La Growth Machine

The engine underneath is mechanical and the price list confirms it. Strip the models and what remains is the entire product: a sequence builder with branching conditions and split testing, connection requests, direct messages, profile visits, likes, follows, tweets and retweets, email sequencing, waterfall enrichment across eleven providers, a company database, a unified inbox, system of record synchronisation, an interface and the proxy infrastructure. None of that requires a model.

The writing assistant is the model dependent piece and the feature comparison table marks it absent from the entry tier, so the cheapest plan is sold complete without it. Voice message generation is the other, and it operates on a recording the seller supplies rather than producing speech from nothing. Recorded as observed: the entry plan card lists the writing assistant among its contents while the comparison table below marks it as excluded, which a buyer cannot reconcile from the page.

Real controls exist and they sit at the campaign design layer rather than at the point of action. The operator configures the sequences, the conditions and the sending limits, manual task steps can be inserted so a human performs a given step, replies are checked across every channel so a response on one stops the sequence on all of them, exit actions can be defined, and out of office detection pauses a lead automatically and resumes when they return.

That last one is a genuine courtesy control and few vendors here have it. What is absent is oversight of what the system actually sends. Messages, voice notes, connection requests, likes, follows and retweets all dispatch without an approval step, and nothing published describes a review queue, a confidence threshold, a withholding behaviour, an escalation path or an audit trail of actions taken under a seller's own identity. Configuring limits in advance is a governor on volume rather than a check on content.

Two model dependent capabilities are marketed and neither is explained. A writing assistant generates message copy, and the voice feature adds generated introductions to a recording the seller made once, so that each recipient hears an opening addressed to them personally. No provider, model family, version, inference location or data handling position is published for either.

The gap that matters most is on the voice feature, because that is the one producing synthetic speech in a real named person's voice: nothing states how the generated portion is produced, whether the seller's voice is modelled or merely spliced, where the audio is processed, how long it is retained, or what prevents the same voice being used for content the seller never approved. For a vendor domiciled in Europe selling a synthetic speech feature to European operators, that is the disclosure a buyer most needs and none of it is present.

Three customer stories carry a named individual, their job title, their company, a description of that company's business and a quantified result: a threefold increase in outbound pipeline with the share of total pipeline moving from roughly a tenth to a third, a stated pipeline figure in the low millions of euros from intent led outreach, and outreach running two and a half times faster. Two further testimonials elsewhere are similarly attributed.

That combination of name, title, company and number is what this axis asks for and most vendors here do not supply it. A customer stories section, a public wall of testimonials, a stated base of more than ten thousand teams and an independent review score are all present, and the displayed logos include several well known technology companies.

Off the top band because the headline claim that multichannel produces three and a half times more replies is repeated across the site with no methodology, sample or comparison basis attached, and because a cost comparison against hiring an assistant cites a monthly salary range with no source.

Four regulated surfaces run through this product, being email, the professional network, a second social platform and telephone calls, and no acceptable use policy, sending policy or anti spam policy was located anywhere on the surface read. No statute is named on any page examined, which is notable for a company domiciled inside the jurisdiction that wrote the strictest of them and selling to operators contacting European individuals.

Two things hold this at the middle band rather than lower. A processing agreement is published as a downloadable document rather than offered on request, which is a real instrument and more than most here provide. And several product behaviours are compliance adjacent in substance even if never framed that way: contact verification runs before a send so invalid addresses are not mailed, out of office replies pause the sequence, and a reply on any channel halts the sequence on all of them. Those reduce nuisance contact without ever being presented as a compliance position.

The estate is better assembled than most and its substance was not read this pass. Four legal items sit in the footer of every page: terms, a privacy policy, a downloadable processing agreement published as a document rather than promised on request, and a dedicated page on account safety. A European domicile places the vendor inside the strictest data protection regime by default rather than by election, and support is stated to be handled internally in France rather than outsourced.

What could not be assessed is the content: legal bases, retention, transfer mechanisms, sub processors, data subject rights and the position of the enriched individuals were all left unread and are recorded here as unexamined rather than absent.

The question a reader of the full policy should put first concerns the people in the enrichment output rather than the customers, since the product returns personal email addresses and an inferred gender field for individuals who have no relationship with either party.

The supply chain is quantified and anonymous at the same time. Waterfall enrichment is stated to query nine email providers plus two verification tools, which is an unusually precise count and more than most vendors disclose, and not one of the eleven is named. A company database of more than twenty seven million records sits alongside it with no sourcing statement.

