Agreed cleansing recipe
Column-level allowances for trims, coercion, duplicate handling, and missing-value policy recorded up front.
Skip to content
Capability overview — spreadsheets & CSV workflows
Data Cleaning toolkit
Tabular data quality & preprocessing
Teams waste hours normalizing sloppy exports before reporting or downstream models. A scoped preprocessing pass standardizes parses, trims duplicate rows, aligns categories where configured, prepares date/numeric coercion, fills or drops missing values per agreed rules, and outputs files plus a shareable HTML run summary—ideal for pilots and internal tooling, not turnkey SaaS.
Static overview only. Use the Live Streamlit demo button for the labs-hosted sandbox.
Python / pandas
Streamlit
CSV · Excel · Parquet · JSON
pytest
Limitations & scope
- Not SaaS provisioning, not multi-tenant managed MDM catalogs, not magic autonomous AI wiping columns.
- Pilot scope agrees file types, thresholds, allowable drops/fills, and review gates—beyond that needs new scope.
- Large files need sampling or server-side ingestion projects; defaults target practical internal batches.
- English-language UI today; multilingual labels require agreed extensions.
Operational pain. Finance, RevOps, and CS teams routinely export spreadsheets and CSV snapshots that contradict each other: duplicated customer rows, collapsing date formats, free-text regions masquerading as categories, blanks where ERP rules expect numbers. Manual cleanup slows reporting cycles and hides mistakes until dashboards look “off.”
VahdetLabs scopes short preprocessing pilots that bring those exports to a repeatable baseline before BI refreshes or internal tooling handoffs—not a flashy SaaS promise, just disciplined tables.
What a pilot produces
Handoff-ready files
CSV / Parquet / JSON exports whichever downstream stack prefers—with identical transforms each run.
HTML run summary
Sharable synopsis of counts, deltas, warnings—anchors internal reviews without unlocking production databases.
Scope boundaries
Explicit what is in/out—no stealth expansion into unmanaged MDM catalogs or enterprise workflow orchestrators without renegotiating.
Standardization scope
-
Structure & duplicates
Optional business keys constrain duplicate scans; merges from multiple uploads keep provenance for reconciliation.
-
Types & formats
Dates and numerics coerce only when instructed; outliers surface as warnings—not silent deletions unless that rule is negotiated.
-
Categories & whitespace
Case normalization and trimming eliminate “West ” vs “west” drift feeding regional charts.
-
Missing data policy
Stakeholders choose between leaving gaps, pruning rows with explicit rationale, or filling with audited constants tied to KPI definitions.
Collaboration rhythm
- Discover: sample files + KPI intent + risk notes (regulated fields stay client-side).
- Prototype: show detection output, reconcile edge cases together, freeze the cleansing checklist.
- Operationalize: deliver repeatable exports plus summary HTML; train owners on rerunning or scheduling follow-on work separately.
Commercial guardrails
- Not an always-on multitenant SaaS SKU—pilots intentionally bound time, datasets, environments.
- Not an enterprise master-data catalogue; semantic merges across unrelated systems belong to larger programs.
- Large-file ingestion, VPC placement, SSO, ticketing hooks are follow-on scopes once table discipline proves value.