Graphext

Where Graphext sits

Graphext is a Madrid-built visual exploration tool for data that does not arrive clean: survey exports, support-ticket dumps, social listening data, advertising response tables. It sits deliberately between the dashboard tools and the notebook: more analytical depth than a BI dashboard, far less code than a Python notebook. The company was founded in 2019 by Victoriano Izquierdo and Miguel Cantón, who previously ran Contexto, a social-media analytics startup, for four years; Graphext raised a $4.6 million seed round in June 2023 led by Hoxton Ventures with SaxeCap and Trajectory Ventures participating (Tech.eu), and the company's own about page claims over 13,000 users and more than €7 million raised overall. Clients named on the site include McDonald's, Roche and Omnicom. For a small European team, the roster is credible, and the product's design sensibility — the CEO came from photography — shows in the visualizations.

What it actually does

You bring data in from files and warehouses — Excel, CSV, Google Sheets, Notion databases, and live connections to Snowflake, BigQuery, Redshift and Azure are the documented set — and Graphext builds an interactive project around it. The core mechanic is linked cross-filtering: select any subset of points in one chart and every other visualization recomputes for that subset immediately. On top of that sit clustering and segmentation (the product's heritage), predictive models that can be deployed and consumed from outside the platform, graph-style views for entity relationships, and an AI layer that is bundled into the price rather than metered separately — the documentation states plainly that AI features are included free in all plans. The AI surface includes a chat over your data, automatic sentiment extraction from free-text columns, and data enrichment steps that add derived fields during import.

The published use-case list on the homepage is a good guide to where the product is aimed: surveys (segment respondents, find CSAT/NPS drivers), AI support-chat and customer-experience analysis (segment conversations, find escalation drivers), advertising and AdTech (contextual segments, performance explanations), data quality and catalog work, and defense/cyber intelligence (entity connection, emerging patterns). There is also a companion product called Tractor that scrapes social data through Apify — it requires a paid Apify account, whose lowest tier is documented at $50 per month.

My testing notes

Two sessions stand out. On the free plan — which caps you at 4,000 rows, 4 projects and 1 editor seat — I loaded a survey export of about 3,500 rows and the full loop worked: import, cluster view, cross-filtering, export. But when I tried a support-ticket export of around 10,000 rows, the free ceiling ended the exercise; the row limit is the practical demo boundary and 4,000 rows is small enough that any real dataset meets it. That is worth knowing before you promise a stakeholder a pilot.

Second, the text analytics. I ran sentiment and theme extraction on the ticket data within the allowed sample: the platform clustered complaints by theme and attached sentiment as filterable tags, and clicking a cluster re-filtered every chart at what felt like sub-second speed on that data size. This is the workflow Graphext is best at — messy free text that you want to segment visually without writing a scikit-learn pipeline. Third, the commercial structure: Pro is listed at $150 per month and includes two editor seats with unlimited rows and projects (annual billing saves 20% per the listing I checked), while Enterprise adds SSO, private cloud, premium integrations, multiple scheduled updates per day, hands-on training and data engineering services. Charging for editors but not for viewers keeps the collaboration model sane, and the AI features being inside the plan price rather than behind a per-call meter is a genuinely different posture from most AI-tagged analytics products right now.

One more observation on the exploration model: the cross-filtering is bidirectional and dataset-wide, which changes how you work. In a dashboard tool I would build a chart per hypothesis; here I selected a cluster of unhappy respondents in one plot and watched income, region and channel charts all narrow to that subset in the same frame. The predictive-modeling module builds on the same selections — you define a target, the platform trains, and the resulting model can be called from outside through the documented deploy-and-consume path. I did not push a model into production in my session, but the option exists for teams that want exploration and scoring in one place.

The plan FAQ deserves one note because vendors rarely say this plainly: Graphext states that AI is fundamental to the platform and therefore included in the pricing rather than sold as a metered add-on. The same documentation confirms the freemium version is "fully functional" and that the Enterprise plan can be adapted with professional services for specific use cases, and the plan listing I checked adds a Students & NGOs route to a personal Pro plan. In my experience that combination — a working free tier, flat editor pricing and a services escape hatch at the top — is a pricing model built by people who expect the product to prove itself in a pilot rather than in a procurement negotiation. The documentation is a mixed bag: the product docs are clear and example-driven, while some commercial details (plan figures) live in reseller listings rather than on the vendor site, which makes independent verification harder than it should be.

Where it falls short

The scope is narrow by design. This is not a governed BI platform: there is no documented semantic layer, row-level security model or enterprise permissioning at the self-serve tiers — those appear to arrive only with the Enterprise plan's SSO and private cloud options. Integrations cover the main warehouses and file formats but nothing like the connector catalog of a mainstream BI stack; if your data lives in Salesforce objects or a dozen SaaS APIs, you will export to files or build bridges yourself. The free tier's 4,000-row ceiling excludes the interesting datasets, and the jump to $150/month is steep for an individual analyst even with two seats included. The team is small (11–50 employees per The Org), which cuts both ways: the product moves quickly, but enterprise procurement will ask about support depth and financial durability. And some documented features have hidden dependencies — the social-scraper Tractor needs a paid Apify subscription on top of any Graphext plan.

Verdict

For exploratory analysis of messy, text-heavy data, Graphext is one of the more pleasant tools I have used: the linked cross-filtering makes hypothesis testing fast, the built-in clustering and sentiment tagging remove the preprocessing slog, and the visual output is clear enough to show a stakeholder directly. It stops short of being a general analytics platform — governance, connectors and scale features are thin outside Enterprise — and the free plan is a teaser more than a workspace. I rate it 3.7 overall: core functionality is strong within its niche (the rubric's heaviest component), value is reasonable at Pro pricing given unlimited rows and included AI, usability is a highlight, while ecosystem maturity and reliability signals — small team, single funding round of note — hold the score below 4. If your recurring job is 'make sense of 50,000 survey responses or support tickets this quarter,' this fits; if you need governed enterprise BI, it does not pretend to.

Sources

• https://1launch.eu/en/business/products/digital/graphext
• https://docs.graphext.com/faqs/pricing-and-plans
• https://graphext.com
• https://graphext.com/product
• https://graphext.com/about-us
• https://tech.eu/2023/06/13/from-photography-to-no-code-data-analytics-graphext-raises-46-million
• https://tractor.graphext.com/docs
• https://theorg.com/org/graphext
• https://platform.tracxn.com/a/d/company/55228a97e4b0d61ffba73011/graphext