FirstTouch

The removal test is settled by the vendor's own competitive positioning rather than by inference. Every comparison it publishes against rival tools turns on where the action fires and where it is logged, being triggered from a workflow inside the system of record and attributed on the contact timeline, rather than on the quality of anything a model produces.

Strip the research and qualification agents out and that orchestration and attribution layer remains, complete and saleable, and it is what the vendor argues is the difference between itself and a logging extension. The meter supports the same reading: the included credits are denominated in enrichment, contact discovery and engagement signals, not in generated messages.

At the very top of the band and the best articulated oversight position in this category. The vendor names four controls and describes them as enforced rather than advisory: human approval on every step before anything is sent, enforced pacing with automatic withdrawal, qualification gating so only profile matched prospects are contacted at all, and a full audit trail.

It adds a circuit breaker most products here lack, stating that any platform warning on any seat pauses that seat the same day with no exceptions, and a one tool policy requiring the buyer to disconnect other automation before starting. Its external agent interface is described as running actions with the operator's approval rather than freely.

One piece of this is architecturally real rather than asserted, and it is the important one: the audit trail is written into the customer's own system of record, so the log lives somewhere the vendor does not control and the customer can verify.

Held off the top because no configuration surface, rule syntax or numeric pacing limit is published, the safety checklist that would contain them sits on a documentation site that was not read, and the claim set is presented in a marketing article rather than in product documentation.

Research and qualification agents are central to how the product is sold and the system behind them is undisclosed. No model, provider, family or version is named anywhere located, and no accuracy, confidence treatment or fallback is published for the qualification judgement, which is consequential because that judgement decides which people get contacted at all. The nearest thing to a disclosure is a claim about output character rather than about the system, being that the messaging is crafted to read as more human than most.

A linked case study names the customer and the figure, reporting forty percent more qualified opportunities in a single quarter with two named individuals quoted inside it. Beyond that, testimonials carry name, title and employer rather than initials, including representatives at three well known software companies, and one quantifies a personal result of three additional opportunities in a first month.

A marquee reference is named with a public figure attached, describing a social first motion behind a combined thirty million dollars of recurring revenue. Investors are named and checkable and include the founders of the two companies this team came from.

Held off the top because no methodology, baseline or period accompanies any figure, the headline counters on the site are animated elements that did not resolve on retrieval so the numbers could not be read, and no independent review platform record exists at all for a vendor claiming more than a thousand users.

At the top of the band for one control that genuinely restrains who gets contacted rather than how fast. Qualification gating is described as ensuring only real profile fits are touched, which is a targeting discipline almost nothing else in this category offers, and the one tool policy plus the same day seat pause on a platform warning are real operating rules. What is absent is any obligation.

No consent standard, opt out mechanic, suppression capability or regulation of any kind is named for the messages the customer sends, and for a product whose whole proposition is putting social outreach on the same footing as email, the compliance machinery that surrounds email is not carried across with it.

Strict data governance is claimed alongside the certification claim, and the architecture has a real privacy property worth crediting: because activity is logged into the customer's own system of record rather than accumulating solely in the vendor's platform, the customer retains visibility of what was done in their own environment.

Against that, the privacy policy and terms were not read this pass and are recorded as a retrieval limit rather than as an absence, so no legal basis, retention period, transfer mechanism, sub processor list, officer or supervisory authority could be verified.

The population that most needs addressing is the one the product identifies rather than the one that signs up: people surfaced by their public reactions to posts, enriched with work email and mobile number, and scored against a profile they never saw.

The published provenance is a class list rather than a supplier list: public social profiles, professional networks, market trends and industry reports, with data accuracy and relevance asserted. No vendor, licence, lawful basis or refresh position is named, and the credits that meter enrichment and contact discovery buy work email addresses and mobile numbers whose origin is unstated.

The mechanic worth naming separately is engagement signal sourcing, which takes the people who liked or commented on a relevant post and turns them into a qualified audience. That is a milder relative of the follower harvesting graded elsewhere in this category, since the reaction is public and a qualification step sits between collection and contact, and it still creates a record about a person on the basis of a gesture they made toward someone else's content. No notice, lookup or removal route exists for any of them.

At the top of the band, and the placement is uncomfortable precisely because the oversight around it is the best in this category. The vendor states that it uses dedicated agents to simulate human interaction and timing, with dedicated proxies, to keep the account safe.

Proxying the network origin of automated actions is a materially heavier exposure than the pacing and delay mechanisms graded elsewhere here, because it is aimed at making server side activity indistinguishable from a person at their own machine. It is paired with an unqualified promise that the buyer will not be banned if they follow the guide, supported by a claimed record of zero restrictions across more than two hundred connected accounts.

That guarantee is the opposite of what a peer vendor in this same session wrote into its terms, which is that it cannot guarantee any such thing and the customer accepts the risk. What holds this at the top of the band rather than lower is that the protective controls are real and specific: approval on every step, enforced pacing with automatic withdrawal, a same day seat pause on any platform warning, and a requirement to disconnect competing automation first. No account rotation or multiplication is sold.

