Service
Data work at scale: fast, accurate, verifiable
Every business runs on data someone has to type, check, clean and maintain. We industrialise it: SOPs, double-key QA, throughput dashboards, and per-record prices that make backlogs disappear.
Overview
Data services at Aadhya
The data desk handles the full lifecycle: entry (catalogues, documents, forms, research capture), cleaning (dedupe, standardise, validate) and enrichment (append emails, phones, firmographics). Since 2010 this has been Aadhya’s core muscle — the same QA discipline now powers our support and sales desks.
Everything is measured: items per hour, error rate from daily samples, backlog age. You get a dashboard, not a promise. One-off backlogs are fixed-quoted; continuous queues run per-seat or per-thousand-records.
At a glance
- Work types
- Entry, digitisation, catalogue, research, labelling, CRM upkeep
- Accuracy
- 99%+ stabilised, double-key ramp
- Scale
- 1 seat to 20-seat surge teams
- Turnaround
- Overnight batches (IST advantage)
- Formats
- Any: PDF→structured, image→text, web→sheet
- Pricing
- Per record / per seat / fixed project
What you get
Built for accountability, not activity
- Double-key accuracy: critical fields entered twice by different operators and machine-compared — the boring method that actually hits 99.9%.
- Overnight turnarounds: files dropped at your 6pm are done by your 8am, because your night is our workday.
- Format agnosticism: scanned PDFs, handwriting, legacy exports, web sources — structured into whatever your system ingests.
- Catalogue expertise: Shopify/Amazon product data entry with attribute discipline, image association and category mapping.
- Research capture: structured web research (competitor lists, directory building, contact sourcing) with source URLs on every row.
How it starts
- Sample
- Send 50–100 representative records; we return them done, with a fixed quote and an SOP draft.
- Ramp
- Double-key mode until measured accuracy clears target on your data.
- Run
- Batch or continuous, dashboarded daily.
- Audit
- You spot-check anytime; disputed records reworked free.
Pricing shape
How this service is priced
Exact rates live on the pricing page — published, because serious buyers filter on it.
- Per 1,000 records
- For countable, uniform work. Most projects land here.
- Per seat
- Continuous queues and mixed work.
- Fixed project
- Backlogs and migrations, quoted from the sample.
Fit check
Built for some teams. Wrong for others.
Honest scoping saves both sides a month. This desk fits when:
- Teams with backlogs measured in thousands of records and dread
- Businesses whose catalogues, CRMs or documents grow faster than anyone maintains them
- Analysts spending research hours on capture instead of analysis
Probably the wrong desk if: Datasets under a few hundred rows — an intern afternoon beats a vendor; Work requiring domain judgment on every record with no writable rules; Anyone wanting accuracy promises without a sample batch first.
The Aadhya way
Data work rewards humility: every dataset lies about itself until you process the first hundred rows. That is why everything here starts with a free sample — it prices the real work, exposes the edge cases, and replaces negotiation with evidence before a single invoice exists.
Questions buyers ask
Simple structured entry from ~$4–$8 per 1,000 fields; complex document extraction more. Seats from ~$550/month. The free sample batch produces your exact fixed quote — see pricing.
Double-key entry on critical fields (two operators, machine comparison), field-validation rules, and daily QA sampling with error-rate tracking. Accuracy is measured on your data during ramp, not asserted from a brochure.
Overnight for batch work up to a few thousand records — the IST timezone means your end-of-day handoff is complete before your morning. Bigger volumes get a scheduled cadence you set.
Yes — human keying with OCR assist where useful, confidence flags on illegible fields rather than guesses. Ambiguity rules are agreed in the SOP.
A specialty: SKU creation, attribute completion, image association, variant matrices and category mapping for Shopify, WooCommerce, Amazon and Magento — with the attribute discipline that stops filter pages breaking.
Office-only workstations, least-privilege access to your systems, NDAs, no local copies beyond the working batch, and deletion on completion certified in writing if you need it.
Practically, ~$300 of work — below that the setup overhead dominates. The free 50–100-record sample has no minimum at all.
Most clients start with a backlog and stay for the maintenance queue: daily catalogue updates, CRM hygiene, weekly research refreshes. Continuous work runs per-seat with the same dashboards.
Yes: classification, annotation, transcript cleanup and RLHF-style preference labelling with inter-annotator agreement tracking. The QA machinery transfers directly.
Yes — reconciliation queues are a desk specialty: match rules written with you, discrepancies classified by cause, and a resolution log your auditors can read. The deliverable is not just matched records; it is the documented reason the systems disagreed.
Automate what is automatable — we will tell you which steps those are, and script them. What remains is judgment, messy sources and exceptions: exactly the work where a QA-disciplined human team beats a hallucinating parser. Most clients end up with a hybrid, cheaper than either extreme.
Next step
Start with a pilot, not a contract.
Describe the queue, the list or the workload. You get a written pilot plan and a fixed quote within 48 hours — and the pilot itself proves us before you commit to anything longer.