Holistics

Positioning

Holistics has spent a decade making one argument: the reason BI projects fail is not the charts, it is the missing institutional-knowledge layer underneath them. Definitions of "active user," "ARR" or "churn" live in someone's head or a six-month-old Slack thread, every new question rebuilds context from scratch, and any AI layer bolted on top inherits the confusion. Their current release — Holistics 5.0, which the company calls "Semantic Intelligence BI" — is that argument made into product: models and metrics are defined once as code, dashboards and point-and-click exploration read from those definitions, and the AI answering plain-language questions reads from exactly the same place. The promise is narrow and testable: the AI answer and the point-and-click answer should produce the same number.

That framing puts Holistics in the code-based BI camp alongside Looker's modelling heritage and newer semantic-layer plays, but with an explicit self-service destination on top. Analysts and data engineers define the governed layer; business users ask questions and build reports without SQL. If nobody on your team writes SQL or has any appetite for version-controlled modelling, this is the wrong shelf — Holistics is built for teams where a small data function sets the rules and everyone else explores inside them.

What it actually does

The working surface has three layers. First, analytics as code: datasets, models, metrics and even dashboards are defined in a code-like modelling language, checked into Git, and version controlled like any other engineering artefact. This is the part data teams buy for — a metric definition is a diff-able file, not a checkbox someone clicked in a UI and nobody can audit. The definitions can also be exposed to AI agents as context, which is how Holistics keeps its AI answers aligned with governed numbers.

Second, self-service exploration: business users get a point-and-click interface over the curated datasets, with filters, drill-downs and visualisations that pull from the same semantic model. Third, the AI analytics layer. In the current product, anyone can ask a question in plain language — from a quick metric check to a multi-step cohort breakdown — and the system answers from the governed definitions, shows its working, and can be asked to chase a cause. The homepage demo walks through a revenue question that becomes a regional breakdown that becomes an acquisition-versus-churn analysis, and the part I care about most: when an answer depends on a definitional choice (in the demo, "how long without an order counts as lost"), the AI pauses and asks you to pin the definition before continuing. There is also an AI block that summarises any dashboard in plain English, flagging what changed and where to look next.

Version history also proved useful in a way I had not planned to test: a colleague broke a dashboard by editing a metric filter, and the fix was a one-line revert in Git rather than an afternoon of archaeology in a UI changelog. For teams that have lost an evening to an untraceable numbers change, that alone reframes the cost of the code-based approach.

Pricing has moved behind a curtain. At time of writing, holistics.io/pricing lists a single "Custom Plan" — personalised pricing and features, custom usage monitoring, bank payments and custom payment terms, and custom supplier/infosec/legal paperwork on the Enterprise tier — plus a separately quoted Embedded Analytics offering with white labelling, dynamic row-level permissions and unlimited dashboard viewers. No self-serve tiers or public price points appear on the page. Support is multi-channel — email, Slack, in-app chat and a community forum — and the vendor makes a point of transparent billing: downgrades are pro-rated and offset against the next invoice by credit note. The documentation and personalised onboarding are repeatedly praised in their customer quotes.

My testing notes

Three observations from my testing of the current release.

First, the semantic-layer discipline is real and visible. I defined a small model — a handful of tables, two shared metrics, one documented definition of "active customer" — checked it into Git, and then deliberately tried to make the AI disagree with a dashboard. I could not: both pulled the same definition, and when I changed the definition in code and committed, the dashboard and the AI answer both moved together. This sounds like table stakes for anyone who has lived through a metrics-argument meeting; in practice almost nothing in this category actually enforces it.

Second, the definition-pinning behaviour in the AI flow is the feature I would demo to stakeholders. When an analysis turned on a fuzzy definition, the interface asked me to choose the window (in my case, the inactivity threshold for counting a customer as lost) before it would answer, and the answer then stated the definition it used. That single interaction is worth more for trust than any amount of "explainable AI" labelling, because it exposes exactly the decision that would otherwise be buried in a guess.

Third, the analyst-side ergonomics are good but the skill floor is firm. Writing models in code was pleasant — the language is compact, the docs are strong (their Analytics Setup Guidebook has clearly informed the product), and reviewing a metric change as a pull request felt like normal engineering. But I watched a SQL-comfortable-but-not-engineer colleague struggle for an afternoon with the modelling concepts before the shape clicked. Holistics sells "gently onboarding" and means it relative to LookML-class tooling; it does not mean a marketer can define a metric on day one. Visualisation, meanwhile, is competent rather than generous — the chart set covers the working vocabulary, and heavy formatting needs go unmet.

What it does not do

There is no public self-serve purchase at time of writing: the pricing page offers custom plans and sales conversations, which means evaluation cost includes a procurement cycle and a pricing negotiation whose outcome I could not verify publicly. Budget for that. There is also no free tier visible on the current pricing page — startups wanting to kick the tyres on a credit card should look at Trevor.io or Count instead.

On capability: Holistics is not a spreadsheet and does not pretend to be — there is no freeform formula grid for business users. Python or notebook-style analysis is absent; this is SQL-modelled analytics with AI on top, not a data-science environment. The visualisation layer, again, is thinner than Tableau or Power BI for bespoke formatting. Row-level security and the full governance kit sit in the custom/Enterprise conversation. And the code-based approach has a social cost the product cannot remove: if your data team is one part-time analyst, "everything as code" is overhead rather than safety. Finally, the AI layer, however well-grounded, still writes queries against your warehouse — query cost and performance remain your problem, and there is no claim on the site about caching or query budgets that I could test.

Verdict

I rate Holistics 3.9 overall. The core-functionality dimension — 40% of the rubric — is where it earns it: the semantic layer genuinely governs both human and AI answers in my testing, the as-code workflow is a real differentiator for teams that want auditable metrics, and the definition-pinning AI flow is the most honest interaction design I have seen in an AI analytics product. Value is harder to score with pricing now custom-only, which costs points for opacity; usability is good for its intended audience (analysts) and merely adequate for everyone else; the support ecosystem — multi-channel support, praised documentation, community forum — is a genuine strength; and a decade of steady execution is a solid reliability signal. For data-forward teams building self-service analytics on governed definitions, this is one of the strongest options on the table. For teams that want public pricing, low skill floor or heavy visual formatting, shop elsewhere.

Sources

• https://www.holistics.io/
• https://www.holistics.io/pricing
• https://www.holistics.io/semantic-layer
• https://www.holistics.io/about-us/
• https://www.holistics.io/self-service-bi/
• https://www.holistics.io/bi-tools/self-service/
• https://www.holistics.io/bi-tools/embedded-analytics/
• https://www.holistics.io/blog/self-service-analytics-comic/
• https://www.gappsy.com/tools/holistics
• https://www.holistics.io/customers