Surfe

A data plumbing product that makes few claims about models and does not need to. The core mechanism is sequential querying: a profile identifier goes out to more than fifteen contact data providers in turn until one returns a verified address or mobile number, and the result is written into the customer's record system with field mapping applied. That is a waterfall, which is orchestration and deduplication rather than inference, and it predates the current vocabulary by years.

Search, validation, bidirectional synchronisation and the browser overlay are all deterministic. Two capabilities do involve inference and both sit at the upper tiers: account lookalike generation, which infers similarity from firmographic patterns, and market signal playbooks. The vendor's own answer to a question about return on investment mentions prioritisation of prospects, which is the one place model language appears.

Stripping all of it leaves the product that fifty thousand users are described as running on. The protocol server for artificial intelligence assistants is a distribution route for other people's models rather than a model of the vendor's own. Ask what powers account lookalikes and prioritisation, and whether any ranking is applied to search results.

Little runs unattended, and the one thing that does write to the record system automatically is ungoverned. Most actions require a representative to be looking at a profile and to click, which independent reviewers describe as a limitation for scaling teams and which is also a real constraint on unattended behaviour: there is no background bulk enrichment, so nothing happens to a customer's data while nobody is watching. Credit allowances cap consumption per user by construction.

The exception is bidirectional synchronisation, which is published as a feature of the paid tiers and by definition writes to the record system without a person confirming each field. Nothing published describes what happens on conflict, whether enriched values overwrite existing ones, whether a mapping can be reviewed before it commits, or how a wrong value is unwound once it has propagated into a record system and from there into a sequencing tool.

Access logging appears only at the quoted tier, so on the self serve plans there is no published way to see what was written or by whom. Ask what governs conflict resolution in bidirectional synchronisation, whether enriched values overwrite existing record data, and what audit trail exists below the quoted tier.

Few model claims are made, so there is little to disclose, and the small amount that exists is undescribed. Account lookalike generation and market signal playbooks both appear as upper tier features and neither is explained: nothing states what similarity is computed from, what a signal is derived from, or how either is validated. Prospect prioritisation is mentioned in the vendor's own answer about return on investment without further detail.

No model provider, family or version appears anywhere, no model card exists, and no accuracy or evaluation figure is published for any inferred output. The more consequential transparency question for this product is not about models at all but about the waterfall, and it is equally unanswered. The vendor states that more than fifteen providers are queried in sequence and names none of them on its own surface, though independent accounts list several.

Which provider supplied a given contact detail determines its likely accuracy, its licensing basis and where a data subject would go to correct it, and none of that reaches the buyer. Ask which providers sit in the waterfall, in what order, and whether the source of each returned field is disclosed.

Named customers quoted by name and title, and no number for the one thing this product is judged on. The customer evidence is better attributed than most: case studies identify the company, the individual and their role, with quotations describing what the deployment actually did, and the logo wall carries recognisable names across software, logistics, property and industrial manufacturing.

Scale is stated at more than fifty thousand users and over a million enrichments a month, and review platform standing is cited with a count. What is missing is the find rate. For an enrichment vendor the outcome that matters is what proportion of lookups return a usable verified address or mobile number, and no figure appears anywhere.

The vendor's own frequently asked questions pose the accuracy question directly and answer it with method rather than measurement, describing multi source verification across fifteen providers and real time rather than periodic refresh, which explains why the answer should be good without saying what it is. That gap has commercial consequences a buyer can compute: credits are consumed by searches, and if half of them return nothing the effective cost per usable contact doubles. Ask what percentage of email and mobile lookups return a verified result, measured across what population, and whether failed lookups consume credits.

A published route for the people in the data to get themselves out of it, which is rare in this category. The vendor maintains a dedicated opt out mechanism as its own page in the site footer, alongside a data protection page and support for data subject access requests stated on its main site, and describes granular opt out as a capability rather than a policy aspiration.

