Methodology
How the tools on this site work, how the guides are produced and checked, where the data comes from, and what the tools cannot do.
How each tool class works
Every tool on this site runs locally. The cleaning tools (CSV Cleaner, Duplicate Remover, JSON to CSV, Email Extractor, Data Health Score), the Data Workspace, and Verify use a deterministic, rule-based engine in your browser: deduplication, whitespace and case fixes, email, phone, date and name standardization, and formula neutralization, applied in a fixed order and recorded in a change log.
The writing and scoring tools (Cold Email Generator, ICP Generator, Lead Quality Analyzer) also run in your browser, generating drafts and scores from built-in templates and rules. No tool on this site uses an AI model, and nothing you type or upload is sent to a model service or stored on our servers.
The command-line tool and MCP server run the same engine locally on your own machine, with no network calls. Paid cleaning services are different: you email us a file and a person cleans and checks it by hand, under our Service Standards.
Data sources
The browser tools process only the files and text you supply. The guides use invented sample files (see the disclosure below). The demand snapshot documents its own method on its page: public marketplace listings, viewed on the date stated there, presented as a snapshot rather than a market average. There are no live data lookups: email checks are format-only, and phone standardization does not confirm that a number is in service.
Update frequency
When the cleaning engine changes, the tools, these method notes, and the worked-example guides are regenerated together, so a guide never describes behaviour the current tool does not have. Prices change only on the service pages themselves; the demand snapshot is dated on its page and is not silently refreshed.
Simulated-data disclosure
Every sample file, worked example, practice task, and the sample cleaning report on this site uses fictional, invented data — no real person, company, or customer file appears in any example. Samples are labeled as fictional where they are offered. Each guide shows its sample file, the before and after tables produced from it, and the cleaning report, so any reader can re-run the same file in the Data Workspace and compare.
Limitations
- Email checks cover format only; they do not confirm a mailbox exists.
- Ambiguous dates are left unchanged and listed for a person; two-digit years are untouched unless you opt in, and conflicting duplicate records are kept and listed rather than silently merged.
- Drafts, profiles, and scores from the writing and scoring tools are starting frameworks built from your inputs — they are not verified facts about any person or company.
- There is no hosted API; the agent tooling runs locally.
Error reporting
If you find a mistake in a tool result, a guide, or this page, email hello@smartdataai.ai. Corrections are made to the page itself and, for guides, by regenerating the worked example with the corrected engine.
Who operates SmartDataAI
SmartDataAI is a Frontier Oak Property Holdings project, operated by Frontier Oak Property Holdings LLC. It is run by a small team that works on data cleaning and lead operations; every paid order is handled and checked by a member of that team.