lemlist

The company predates the generative wave by several years and built its reputation on custom image personalisation and mailbox warmup, neither of which requires a model. Apply the removal test and a complete, mature and differentiated product remains: multichannel sequences mixing email, professional network actions, calls and messaging apps, a large lead database, waterfall enrichment, a separately sellable warmup product, inbox rotation, split testing and a unified inbox.

The model layer adds generated personalisation, a voice message feature and automated extraction of lead details from profiles and websites, all of which sit on top of a working platform rather than constituting it. Graded at the middle band on vintage rather than marketing volume, the same reading this index applies to every established platform that added a programme to a product that already sold.

Sequences run unattended across four channels once configured, mixing email follow ups, professional network actions, messaging app messages and calls, with steps advancing on recipient behaviour. Three constraints exist and all three are commercial rather than editorial. The email plan carries a published ceiling of fifty thousand messages a month. Sending addresses are limited per plan with additional slots sold individually.

And a strict one account to one person rule is enforced, so a single subscription cannot be operated as a shared sending pool. What is absent is anything at the point of action: no approval step stands before a generated message or an automated professional network action, no review queue exists, no escalation path is published, and no per action audit record was located. The unsubscribe suppression is a genuine automatic stop, but it triggers on recipient behaviour rather than operator oversight and only when the operator included the link.

Several model branded features are marketed, covering generated personalisation, automated extraction of lead details from profiles and websites, and generated voice messages, and none of them carries a model, provider, version, hosting location or processing boundary in the material read.

The voice feature is the one where the absence matters most, because a synthesised voice reaching a prospect raises questions about how it was produced and from what input that no published documentation answers. The personalisation claim compounds it, since the stated design goal is output that does not read as machine written, and a buyer relying on that has no description of the mechanism producing it. The gap sits oddly beside an otherwise specific disclosure posture, where hosting provider, country, statutes and per unit credit costs are all named plainly.

Market position here is genuinely established and is visible from an unusual angle: seven independent vendors have published detailed pricing breakdowns and alternative comparisons about this product, which makes it the reference point competitors position against rather than a vendor claiming a position. A profile exists on a major software review platform with product detail and media supplied by the vendor. Against that, no customer evidence was read this pass.

No named customer, case study, quantified outcome, attributed testimonial or review count was examined, and the vendor's own customer pages are recorded as unexamined rather than absent, which is the honest position and is flagged for re verification. Market prominence is not the same thing as a published result a buyer can check, and this axis measures the second. The database size claim also varies between sources at four hundred and fifty million and six hundred and fifty million records without a dated figure to reconcile them.

This vendor does something almost none of its peers do: it tells the buyer that the regulation it addresses is not the only one that applies. Its support documentation names four separate regimes, the European data protection regulation, the United States commercial email statute, the Canadian anti spam law and the United Kingdom electronic communications rules, and states plainly that laws beyond the first may apply to the buyer's outreach.

The controller and processor split is stated explicitly, with the buyer named as responsible for ensuring their outreach complies. Sourcing is described as using only publicly available information, exports of contact data require the user to confirm compliant use, and recipients who unsubscribe are removed from the campaign and cannot be contacted again by that user. A recipient can also request deletion directly by email.

Off the top band on one structural weakness that matters: the unsubscribe link is optional. The documentation says a recipient is suppressed when a link is included and strongly recommends including one, which means the suppression mechanism only operates if the operator chose to switch it on. A compliance control the sender can decline is not the top band. No separate acceptable use policy was located.

The instruments are published rather than gated, which is the first thing this axis looks for. A processing addendum sits at its own address with the standard document viewable directly and a form for a countersigned copy, it is stated to supplement the terms and privacy policy, and it references model clauses for transfers.

The controller and processor roles are named explicitly, hosting is identified by provider and country, storage is stated to be within the European Economic Area, and recipients who are not customers have a published route to request deletion by email. Sourcing is characterised as publicly available information only. Off the top band on four counts. No sub processor register was located anywhere. No data protection officer or supervisory authority route is identified.