That matters because the individuals whose work addresses and mobile numbers this product supplies are not customers, have no relationship with the vendor, and in most competing products have no visible way to reach it. Compliance with the European data protection regulation and the California statute is published as a row in the plan comparison, applying at every tier including the free one rather than being sold as an enterprise feature.

The product itself sends nothing: outbound runs through the customer's own connected sequencing tools, so campaign obligations, suppression and consent records sit there. Held below the top band because no conformance position is stated with the professional network the extension operates on, and because nothing describes how an opt out propagates to the fifteen upstream providers whose data the waterfall returns, which is where a suppressed contact could reappear. Ask how an opt out propagates upstream and how quickly it takes effect.

A genuinely thorough apparatus for a company of this size, published rather than promised. The site footer carries a privacy policy, a cookie policy, a legal notice, a dedicated data protection page and separate terms governing the programmatic interface, alongside a trust centre at its own path.

The security page states compliance with the European data protection regulation in both capacities, as processor and as controller, which is a meaningful distinction most vendors leave vague and which matters here because supplying third party contact data makes this company a controller of personal data about people who never contacted it. A certified information security management standard is held.

Support for data subject access requests is stated, and European data residency is offered on request, which is a real election even though it is arranged rather than published as a configurable option. Two things hold it below the top band. The residency offer has no published terms, so a buyer cannot establish what it covers or costs before asking.

And the processing instruments themselves were not reached on this pass, so no subprocessor list, retention period or transfer mechanism could be confirmed, and the trust centre that would carry them was not retrieved. Ask what European residency covers and on what terms, and for the subprocessor list and retention periods.

The vendor publishes the document that establishes its legal basis for holding personal data, which almost nothing else in this category does. A legitimate interests assessment appears as its own page in the site footer. That is the balancing exercise European data protection law requires before an organisation may process personal data without consent, weighing its commercial interest against the rights of the people in the data, and publishing it rather than keeping it as an internal file is a substantive act: it exposes the reasoning to challenge.

Its contents were not read on this pass and its existence is the finding. Around it sit a stated position of lawful sourcing, a claim that only permitted fields for permitted contacts are supplied, and controller as well as processor status, which together describe a company that has thought about where its data comes from rather than only about what it costs.

The architecture supports the same reading: there is no proprietary scraped database being resold, and every lookup queries licensed providers live. Held below the top band because the providers in that waterfall are not named on the vendor's own surface, so a buyer cannot assess the licensing behind any individual result, and because no indemnification position was located. Ask which providers supply the waterfall and under what licences, and for the indemnity position on supplied contact data.

A browser extension whose entire purpose is reading a professional network, and a pricing page that never names it. Throughout four plan descriptions and a comparison table the capability appears only as business network integration, and the network in question is not identified anywhere on that page even though it is the product's whole premise and is named freely in independent coverage.

A vendor choosing a circumlocution for the platform it depends on is making a statement about that relationship, and it is the clearest signal available because no conformance position, terms reference or continuity statement appears in either direction. Two facts sit on the other side and both are real.

The architecture limits harvesting: enrichment fires on a profile a representative is actively viewing rather than in bulk, which independent reviewers describe as a scaling limitation and which also means the extension cannot mass extract. And a competitor selling directly against this product states that its sales navigator support and message synchronisation run through the user's own connected account rather than by scraping, which is a concession against interest and worth weighing.

Contact details come from licensed providers rather than from the network. Ask what the extension reads from profile pages, what method the message synchronisation uses, and what the stated conformance position is.

A contact data supplier has opened a route for third party artificial intelligence assistants to query personal data directly, and published no governance for it. The vendor operates a protocol server with its own documentation, promoted as letting widely used assistants plug into its enrichment waterfall, and names access for artificial intelligence agents as a use case in its own right alongside sales and marketing.

What can be retrieved through that route is a named individual's verified work address and personal mobile number. Nothing published states what limits agent access, whether queries through that route are rate limited or logged differently from human ones, what the assistant provider may retain of the returned personal data, whether the customer's opt out and suppression obligations follow the data into the assistant, or whether anything distinguishes a query made by a person from one made autonomously.

