CueGrowth

A genuine middle case, the second on this axis after Clevenio. What survives removal of the models is real: a searchable profile database, list import, a campaign sender and an inbox, which is a saleable product. What survives is not what the vendor sells.

Every step it differentiates on is model dependent, from ingesting a presentation or video into structured positioning, to generating ICP search queries rather than filling filter boxes, to enrichment columns computed per lead, to drafts that cite a prospect's stack. The vendor's own comparison set is Clay, Apollo and Outreach, and the argument against all three is context carried by a model across steps. No edition without the models is offered, including the free tier, which ships offering intelligence and draft generation.

An approval position is stated three separate times and one of them sits where it counts. The section explaining the workflow is headed with the claim that the operator reviews while the model does the heavy lifting, the six step strip ends with an approve and send step, and the free tier feature list includes human in the loop review before anything sends, so oversight is described as a property of the cheapest plan rather than an enterprise control.

Off the top of the band because the review is asserted rather than specified: whether it can be disabled, what happens at volume, what the model is prevented from doing, and whether any audit trail records what was drafted against what was sent are all undescribed.

Grounding is described better than most and the model itself is not described at all. On the good side, the inputs are enumerated: uploaded positioning, proof points and ICP criteria, 50 or more enrichment signals, and named public sources including GitHub activity, documentation and job posts, so a buyer can see what the drafts are built from.

Provider, model family, version and inference location go unnamed, and no accuracy or error rate is published for enrichment, for ICP matching or for the intent tags that decide which prospects a rep works. Held at the top of the band on the strength of the input disclosure.

The surface looks strong and does not survive being read closely. Twelve named customer logos include substantial companies, among them Forrester, Pendo, SnapLogic, QuestionPro and IdeaScale, and a full testimonial carries a name, title, company and photograph. The problem is that the same three figures appear against different customers on the same page. A 340 percent reply rate lift is attributed to Skillfully in the outcomes panel and to SnapLogic in the ticker beneath it.

An 85 thousand dollar result is attributed to QuestionPro as pipeline in 60 days and to IdeaScale as revenue in two months. More than 200 meetings a month is attributed to Lava in one place and to Forrester in another, while Lava appears elsewhere against a different figure entirely. No method, sample or period sits behind any of them.

Regulation goes unaddressed across every page read, on a product that sends cold email at volume and, on its own integrations page, connects with prospects on a professional network automatically and follows and interacts with them on a second platform. The controls that exist are commercial rather than legal: sending limits and bounce monitoring in the campaign dashboard govern throughput and reputation, not consent.

A lawful basis position, a suppression mechanism, an unsubscribe commitment and a data subject request route are all absent from the product surface, and a published data processing addendum exists but was not read this pass.

Three legal documents are published and footer linked, and one of them is a standalone data processing addendum, which is uncommon and creditable at this size. None were read this pass. Held at the top of the band rather than higher because the gap is structural: a stated 900M profiles sit behind the search, enriched with 50 or more signals per person, and the people in that database have no notification, self service lookup or removal route described anywhere on the vendor's surface. A data protection officer, an EU representative and a sub processor list are also absent from the pages read.

Scale is published and sourcing is not. A stated 900M profiles support the ICP search, enrichment runs across 50 or more signals, and no provider, licence, collection method or refresh cadence is named for any of it. The enrichment sources that are named introduce a shape the index has not graded before: GitHub activity, documentation and job posts, so a developer's public commit history becomes a qualification signal in a sales database they never entered. That is public information used lawfully in most readings and it is still a profile assembled without the subject's knowledge, with no notification and no removal route published.

The current homepage sells email, and the professional network exposure sits one layer down and is still live. The integrations page describes identifying website visitors and connecting with them on that network automatically in real time, and following and interacting with leads on a second platform through a workflow template.

The vendor's own product screenshot settles what it is: the offering intelligence demo is loaded with CueGrowth's own positioning, which reads as an AI powered outbound platform for that network, and its stated customer problem is reps spending three to five hours a day on manual work there. The help centre still refers to campaigns running on a predecessor server. No conformance position, rate limit statement or account risk allocation was located. Held at the middle of the band because rotation, proxies, multi account stacking and undetectability marketing are all absent.