The privacy policy itself was not read this pass and is recorded as unexamined. And the vendor states it will not sign a customer's own addendum or make individual changes to its standard one, which is candid but leaves an enterprise buyer with a take it or leave it instrument.

Two statements do real work and neither identifies a source. The vendor states that its database is built from publicly available information only, which is a sourcing characterisation rather than a legal basis but is more than most offer, and it publishes a route for a person in that database to request deletion by writing to a named address. Contact exports require the user to confirm compliant use.

What is absent is the chain itself: the enrichment is described as a waterfall drawing verified addresses and telephone numbers from the market's top providers, and not one of those providers is named anywhere in the material read. The database size claim varies between four hundred and fifty and six hundred and fifty million records across sources with no dated figure published to reconcile them.

This sits at the top of the middle band rather than higher because a named removal route and a stated sourcing basis are real, and the vendors placed above it on this axis all named at least one actual supplier.

The sending and record paths are approached conventionally: messages leave through the customer's own connected mailboxes, and three major systems of record are named as integrations. Nothing located markets account rotation across professional network profiles, proxy infrastructure, undetectability or any language about avoiding platform enforcement, which keeps this well clear of the lower bands.

The exposure is the professional network automation itself, which runs as a step type inside sequences and is characterised by a third party as executing server side rather than through a browser extension. That is architecturally similar to a favourable posture recorded elsewhere in this session, since actions performed from vendor infrastructure move enforcement risk away from the buyer's own account, but this vendor makes no such claim itself and no conformance position of any kind was located. The platform's user agreement is nowhere mentioned in the material read.

No training position, cross tenant boundary commitment or retention statement for model inputs and outputs was located, and no model provider is named so no provider terms can be inspected either. The architectural facts that do exist bear on the question without answering it: data is stated to rest on named French servers within the European Economic Area, and a processing addendum is published, both of which constrain where content can sit without describing what is done with it.

The exposure is real, since the platform holds campaign copy, conversation content across four channels, uploaded personalisation assets and a database of contact records for people who never contracted with anyone, and generated personalisation necessarily reads lead detail drawn from profiles and websites to produce each message. The published addendum was not read this pass and is flagged.

One line of the vendor's own product material states the concealment goal outright, promising personalisation at scale with a model while not sounding like a model. That is the manufactured impression of human authorship named as the objective rather than implied.

Around it sit several surfaces pointing the same way: generated voice messages, custom images and video carrying the recipient's own name and company marks, automated professional network actions performed under the sender's real identity, and messaging app automation. Against that stand real mitigations the vendors graded lower on this axis did not have.

A recipient who unsubscribes is suppressed from further contact by that sender, a recipient can request deletion directly by email, the controller and processor split is stated, and four separate regulatory regimes are named in the vendor's own guidance. Held at the middle band on that balance, and specifically not lower because nothing establishes that the voice feature reproduces a particular person's voice, which was the aggravating fact in the comparable case. No disclosure of automation at the point of contact is published, and the European transparency obligation is nowhere addressed.

Three major systems of record are named as integrations, a browser extension is published and is in fact the entire content of the free tier, and a structured support centre with individually addressable articles is maintained to a standard well above most vendors in this index. A unified inbox consolidates activity across email, professional network, messaging apps and calls, which is an integration of channels rather than of systems.

What was not confirmed from vendor sources this pass is the programmatic layer: no interface reference, webhook surface, automation connector listing, marketplace or agent protocol server was located in the material read, and grading from general reputation rather than from published evidence is exactly what this index declines to do. Recorded as unexamined rather than absent and flagged for re verification, since a documented interface would move this grade.

The residency claim names the thing most vendors leave vague, which is who actually holds the machines. The vendor's own support documentation states that all data stays on French servers hosted by a named European cloud provider, and separately that data is stored within the European Economic Area.

Naming the hosting provider and the country together, in an unqualified sentence about all data rather than a hedged one about some, is a stronger disclosure than the general European commitments that earn this band elsewhere in the index, and it is published in support documentation a customer can find rather than buried in a policy. No regional tier is sold on top of it, which is consistent with a single stated location rather than a menu.