Access logging appears only at the quoted tier, so on the self serve plans there is no published record of what an agent retrieved. Nothing was located on whether customer usage trains any model. Ask what governs access through the protocol server, what an assistant provider may retain, and whether agent queries are logged separately.

The people whose details this product supplies are never told, and the vendor has built more of a route back than most. What is delivered is a named individual's verified work address and, at every paid tier, their mobile telephone number, obtained from third party providers and written into a stranger's record system without that person's knowledge.

A personal mobile is the more intrusive of the two by some distance, since it reaches someone outside working hours and on a device they did not give to anyone in this chain. Against that sit real mitigations that most competitors do not publish: a dedicated opt out mechanism, stated support for data subject access requests, granular opt out, and a published assessment of the legal basis for holding the data at all. Those give a person a route to object, if they discover the vendor exists.

Nothing published addresses the discovery problem itself, which is that a contacted individual has no way to learn which supplier furnished their number, and nothing describes whether a suppression request propagates to the upstream providers or only stops this vendor returning the record. The product sends nothing itself, so no impersonation arises. Ask what a contacted individual is told about the source of their details, and whether suppression propagates upstream.

Public developer documentation and a first class protocol server for assistants, both reachable without a sales conversation. Named integrations cover four customer record platforms, three sequencing tools, a spreadsheet connector and the browser extension itself.

Around them sits a developer surface that is more substantial than the connector count suggests: documentation on its own subdomain covering endpoints, schemas and development kits, separate documentation for a protocol server that lets assistants query the enrichment waterfall, a public status page, a partner programme and a help centre.

Publishing the interface documentation openly while gating the interface itself lets a technical buyer evaluate before committing, which is the right order. Two things hold it below the top band. The interface is available at the quoted tier alone, so every self serve customer has documentation for something they cannot use, and the covered record platforms number only four, which independent reviewers raise repeatedly as the reason teams switch. Browser support is limited to one browser. Ask at which tier the interface becomes available and what it costs, and whether further record platforms are planned.

One published election and nothing else. The vendor states that European data residency is available on request, which is a genuine option and more than most competitors offer, and it is arranged rather than published, so a buyer cannot establish before asking what it covers, which tiers qualify, whether it costs anything or whether it extends to the upstream providers in the waterfall.

That last question is the substantial one and it is specific to this architecture: a lookup queries more than fifteen third party providers in real time, so a query about a European individual may traverse several jurisdictions during a single enrichment regardless of where this vendor stores anything. Nothing published addresses it.

Beyond that no hosting provider, region, tenancy model or recovery objective was located, and the trust centre that might carry them was not retrieved on this pass. Three office locations across two continents are published, which describes the company rather than the infrastructure. A public status page reports availability. Ask what European residency covers and whether it extends to upstream provider queries, and for the hosting regions and recovery objectives.

A certification held rather than an attestation asserted, and nothing a reviewer can verify from the pages reached. The vendor holds a certified information security management standard, described on a dedicated security page as covering people, process and infrastructure and as requiring regular risk assessments, and that is a stronger class of claim than a self commissioned report because it depends on an accredited certifying body.

A trust centre exists at its own path in the site navigation and was not retrieved on this pass, so material may well sit behind it and nothing here asserts otherwise. Product level controls are published per tier, with single sign on, directory provisioning, and security and access logging all appearing at the quoted tier alone, which means access governance on the three self serve plans is undescribed.

What could not be established: no certificate number, certifying body, audit period, scope statement, report request route, penetration testing summary or vulnerability disclosure route. One discrepancy belongs on the record. The certification badge grid displayed on the pricing page includes an audited control report alongside the standard the vendor claims in writing, while no written statement anywhere on that page mentions it and the comparison table omits it. Ask which certifications are actually held with their certifying bodies and scopes, and what the badge grid is asserting.