The output list is where this becomes pointed: an enriched lead may return a professional email, a personal email, a telephone number, a social handle, company details and an inferred gender. A personal email address and an inferred demographic attribute for a person who never contracted with anyone is the category this axis exists to surface, and no notice, lawful basis, source disclosure or removal route for those individuals was located.

Held at the middle band rather than lower because the count of suppliers, the distinction between finding and verifying, and the per unit consumption of each enrichment type are all published, which is real structural transparency about how the pipeline works even without naming who is in it.

The most explicit detection avoidance positioning in this index, and it is the vendor's own marketing rather than an inference. A homepage section is headed with a statement about staying under the radar. A pricing page section states the product is built for stealth. The cloud architecture is sold on the basis that there is no browser extension to detect and that the platform is invisible to the professional network's detection systems.

Every connected identity is given a dedicated mobile proxy positioned near that person's real location, described as making the traffic indistinguishable from a human browsing on a phone. Action timing is randomised to imitate human behaviour. That is infrastructure built and marketed for the purpose of defeating a platform's enforcement.

The exposed action set is also the broadest graded here, spanning connection requests, messages, voice notes, profile visits, automated likes and follows on one platform and tweets, favourites, retweets and follows on another, with the buyer's own accounts performing all of it. Sending addresses are additionally rotated five to ten per identity, described as reducing the risk of being flagged.

Genuinely creditable and recorded: limits are configurable and published as derived from real usage data, proxies are dedicated rather than shared, and team members reply without sharing passwords. None of that changes whose account is restricted when detection succeeds.

Nothing addressing model safety or data stewardship in a model context was located on the surface read. The product holds connected mailbox access, professional network session access, message content in both directions across several channels, a unified inbox of live conversations, and a recording of the seller's own voice.

No statement was found on whether any of that material trains, tunes or evaluates models, whether processing is isolated between customers, which providers touch it, how long generated audio or message content is retained, or what happens to the voice recording when a subscription ends.

The voice sample is the most sensitive item and the one least addressed: it is biometric in character, it is supplied once and reused indefinitely across campaigns, and no retention, deletion or misuse position accompanies it. Recorded as unexamined rather than absent for the parts covered by the unread policy and processing agreement, both of which are published and should be read before a buyer relies on this grade.

The bottom band is reached here on accumulation rather than on any single mechanism, and the vendor's own vocabulary settles it. Five separate manufactured authenticity surfaces run under the seller's real identity with no disclosure position anywhere. Voice notes are generated per recipient by adding synthetic introductions to a recording the seller made once, in a format whose entire persuasive weight comes from sounding personally spoken.

A mode named for simulating conversation is sold on producing exchanges that read like a real one. Automated liking of a prospect's posts, following and profile visiting manufacture social signals whose only meaning is that a human chose to perform them. The same applies on the second platform through automated favouriting, retweeting and following. And message copy is model generated while presenting as the sender's own words.

Elsewhere in this index each of these has been held at the middle band individually, on the reasoning that one synthetic surface is comparable to a pre recorded voicemail. That reasoning does not survive five of them stacked under one identity, marketed with the words simulate and stealth, from a vendor inside the jurisdiction whose transparency obligation for synthetic content applies most directly and which is not mentioned anywhere.

The deepest agent era surface of any vendor in this index, and it is not close. Three distinct artefacts sit in the site footer under product: an agent protocol server with its own page and a separate toolkit page, a published skills library described as skills for outreach, campaigns and replies built for a named assistant, and a public source repository holding the same material.

No other vendor graded here publishes a skills library at all, and this is the sixth agent protocol server in the index. Underneath that sits a conventional integration surface that would earn a strong grade by itself: public interface documentation with enumerated operations for creating, stopping, converting and removing leads and for building an audience from a search address, webhooks, bidirectional automation platform connections that both receive and emit triggers, connectors to a data orchestration platform and a messaging platform, and native system of record integrations whose synchronised fields are enumerated line by line across both directions, including message content sent and received on every channel. Any other system of record is reachable through the automation layer with named examples. The one reservation, recorded rather than penalised: the interface is gated out of the entry tier.

No residency position was located on the surface read, which is a conspicuous omission for a European vendor whose likely buyers ask about it first. What is known bears on the question without answering it. The company is French, its support organisation is stated to be internal and based in France rather than outsourced, and the site is published in four European languages, all of which point to a European operating footprint without committing to one.