Two items are recorded rather than treated as disqualifying: no sub processor register was located, so where the enrichment suppliers and telephony providers sit is unaddressed; and the published processing addendum refers to model clauses, which implies some transfer path exists that the residency statement does not describe.

No certification of any kind was located, and there is no trust centre, audit report or period, penetration test, dedicated security page, vulnerability disclosure route or status page in the material read. What the vendor does publish in place of that is unusually candid and is worth recording precisely because it is the kind of thing most companies conceal: on its own processing addendum page it states that it cannot make individual changes to its standard agreement because, as a small team, it does not have a legal team on staff.

That is honest, it explains the take it or leave it posture on contracting, and it is also a material signal for any buyer whose procurement process expects negotiated terms and independent attestation. Behind the silence sit connected mailboxes across a customer's whole team, conversation content on four channels, and a contact database covering hundreds of millions of people.

The vendor's own support documentation publishes a clear and current picture: an email plan priced by monthly sending volume at sixty nine dollars a month for fifty thousand messages, falling to sixty two quarterly and fifty five yearly, and a multichannel plan at one hundred and nine dollars per user per month on the same discount ladder of ten percent quarterly and twenty percent yearly. An enterprise tier is quoted, annual only, with a stated five seat minimum.

A fourteen day trial requires no card. Consumption is priced to the unit at one cent per credit, five credits for a verified address and twenty for a telephone number, charged only when a lookup succeeds, and additional sending addresses are published at nine dollars each per month above the included allowance. The vendor also announced a ten dollar increase in its own documentation and published the grandfathering rule alongside it, which is more transparency than most price rises get.

Off the top band because calling on the multichannel plan carries usage based telephony costs with no rate published, because the enterprise tier is quote only, and because the catalogue was restructured mid year while legacy accounts were grandfathered, so seven independent pricing trackers published seven materially different pictures within months of each other and a buyer's structure depends on when they arrived.

Export is a working capability rather than a promise, and it sits at the very bottom of the range: the free tier consists of nothing but a browser extension that finds, verifies and exports lead records, so extraction is available before any money changes hands. The published processing addendum carries the contractual terms governing personal data, and a recipient level deletion route is published separately. Against that, the account level questions are unaddressed in the material read.

No export format or schema is specified, no retention window or deletion timeline is published for campaign history, conversation content or lead records after cancellation, and no programmatic route out was confirmed. One adverse term is recorded from third party reporting rather than vendor documentation and flagged as such: enrichment credits are described as expiring at the end of each month with no rollover, so prepaid capacity does not survive a quiet period.

Mailbox warmup here is not a feature checkbox but a separately branded product with its own standalone price of twenty nine dollars per seat per month, included at no extra cost on every paid plan, which is a materially stronger commitment than the warmup toggles most competitors ship. Around it sit inbox rotation, a deliverability boost, and address verification charged only when a lookup succeeds so the commercial design rewards not sending to bad addresses.

Additional sending addresses are published at a stated monthly price rather than bundled invisibly. The email plan carries a published ceiling of fifty thousand messages a month, which is a real volume boundary and rarer in this category than it should be.

Off the top band because the published discipline stops at the tooling: no complaint rate threshold, bounce handling policy, authentication guidance, blocklist monitoring or list hygiene rule was located, and inbox rotation is offered without any stated position on what it is for, which matters because the same mechanism is sold elsewhere in this index explicitly as a way to send past limits.

Buyer types are named plainly and differ by plan, with the email tier aimed at cold emailers, recruiters, solo operators and founders, and the multichannel tier at sales teams needing professional network steps, calls and messaging apps. Two real structural boundaries exist. The enterprise tier carries a stated five seat minimum on annual billing only, which is a floor rather than an adjective.

And a strict one account to one person rule is enforced, with the support documentation explicitly answering that multiple people cannot share one subscription, which is a genuine constraint a buyer needs to know before modelling cost.

Off the upper band because nothing states an upper ceiling, an unsuitable use case or a disqualifying condition, because no industry or region is named despite a European operating base, and because the plan naming and structure changed during the year, so the segment a given tier addresses depends on when the buyer arrived.