Four tiers published at both billing cycles with credit allowances stated in full, and three inconsistencies on the same page about what a plan actually includes. What is published is substantial: monthly and annual rates for two paid tiers, email and mobile find credits given separately for every tier in both monthly and annual units, search result limits, a free tier requiring no payment card, a full comparison table across five capability groups, and support level by tier.

The annual saving is honestly stated, with the advertised ceiling of twenty percent matching the actual maximum. The inconsistencies are on the page rather than between the page and third parties. The free tier's feature list differs between the monthly and annual views of the identical zero cost plan, listing record system synchronisation in one and omitting it from the other. The second tier's one click capture is marked as limited in one view and unqualified in the other.

And the connector that makes this product what it is, the module that writes to the customer record system, appears in the feature lists of both paid tiers while a line beneath the frequently asked questions states that it is an independent module purchasable at twenty nine dollars a month. A buyer cannot determine from the page whether their tier includes it.

Nothing states what happens when credits are exhausted or whether an annual allowance is one pool or twelve monthly caps, a question independent reviewers have answered in opposite directions. Ask whether the connector is included at the quoted tier, and what an exhausted credit allowance costs.

The architecture solves most of this problem before it arises, and the remainder is unaddressed. Because the product's purpose is writing enriched contacts into the customer's own record system, the valuable output already lives somewhere the customer controls: a departing customer keeps every contact, address and mobile number that was ever synchronised, held in their own platform under their own retention rules, with field mapping already applied.

That is genuine portability by design rather than by policy, and it distinguishes this from products that accumulate value inside themselves. What remains inside is smaller but not nothing: contact lists built in the vendor's own application, search histories, account lookalike outputs and prospect lists, none of which has a published export route, retention position or deletion timeline. The commercial terms of leaving are also unstated.

Nothing published describes what happens to an unused annual credit allowance on cancellation, whether any part is refundable, or how notice works on an annual commitment, which matters because the annual tiers are the discounted ones and the allowance is bought upfront. Ask what happens to unused credits on cancellation, whether contact lists and prospect lists export, and what notice applies to an annual term.

The product sends nothing and contributes one control that genuinely helps the customer's sending. No campaign capability, sequencer or mailbox exists here; outbound runs through the customer's own connected sending tools, so infrastructure, authentication, bounce handling and sending reputation all sit with them. The contribution worth crediting is email validation, published as unlimited on every tier including the free one.

Validation before sending is the single most effective control on bounce rate, and bounce rate is what determines whether a sending domain retains its reputation, so a vendor giving it away without metering is materially reducing a risk its customers carry. Most competitors meter validation as a credit consuming action. The exposure runs the other way and is worth naming.

This product supplies verified addresses and mobile numbers for people who never gave them to the sender, and those contacts flow directly into sequencing tools by design, so the volume of unsolicited outbound leaving a customer's domain rises with adoption while none of the consequence touches this vendor. Ask what validation checks are performed, and whether contacts are marked as enriched rather than opted in when written to the record system.

Roles addressed individually rather than as a single buyer, and a ladder that genuinely starts at zero. The vendor publishes separate pages for representatives, account executives, revenue operations and marketing teams, adds artificial intelligence agents as a segment in its own right through a data supply positioning, and names its intended users as those four groups plus recruiters and agencies.

The commercial ladder runs from a permanently free tier with real allowances through two published paid tiers to a quoted one, so an individual can start without approval and a large team has somewhere to go. Geographic evidence is concrete rather than adjectival: offices on two continents, a customer described as operating across six countries, another across more than thirty regions, and a European origin that makes the residency offer credible.

Two limits hold it below the top band and both come from the same place. Only four customer record platforms are supported, which independent reviewers raise consistently as the reason teams look elsewhere, and the extension runs in one browser only, so an organisation standardised on anything else is excluded regardless of budget. Ask which record platforms are supported today and whether other browsers are planned.