The corpus this product ingests makes the training question sharper than usual. Offering intelligence takes customer presentations, videos and documents and turns them into structured positioning and proof points, so what lands in the platform is the customer's own competitive material rather than only contact records, and it sits alongside reply text and campaign performance. Whether any of it trains or improves models serving other accounts goes unstated on every page read. A published data processing addendum is the likely home for a purpose limitation clause and was not read this pass, which is the one document that could move this grade.

The Coldreach formulation applies exactly: the research is real and the researcher is not. Published example drafts cite a prospect's adopted tooling and architecture, drawn from genuine public sources, and the recipient receives the individual attention that once meant a person had spent an hour on them, produced by a model in seconds. Article 50 goes unmentioned.

Held at the middle of the band on three counts that matter: messages send from the operator's own identity, nothing impersonates a named human or uses a synthetic voice, and the vendor states that a person reviews before anything sends, which puts a human between the model and the recipient even though the words are not theirs.

A real integration surface exists and it points sideways rather than into the revenue stack. The published set covers Zapier and Make for workflow automation, a custom API webhook for full data flow control, website visitor identification feeding campaigns, and audience population into Meta Ads, Google Ads and a social platform for retargeting, alongside a partner programme and a documentation site.

The conspicuous absence is a named CRM connector on the current product surface, which is unusual for a product positioning itself as the operating system for outbound, and the top plan's promise that all integrations are included does not enumerate what they are. No public API reference, developer portal or agent facing endpoint was located.

The residency question goes unanswered across every page read. Hosting provider, processing region, storage location and sub processor list are all absent, and the only infrastructure reference located anywhere is a help centre article describing campaigns activating on a predecessor server. The gap carries weight given what the platform holds: a stated 900M profile database, customer sales collateral ingested as offering intelligence, and connected mailbox activity.

The security posture appears once, as a line item in the enterprise tier promising single sign on and advanced security, which places authentication controls behind a quote rather than describing them. No certification is claimed, and no security page, trust centre, audit, penetration test, control set or status page was located.

Two things keep it mid band rather than lower: no certification is asserted and left unsubstantiated, and a published data processing addendum exists, which is where technical and organisational measures are usually enumerated and which was not read this pass.

The published structure is better than most: a free tier with the workspace, offering intelligence, list import, enrichment and draft previews included, then Accelerate at 299 US dollars per user per month carrying 15,000 credits with an overage rate published at 0.025 per credit, then a custom enterprise tier. The plan description also states what is not capped, naming no limits on campaign contacts, saved contacts or inbox responses, which is a useful inversion of the usual gate list.

Off the top of the band on two counts. A credit consumption table was not located, so a buyer cannot tell whether 15,000 credits is a month of work or a week, and at 0.025 each that unknown decides the real bill. Single sign on and advanced security sit only in the unpriced enterprise tier, which makes the authentication posture a quote rather than a price.

The routes in are documented and the routes out are not. Lists arrive by CSV upload and integration, and the custom API webhook offers full flexibility over data flow, which in practice gives a technical customer a way to move enriched records and replies into a system they own. What is missing is any statement of it as a commitment: an export function, a post termination retention period, a deletion timeline and a deletion artefact are all absent from the pages read. A published processing addendum is the likely home for those terms and was not read this pass.

Deliverability is instrumented as a product surface rather than claimed as a virtue, which earns the top of this band. The campaign dashboard is described as monitoring sending limits, bounces and performance across active campaigns, and the reporting step promises deliverability, reply and success rates with an explanation of why outcomes moved.

The standard control set is absent from every page read: warmup, authentication named by protocol, bounce or complaint thresholds, blocklist monitoring, placement testing and suppression handling. The sending architecture is never described either, so a buyer cannot tell whose reputation carries the risk.

One of the clearer market definitions graded, and it is corroborated by build decisions rather than page copy. The stated audience is teams selling technical products to technical buyers, with four dedicated solution pages for developer tools, cloud and infrastructure, enterprise software and complex B2B services, and three named roles in technical sales development, enterprise account management and revenue operations.

The proof that this is real rather than positioning is the enrichment set: a vendor that says technical buyers and then ships GitHub, documentation and job post signals has spent engineering on the claim. The logo set matches the claim as well. Off the top of the band because no headcount band, customer count, size distribution or ceiling is published anywhere.