Terret

The forecasting engine is modelled and it is the reason the product exists. The vendor was founded in 2018 as a predictive platform extracting insight from unstructured data, and the components it names are machine forecasting, projections, deal and engagement risk scoring and activity matching, all of which are inference rather than aggregation.

Activity matching in particular is a probabilistic problem: deciding which email thread and which meeting belong to which opportunity across systems with no shared key cannot be done with rules at useful accuracy. Strip the models and what remains is the business intelligence module, meaning dashboards over revenue data the customer already had, which is a materially smaller product rather than the same one with a feature disabled.

Held below the top band because that business intelligence module is a substantial named part of what is sold and is conventional analytics, and because conversation intelligence was added natively only in 2025, so part of the current surface postdates the modelling core. Ask which capabilities stop working without the forecasting engine, and whether the business intelligence module is separately licensed.

An automation component is named and nothing describes what it may do. The platform includes an assistant marketed for automating tasks and improving productivity, and independent coverage describes a focus on agents that both analyse and act across the revenue lifecycle, which is the language of systems taking action rather than surfacing recommendations.

Nothing published states what those agents may change, whether they write into the customer record system, whether any approval precedes an action, what an administrator can constrain by role, or what halts activity in progress.

The exposure is bounded in one respect worth recording: the platform's outputs are forecasts, dashboards and risk scores consumed by revenue leaders rather than messages sent to buyers, so a wrong output misleads an internal decision rather than reaching a customer. That limits blast radius without describing controls. Ask what the automation component can change without approval, whether it writes to the customer record system, and what an administrator can disable.

The architecture is named at the component level and nothing beneath it is disclosed. The vendor identifies machine forecasting, projections, deal and engagement risk scoring and activity matching as distinct engine components, which is more granular than a generic assistant claim and tells a buyer where inference happens. Below that nothing appears: no model provider, family or version, no model card, and no evaluation of any component.

The accuracy gap is the material one and it compounds across the product, because activity matching feeds deal scoring which feeds the forecast, so an unmeasured error at the first stage propagates through every number a revenue leader acts on, and none of the three stages carries a published figure. Describing activity matching as best in class without publishing what class means is the specific instance. Ask which models power forecasting and risk scoring, and for measured accuracy at each stage from activity matching through to the forecast.

A forecasting vendor that publishes no forecast accuracy, which is the finding here. The product's central promise is calling the right number early in the quarter, and the supporting claims are superlatives rather than measurements: superior accuracy, best in class activity matching, the widest range of forecast use cases and the fastest time to value in the industry, none of them accompanied by a figure, a comparison basis or a study.

For a category where accuracy is measurable against realised outcomes every quarter, and where the customer's own historical data would supply the ground truth, the absence of any published measurement is conspicuous rather than ordinary. What is genuinely evidenced is the customer base, which is substantial and named across networking, database, data warehousing, education technology and automation software, and which corroborates that the platform operates at enterprise scale.

Independent coverage adds one operational caveat the vendor does not, that setup requires adjustment for complex revenue models. Ask for forecast accuracy measured against realised outcomes across a stated customer population and number of quarters, and for the activity matching accuracy rate.

Terret's agents draft outreach to named buyers and queue follow up emails, so the product now reaches people outside the customer's company. Terret publishes no position on consent, suppression or unsubscribe handling for that outreach, and its privacy policy addresses the data of its own users rather than the prospects in a customer's pipeline. The product also ingests email, calendar, call recordings and CRM activity involving those prospects, who have no relationship with Terret.

Terret's privacy policy, issued by Vocalo Inc. doing business as Terret, covers data collected through the website and through linked email, calendar, phone and CRM accounts, on the basis of the user's consent. It supplements an end user service agreement that is not public. Removal requests go to a support address, and a separate deletion policy lets enterprise customers delete data for all or some users. No data processing addendum, subprocessor list or transfer mechanism is published.

No external corpus is licensed and the provenance question is internal and unanswered. Everything the platform reasons over originates in the customer's own systems, whether the customer record platform, mail and calendar, connected conversation intelligence tools, product analytics or the customer's warehouse, so there is no purchased database, cooperative or third party feed to trace. The question that does apply concerns aggregation.

A forecasting engine improves with historical outcome data, and nothing published states whether patterns learned from one customer's closed deals inform the models or benchmarks serving another, whether any cross customer benchmarking exists, or what form it would take if it did. For a vendor whose accuracy claim is its principal differentiator, whether that accuracy is built partly on other customers' revenue histories is a question a buyer is entitled to ask. Ask whether outcome data from one customer contributes to models or benchmarks used for others, and in what form.

Integration runs through official routes across an unusually wide set of systems, and the vendor's posture toward its own competitors is the notable part. Named connections cover the major customer record platform, two conversation intelligence vendors it competes with directly, two sales engagement platforms, product analytics tools and the customer's own data warehouse.

