Severity score alone is a weak answer, because CVSS describes the flaw, not your exposure. A better answer layers context: is the service reachable from the open internet, does it sit behind authentication, is there a WAF in front of it, is the host a production system or a parked redirect.

Deduplication and recurrence questions reveal real engineering. One weak TLS configuration on 40 hosts: is that one finding with 40 affected assets, or 40 tickets. A finding you fixed in March that reappears in May because somebody restored a template: does the product know it is the same issue returning. Can you suppress a finding, and does the suppression expire, and does anyone get told when it does.

Ask for a typical first-month finding count for an estate your size, and what proportion a customer usually actions. Any vendor with a few hundred customers can answer that approximately. Then ask the harder version: after the initial clear-out, what does month four look like, and how is alert fatigue managed when the exciting findings have gone.

Red flags: an unexplained risk score from 0 to 100 with no stated inputs, and a demo tenant whose results are suspiciously tidy. Real estates are messy. A demo that is not messy is a demo of something other than a real estate.

Coverage beyond CVEs: misconfigurations, certificates and dangling DNS

Ask what the product flags that has no CVE attached. Exposed admin interfaces and login panels on hosts that should not have them. Directory listings. Open storage buckets. Weak or deprecated TLS configuration. Certificates expiring in the next fortnight, and certificates that already expired and nobody noticed. None of these is exotic. They are the ordinary sediment of an organisation that has moved quickly, and they are frequently more dangerous than a mid-rated CVE on a hardened host.

Then run the dangling DNS test question, which is cheap and unusually revealing. A CNAME left pointing at a deleted cloud storage bucket or a retired platform tenant: the record still resolves, the service behind it is gone, and anyone can register that name at the provider and serve their own content from your subdomain until the record is removed. Ask how the platform detects that condition, whether it explains in plain English that an attacker can claim the free name, and whether the recommended fix is to remove or repoint the record. Products with real external exposure logic answer immediately. Asset listers describe the hostname and stop. We cover the mechanics in subdomain takeover and the dangling DNS problem.

Finally, ask how the tool handles assets you do not own. Supplier subdomains, agency-hosted microsites and SaaS tenants resolving under your name are part of your attack surface in the sense that your customers trust them, and not part of it in the sense that you cannot patch them. Ask how ownership is presented, and whether third-party assets can be tagged and reported separately.

Active testing, safety and scope: what does the platform actually touch?

Ask which checks are passive inference and which involve active interaction with your hosts, in the vendor’s own words. Passive fingerprinting from a banner tells you a host is likely running a vulnerable version. It has not tested the flaw. Only an active check, or an authenticated scan, moves “likely”to “confirmed”. Vendors that blur the two inflate their finding counts and your remediation effort at the same time. The distinction is worth understanding before the call, and we have written it up as likely versus definitely.

Then the authorisation questions, which your change board will ask you anyway: what is sent to our hosts, at what request rate, from which source addresses, and can we exclude fragile systems from active interaction entirely.

Two myths are worth clearing in the same conversation. Credible EASM does not exploit systems; it observes and infers, and a vendor offering to prove exploitability on your production estate is offering you a penetration test under another name. And external observation does not make you more of a target. Certificate transparency logs, DNS and registration data are public by design, and an attacker enumerating your estate reads exactly the same sources, whether or not you have bought a product. The question is only whether you see what they see first.

A good vendor will tell you where the product stops without being asked: internal estate, authenticated depth, business logic flaws, anything requiring credentials. If they claim the platform replaces your vulnerability scanner, that is a judgement about their honesty, not about their coverage.

Workflow, licensing and the pilot that settles it

Integration questions are easy to skip and expensive to skip. API access and rate limits, export formats, ticketing and SIEM hooks, per-asset owner assignment, and whether notifications can be routed to the team that owns the record rather than to one shared inbox. If a finding cannot reach the person who can fix it without someone retyping it, the tool has added a step.

Then the commercial questions buyers regret not asking. What counts as a billable asset: a hostname, a resolved IP, a service. What happens when discovery does its job and finds several times your known estate. A mid-sized organisation that believes it runs 60 external hosts and sees 200 plus after discovery has a licensing conversation to have, and it is far better had before signature. Ask whether assets you mark as third-party or out of scope still count, how renewal pricing handles growth, and what you can export, in what format, if you leave.

Settle it with a pilot. Give every shortlisted vendor the same single registered domain, no asset list, and compare after two weeks: total assets found, overlap between vendors, unique finds, and anything attributed to you that actually belongs to a supplier or a former agency. The unique finds and the false attributions are the two columns that matter. Agree your success criteria and a decision date before the pilot starts, because an open-ended trial becomes a procurement process with no end. Our practical checklist for choosing an EASM product is a sensible read before the final call.

The Short Version

  • Write your objective down first. Finding shadow estate, monitoring a known estate and assessing an acquisition need different strengths.
  • Grade discovery by named sources: certificate transparency, passive and authoritative DNS, registration data, ASN and IP ownership.
  • Demand confidence language and a working way to reject a wrong attribution permanently.
  • Get refresh cadence per source, not one word. Hostnames can be minutes old while service detail is weeks old.
  • Make the vendor separate passive inference from active confirmation, and state what touches your hosts.
  • Ask the dangling CNAME question. It sorts exposure logic from asset listing in under a minute.
  • Check licensing against discovery success before signature, then run one seed domain past every vendor and compare.

Frequently Asked Questions

What are the most important questions to ask EASM vendors first?

Start with discovery from a single seed domain and the sources behind it, then refresh cadence per source, then how passive inference is distinguished from active confirmation. Those three expose the method. Workflow, prioritisation and licensing questions matter, but they are easier to fix later than a weak discovery engine.

How can I tell if a vendor’s discovery claims are real?

Ask which source produced a specific asset in the demo and what evidence is attached to it. A real answer names certificate transparency, DNS, registration data or netblock ownership and shows the record. Vague references to proprietary intelligence, with no provenance per asset and no confidence levels, usually mean the detail is not there to show.

Should I ask for a pilot, and how long should it run?

Yes, and around two weeks on your own seed domain is normally enough to compare asset counts, overlap, unique finds and false attributions across a shortlist. Agree success criteria and a decision date before it begins. Longer pilots rarely produce better decisions; they mostly produce drift.

Does an EASM platform scan or exploit our systems?

Credible external attack surface management observes and infers from public data and light interaction, and does not exploit systems. Some products do perform active checks against your hosts, which is why you should ask what is sent, at what rate and from which addresses, and whether fragile systems can be excluded. Exploitation belongs in a scoped penetration test.

How many findings should we expect in the first month?

It depends on estate size and how long the estate has gone unmonitored, so ask each vendor for a typical figure for an organisation of your size rather than accepting a general claim. Expect the first month to be unusually noisy as historic issues surface, and ask specifically what month four looks like once the backlog has been cleared.