Questions we’re asked.

Tabular data, semi-structured data (JSON, XML, HTML, logs), free text, documents and scans, images and audio. Most projects mix several of these.

We check it against its source, related fields and other sources. Confirmed errors are corrected; real events are kept and annotated; anything we cannot settle is flagged to you with the evidence. We don’t delete an outlier simply because it is unusual.

First we look for it: other sources, related records, the original documents. If it can be derived reliably, we derive it and show the method. If it is genuinely missing, it stays empty and is marked with the reason. Our practice is not to invent values.

No. AI agents do the reading, checking and proposing. People review and approve the changes, individually or as rules, before they are applied to your data.

Not necessarily. Many projects start from an export or a sample. Where data cannot leave your environment, we can work inside it.

Often, yes. We can redact or pseudonymise personal data early in the process and work under an NDA. Requirements for regulated data (for example health or financial records) are agreed case by case. See our data handling page.

The curated data, a change log, a data dictionary, a quality report, and the rules used, so the work can be repeated.

Engagements start with a fixed-fee Data Quality Audit. Curation projects are quoted after the audit, once the scope is clear. Monitoring is billed monthly.

It depends on the volume and condition of the data. The audit gives you an estimate before you commit to anything larger.

Yes. The checks and transformations are delivered with the data and documented, so your team can run them on new data.

CSV, Parquet, JSON, Excel, SQL tables, or direct loads into your warehouse.

Yes. ClarionThinking is a CuratedData product: a free investment-research platform built on market data curated with the same method.