Connecting to rival conversation intelligence products rather than requiring customers to abandon them is a low exposure architectural choice as well as a commercial one, because it means the platform consumes data through those vendors sanctioned interfaces instead of building its own capture against the same meeting platforms. Nothing resembling credential storage, scraping or unsanctioned automation appears in anything located.

Held below the top band because no stated conformance position against any specific platform's terms was found, and because the activity capture layer necessarily reads mail and calendar through grants those platforms control, with nothing published about the permission scopes requested or what happens if access narrows. Ask what permission scopes the mail and calendar capture requests, and whether any named platform conformance commitments exist.

A sensitive combined corpus and no stewardship position located. What accumulates here is the revenue team's correspondence and meetings with external parties, native call recordings since the conversation intelligence addition, complete opportunity and pipeline history, and derived risk and performance scoring about named individual representatives.

Nothing published states whether that material trains or tunes any model, which providers process it when the assistant or the scoring engine runs, what retention applies to recordings and captured activity, or whether anything crosses a tenant boundary. No artificial intelligence governance document, evaluation record or independently audited management standard was located.

The employee dimension deserves its own answer, since engagement and performance scoring about individuals feeds leadership dashboards and nothing describes what recourse a representative has to a score they consider wrong. Ask whether captured activity or recordings train any model, what retention applies, and what recourse exists against automated performance scoring.

Messages drafted by Terret Nexus go to real buyers under the seller's name, built from call transcripts and account research. Terret publishes no default disclosure that a message was drafted by AI and takes no stated position on Article 50 of the EU AI Act. Call recording through Terret Conversation Intelligence also captures external participants, with no published notice or consent behavior.

Genuine openness, including toward competitors, plus a warehouse connection that changes the shape of the integration. The platform ingests from the customer record system, two named conversation intelligence vendors, two sales engagement platforms and product analytics tools, and connects to the customer's own data warehouse so that revenue operations teams retain direct control over the underlying data rather than only seeing it through the vendor's interface.

That warehouse link is the substantive point: it means the analytical layer sits over data the customer can also reach independently, which is a materially different arrangement from a closed platform that holds the only copy. A flexible data model is described as accommodating revenue structures that opportunity centric tools cannot represent.

Held below the top band on verification rather than architecture: no developer documentation, interface reference, authentication model or rate limits were reached on this pass, and no integration inventory was counted. Ask for the interface documentation, the current connector list, and how the warehouse connection handles write back.

Terret is hosted entirely on Amazon Web Services. No region, residency option or tenancy model is named, so the location of a customer's revenue data, call recordings and email activity is settled in the contract rather than on any public page.

Terret has held a SOC 2 Type 2 report since 2020 and retests its security and privacy controls every year, alongside an annual application and infrastructure penetration test. It lists CSA STAR Level 1 and AES 256 encryption, and the platform is hosted entirely on Amazon Web Services. A vulnerability disclosure policy is published. Terret runs no public trust center, and access to the SOC 2 report is not described.

Terret publishes no price, tier, seat rate or contract term for any of its three products, and every commercial route leads to a demo request. The one stated offer is a free 48 hour proof of concept: Terret analyzes the prospect's closed lost deals from the last quarter, names the top three to five loss drivers with quotes from real calls and builds an execution playbook from top performers, at no cost and with no obligation. The end user service agreement that governs a subscription is not public.

Enterprise customers can request deletion of data for all or some of their users by email, specifying the data type, the users and the time frame, and an individual user can request deletion once the company approves it. The warehouse connection BoostUp built keeps part of the revenue data in the customer's own infrastructure. No export path for forecasts, call recordings or agent outputs is published, and the subscription agreement is not public.

Terret now produces outreach as well as analysis. Terret Nexus and Terret Forecast draft emails to stakeholders a deal is missing, queue follow up messages when a champion goes quiet and load drafts into sequences for the seller. Nothing published covers how those messages are delivered: no sending infrastructure, warmup, volume limit, bounce handling or reputation monitoring is described, and no statement says whether delivery runs through the customer's own mailbox or engagement tool.

Coverage is stated on the dimension that actually matters for this product, which is revenue model rather than company size. The platform natively supports subscription, consumption and usage based, product led, renewal, expansion and advertising revenue streams in a single forecast, and independent reviewers consistently identify that as the reason to choose it over larger competitors whose opportunity centric models struggle with usage based or product led motions.

A buyer whose revenue arrives in more than one of those shapes can tell immediately whether this fits. The conventional dimensions are also stated: mid market through large enterprise, roughly one hundred to more than five thousand employees, named industries spanning technology, financial services, education, healthcare and telecommunications, and named buyer roles across revenue operations, sales leadership and analysis. The named customer base corroborates the enterprise end.

Held below the top band because the floor is not stated, no localisation was located, and independent coverage reports that complex revenue models require setup adjustment, which is a caveat attached to the very capability the vendor leads with. Ask what the smallest viable deployment is and what configuration effort a multi stream revenue model requires.