Against that, the architecture necessarily distributes: every connected identity is assigned a dedicated mobile network connection positioned near that person's own real world location, so the traffic path is deliberately spread across whichever countries the customer's team occupies, and nothing published explains where the session data, message content or enrichment output produced through those connections is stored or processed. No region, data centre, hosting provider or sub processor register appears anywhere. A processing agreement is published and was not read, and it is the document most likely to carry the answer.

No certification, attestation, trust centre, audit, penetration test or vulnerability disclosure route was located anywhere on the surface read, and the footer legal block contains no security item among its four entries. That matters more here than for most vendors in this index because of what the platform holds: an authenticated session for each connected professional network account, mailbox access including the older protocol path that requires a stored username and password rather than a revocable token, message content in both directions across four channels, and a recording of each seller's voice.

Two things keep this at the middle band rather than lower. A processing agreement is published as a downloadable document, which necessarily carries security commitments even though its content was not read here. And the architecture includes a real isolation property, in that each identity receives its own dedicated network path rather than sharing a pooled one, which the vendor presents as an evasion benefit but which also limits cross customer correlation.

Among the most complete pricing disclosures in this index. Three tiers carry real figures, three currencies are selectable, and four billing periods are offered with the annual saving stated concretely as two months free rather than as a percentage. A full feature comparison table runs to well over a hundred rows with every inclusion and exclusion marked, covering actions, conditions, integrations, field level synchronisation and support.

The unit of billing is defined explicitly and then defended in the answers below, where the difference between a billed identity and a free team member is explained at length, along with what each can and cannot do. Consumption pricing is published to the unit, at five credits for a full enrichment and one for data only, and a working calculator converts a lead volume into a cost. Support is sold with stated response times per tier.

Cancellation, upgrade and downgrade are described as available at any moment. Two further items are unusual enough to name: a fifty percent discount for organisations pursuing social or environmental objectives, with the qualifying process published, and a plain statement that support is handled internally in France by a named team rather than outsourced.

Recorded as observed: the middle and top tiers display the same monthly figure while a comparison elsewhere on the same page cites a different amount for the middle tier, and the two cannot be reconciled from the page.

Portability is a shipped feature rather than a promise, and it is not metered. Spreadsheet export appears with an inclusion mark against all three tiers in the comparison table, so a buyer on the cheapest plan can take their lead and campaign data out without buying up, which is the direct inverse of the credit metered export seen elsewhere in this index.

The vendor states plainly that campaign and lead data can be exported at any time or pulled into the buyer's own dashboards through the interface and the automation layer, and frames it as the reporting stack remaining theirs. Bidirectional synchronisation means message content, activity and lead status also accumulate inside the buyer's own system of record throughout the relationship. Commercial exit is clean, with subscriptions endable at any moment and downgrades available freely.

Off the top band because the post termination half is entirely unaddressed on the surface read: no retention window, no deletion timeline, no deletion confirmation, and no statement of what becomes of stored message content or the seller's voice recording once a subscription ends.

Several named controls, and two of them protect the recipient rather than only the sender. Contact verification runs before a send, using a waterfall that separates finding an address from verifying it and charges differently for each, so invalid addresses are identified rather than mailed. Out of office replies pause the sequence automatically and resume when the person returns, which is a courtesy control almost nobody else here implements.

A reply arriving on any channel halts the sequence across all of them. Sending is distributed across multiple addresses per identity, five on the middle tier and ten on the top, with the stated purpose of protecting the sending domain and scaling volume safely. Daily action limits are configurable and the vendor publishes them as derived from years of real usage data. An automatic signature is added with the stated aim of avoiding spam placement.

Off the top band because the rotation is explicitly framed as reducing the risk of being flagged, which treats the signal as an obstacle rather than as feedback about volume, and because no sender authentication guidance, blocklist monitoring, complaint threshold, bounce policy, warmup process or placement testing is named anywhere.

Segmentation is stated per tier and reinforced by structural limits that make the statements testable. The entry plan is scoped to solo founders and teams beginning outbound and caps identities at three with three team seats. The middle plan addresses scaling teams running multichannel and lifts identities to unlimited with twenty five seats. The top plan is aimed at scaling teams with complex system of record workflows.

Two further plans sit outside that ladder with published qualifying conditions rather than vague enterprise language: an agency plan carrying a stated six identity minimum, and a build your own plan carrying a six month minimum commitment. Three dedicated pages address distinct buyer roles covering engineering minded operators, operations owners and sales leadership.

Off the top band because no headcount band, revenue band, ceiling or disqualifying condition appears anywhere, so the vendor describes who each tier suits without ever saying who the product is not for.