DataLens frequently asked questions

DataLens is a platform covering the whole data lifecycle in one place: connecting sources, preparing and cleaning data, modelling it, orchestrating pipelines, analysing it, delivering the result and governing all of it. The navigation follows that order — Data, Prepare, Model, Orchestrate, Analyse, Deliver, Govern — because it is the order the work actually happens in. Instead of stitching six tools together and maintaining the seams, the same datasets flow through each stage.

Business users, data analysts, data architects, solution architects, data engineers, project teams and students learning data architecture. The interface does not require code for the core work, so an analyst can clean and chart a dataset, while an architect gets source and target models, mappings and lineage on the same data. Generated documents are grouped by persona for that reason — an executive brief and a technical handover describe the same system very differently.

No. Cleaning, transforms, joins, models, mappings and pipelines are all built through the interface. Where SQL is the clearer answer you can use it — database connections have a SQL mode, and models can be generated from SQL — but nothing requires it.

No. The screens are usable without formal training, and the lifecycle order gives you a sensible path through them. That said, the modelling and governance screens reward some understanding of what a fact table or a lineage graph is — which is part of why people use the product to learn those concepts on real data rather than on a textbook example.

A dataset. Upload a CSV or Excel file, or connect a database, REST API, file server, CRM or data lake — every other screen in the platform works on datasets, so getting one in is the only prerequisite. A file upload takes under a minute.

Yes, and that is the common case. Connect the systems that already exist, load the tables that matter, and build a source model of what is actually there. Nothing needs to be greenfield — most of the value in the modelling and lineage screens comes from documenting a system nobody has documented.

The Prepare screen offers 22 operations grouped as Filter Rows, Clean, Transform and Derive — dropping and filling nulls, removing duplicates, capping or removing outliers, casting types, splitting and renaming columns, standardising names and values, appending datasets, lookup enrichment, group-by aggregation, and derived columns including z-scores and percentile ranks. Every operation shows a live preview before it is applied.

A source model documenting the structure you have, and a target model describing the structure you want, kept deliberately separate so current state and intended state never get conflated. Tables can be marked as facts or dimensions, relationships recorded, execution order set, and joins designed visually with inner, left, right and outer join types.

Yes. The Keys & PII tab examines actual column values and reports recognised patterns — email addresses, phone numbers, dates, URLs, postcodes and ZIPs, currency, IP addresses and booleans. Detection is by content rather than by column name, so a personal-data column with an innocuous name is still found. It is a signal for a person to act on, not a compliance judgement.

Yes, on the Pro tier. Quality Rules turn an expectation into a check that runs rather than an assumption nobody tests, alongside a Dictionary recording each column's business meaning and expected format, and Column Naming for consistent naming across datasets. All three carry a Pro badge in the product.

Yes. The Pipeline Workbench assembles loads, cleaning, joins and shaping into a pipeline with an explicit execution order, so a process that lived in one person's memory runs as one unit. A step that fails stops the pipeline rather than letting downstream work run against incomplete input.

Yes. Simulate & Decide offers scenario modelling against a target metric, strategy comparison, impact analysis across metrics, forecasting with linear trend, moving average or exponential smoothing, and shock simulation with step, ramp, spike and sustained-pressure shapes. It runs against your own data, and it produces a structured estimate rather than a prediction.

AI-assisted features run against a model provider you configure yourself, using your own credentials — you choose which models the connection is allowed to use and which is the default. Usage is observable rather than opaque: AI Operations under Govern reports cost, tokens, failures, cache behaviour and estimate versus actual. Ari, the assistant, is available as a widget on every screen.

CSV, TSV, JSON, JSONL, Excel (XLSX) and Schema Only, which exports the structure without the rows. JSONL is the one to choose for large machine-consumed exports; Schema Only is more often what an architecture discussion actually needs.

Yes. A model exports as a PNG image, an SVG diagram, SQL DDL, a JSON model or a CSV schema listing. SVG for documents that get resized, PNG for slides, DDL when someone needs to recreate the structure.

Yes. The API Access section on the Publish screen exposes a dataset as an endpoint, with options to fetch it as JSON or CSV and a JavaScript fetch snippet to copy into an integration. The endpoint serves the dataset as DataLens holds it, not as a live query against your source system.

Bar, line, scatter and grouped bar charts with aggregation and a colour-by breakdown, histograms for distributions, a Pearson correlation matrix across numeric columns, and a pivot table taking a row field, an optional column field, a value field and an aggregation.

Yes — documents are generated from the models, mappings, pipelines and profiling in your workspace, grouped by who reads them. Types include a Data Architecture Document, Data Model Specification, Technical Handover Document, Pipeline Design Document, Source-to-Target Mapping, Pipeline Runbook, Data Governance Summary, Executive Dashboard Brief and Insight Summary Report. Each carries a READY or PARTIAL badge so you know its coverage before generating it.

On the Pro tier. Report exports to CSV and Excel are available on any tier; PDF and Word report export are entitlement-gated, and where a workspace is not entitled the app shows an upgrade card rather than a button that fails.