The training question is unaddressed in every direction. Nothing states whether approved message drafts, research output, enrichment records or the engagement graph feed any model, no tenant boundary is described, and no model provider is named to which such a commitment could attach.

This matters more than usual for a product with a published external agent interface, because assistants connected through it read and write against the same workspace, and the boundary between what an operator's agent sees and what the platform retains is not described anywhere located.

At the bottom of the band. Nothing adopts a false identity, the message arrives from the representative's own account under their own name, and a person reads and approves every item before it goes, so the words that reach the recipient have been endorsed by a human. Three things hold it low.

The vendor markets that its messaging is crafted to bypass filters that detect machine written content and to read as more human than most, which on the kinder reading is a claim about copy quality and on the plainer reading is undetectability offered as a benefit. Timing is simulated by design rather than chosen. And automated profile visits carry the recurring problem in this category, that the entire communicative value of a profile view is that somebody decided to look. Article 50 of the European artificial intelligence regulation goes unmentioned.

At the top of the band and the agent surface is among the strongest in this index. A public agent interface runs on its own subdomain and is stated to connect assistants from five named vendors to the same workspace so they can discover contacts, enrich profiles and run governed outreach.

Role based instruction packs are published as versioned releases on a public code repository, which is a distribution channel almost nothing else here uses, and the interface toolkit is included on every seat rather than sold as an upgrade. Inside the system of record it ships native workflow action cards rather than a webhook, plus a browser extension and list based execution for buyers on tiers without a workflow builder, and a public documentation site.

Held off the top for one deliberate limitation the vendor states plainly: this is a single system of record product, with no support for the other major platform, so the depth is exceptional and the breadth is one lane wide.

No country, region, cloud provider, residency option or sub processor list was located anywhere on the pages read. The one architectural fact that bears on this axis runs in the customer's favour and is worth recording: the activity record is written into the buyer's own system of record, so the durable history of who was contacted and when sits in an environment whose residency the buyer already controls. What sits in the vendor's environment, being the enrichment records, the engagement graph, the qualification output and the connected account session, has no stated location at all.

At the top of the band, and it sits above the two peers graded in this category this session for the simple reason that a certification is claimed at all rather than nothing being claimed. The claim is made without its type, which is the recurring weakness this band was written for: an audited report named without saying whether it covers a point in time or a period tells a reviewer very little.

No auditor, audit period, report request route, trust portal, status page or penetration test was located, and no enumerated control set is published, though a documentation site carries an operational safety checklist that was not read. The platform holds connected account access for its users and runs proxied infrastructure on their behalf, which is what makes the missing detail consequential.

One price, one unit, and every variable around it quantified. Ninety nine dollars per seat per month with five hundred credits included per seat, a single shared credit pool covering enrichment, contact discovery and engagement signals, and overage published at two cents a credit in packs of five hundred. No tiers, no setup fees, self serve signup with no card.

Two boundaries are published that most vendors leave for the call: five or more seats adds a named account manager, and above twenty five seats it becomes a sales conversation, so the buyer knows in advance where self serve stops. The piece that lifts this to the top is the treatment of the external dependency, which this index has penalised repeatedly when it is hidden.

The vendor states plainly that it works on every tier of the host platform including the free one, that the native workflow cards require a professional tier of one of three specific hubs, and that buyers below that tier run through the extension and lists instead. A load bearing prerequisite named with its exact threshold is the disclosure that separates a real price from a headline one.

At the top of the band on an architectural answer rather than a contractual one, and it is a genuine one. The product's entire purpose is writing every connection request, message and reply into the customer's own system of record with its trigger context attached, so the activity history, the attribution and the contact records already live outside this vendor before any exit conversation begins. A departing customer keeps the thing the product was bought to produce.

What is missing is everything the contract would say: no retention period, post termination window, deletion timeline, deletion artefact or export route for the enrichment records, engagement graph and qualification output that accumulate on the vendor's side. The terms of service were not read this pass and are where those would sit.

At the top of the band. No electronic mail is sent by this product, since the host platform handles that side and this one fires social actions, so most of the axis is inapplicable and the grade records what is knowable with that context. Two controls are published and both are more than the category norm: pacing described as enforced rather than advised, with automatic withdrawal of outstanding requests, and a rule that any platform warning on any seat pauses that seat the same day.

A safety checklist covering account health, daily limits, targeting and operating hours is published on a documentation site. What is absent is a single number: no daily limit, ramp schedule, threshold or volume ceiling appears on any page read, so the enforcement is described without being specified.

The sharpest segment boundary in this category, and it is drawn by a prerequisite rather than by a slogan. The buyer is a business to business sales team already running a specific system of record, the vendor says so plainly, and it states equally plainly that it does not support the other major platform. Within that boundary the tiering of the host platform is mapped to what the buyer actually gets, with native workflow cards above one threshold and extension based execution below it.

Two size markers are published, being a named account manager from five seats and a sales conversation above twenty five, so a team can locate itself on the ladder before contact. Held off the top because no headcount band, vertical or geography is stated, and because the ceiling that is published is a service boundary rather than a statement of where the product stops being the right purchase.