Guides · 9 min read · updated 2026-08-30
How to digitise your office in one day
Most “we will go paperless” projects die in week three, when somebody realises that scanning 40,000 pages is a logistics problem, not a software problem. Here is the run sheet that gets one rack — usually the one everyone complains about — fully indexed before the lights go off.
Why one day, and why one rack
One day is not a productivity boast; it is a scheduling constraint that changes behaviour. A project that runs for six weeks gets re-prioritised by the first busy Tuesday. A project that must finish on Saturday noon gets the two clerks, the extension boards and the standing fan that a slow project never gets.
One rack is the correct unit because a rack is what a human can reason about. A typical office almirah holds 35–60 files, roughly 2,500–4,000 pages. At a realistic 150 pages per person per hour for loose, undated, mixed-language paper, one rack with two scanners and one preparer is a 12-hour job — or a 7-hour job if you stop feeding it paper that does not belong in the scope.
The kit list, and what each item actually prevents
Nothing here needs a purchase order beyond the scanner, and the scanner is optional on day one:
- Two phones with 12 MP cameras, or one ADF scanner. Two phones beat one scanner for throughput; a scanner beats two phones for duplex and long receipts.
- A fixed light source and a dark mat. Ceiling tubes create a hot spot across half the page, which is exactly where OCR loses a date. A₹400 table lamp on a black A4 card fixes more accuracy than any software.
- Staple remover, bulldog clips, a roller. A creased fold line splits a number into two numbers. Flattening takes four seconds and saves a correction later.
- A box for “not today”. Anything blank, duplicate, or unreadable goes in the box. This one rule is the difference between finishing and not.
- A printed log sheet. Batch number, file range, operator, start and end time. When a batch turns out to be missing, the log is how you find it in two minutes instead of two days.
The twelve hours
Assume three people: a preparer, two scan-and-check operators. Rotate the boring job every two hours — accuracy falls off a cliff after hour three of page feeding.
| Time | What happens | Output |
|---|---|---|
| 08:00–08:45 | Scope written down: which drawer, which year range, what is excluded. Log sheet printed. | Scope note |
| 08:45–09:30 | Light, mat, table and device battery set up. Ten pages run as a test, reviewed together. | Calibration set |
| 09:30–13:00 | Block 1. Preparer removes staples and sorts; operators shoot and confirm each page is legible before turning it. | ~1,600 pages |
| 13:00–13:45 | Break. Uploads run in the background; the queue drains while you eat. | — |
| 13:45–17:00 | Block 2, plus the first review queue appears: fix dates and names now, not “later”. | ~1,500 pages |
| 17:00–18:00 | Originals go back in the rack in index order — with the file number now on the spine. This is where the physical system improves. | Ordered rack |
| 18:00–19:30 | Exception handling: the “not today” box gets a one-line note per item, and the missing-page list gets written. | Gap list |
| 19:30–20:00 | Ten queries run against the day’s index, out loud, in front of whoever stayed. This is the moment people believe it. | Sign-off |
Stop naming files; start dating them
The instinct is to build a folder tree: 2019 / Sale deeds / Khordha / Patel. Resist it, because a folder tree encodes one decision and you will need four others later (by party, by survey number, by lawyer, by expiry). What actually identifies a record is its dates and its parties — so write those two things down as you scan and let the sorting happen afterwards. If your tool sorts only by filename, you have bought a storage unit, not an archive.
The naming rule that survives ten years is boring and mechanical: YYYY-MM-DD_place_type_number (eg 1998-03-12_khordha_saledeed_241-14). Date first, because it sorts. Never put a person’s name first, because they may have married, misspelt or died.
The quality check that takes ninety seconds per batch
- Pick 5 pages at random, not the first 5. The first five are always the ones you practised on.
- Read the date, the amount and the name on the screen, then on the paper. Those three fields cause 80% of the harm when wrong.
- Count pages. A 12-page deed indexed as 11 means page 12 will be missing on the day it matters.
- Search one thing you know exists. If it does not come back, the index — not the scan — is the problem.
Track three numbers on the log sheet and you will know within a week whether the project is healthy: pages per person-hour, pages flagged for correction per hundred, and corrections closed per day. If the correction queue grows faster than it drains, stop scanning and fix the light, the angle, or the paper.
The five ways this fails
- Scanning without deciding what a “document” is. A 40-page file indexed as one record is unfindable; indexed as 40 pages with a parent is perfect.
- Trusting auto-detected dates on carbon copies. Faded 9s read as 4s. Flag low confidence, do not guess.
- Ignoring the back of the page. The endorsement, the receipt stamp and the noting are on the reverse in a huge share of Indian records.
- One hero doing it after their actual job. Two people for a day beats one person for a month.
- No plan for paper that arrives tomorrow. Digitise the backlog and the flood starts again. Fix the intake desk first — one tray, one photo, five seconds.
When you need more than a phone and a spreadsheet
A rack of current-year invoices is a one-day job with no software beyond a scanner app. It becomes a tooling problem when any of these are true: the paper spans decades, multiple people must find things, the record may be challenged, or the contents are personal data you are obliged to protect. That is exactly the point where encryption and a date-based index start paying for the day you already spent.
If that is your situation, how Chronolipi orders records by date is the honest place to start reading — and the savings calculator tells you whether one rack a week is worth the subscription, using your own hourly cost rather than our optimism.
Questions this raises
How many pages per hour is realistic?
For loose, mixed, undated paper with two people: 120–180 pages per person-hour including the legibility check. A batch that has already been unstapled and sorted runs at 250–300. Anyone quoting 1,000 pages an hour is describing a machine feeding prepped A4, not your almirah.
Should we scan at 300 or 600 dpi?
For a phone: irrelevant — shoot at the highest resolution your camera sustains and let the pipeline downsample. For a scanner: 300 dpi for typed pages, 400–600 for handwriting, faded carbon copies or anything with a stamp over text. Greyscale, not colour, unless a seal or a signature really matters.
What about the original paper — can we destroy it after scanning?
Usually no, not on your own authority. Many record classes have statutory retention periods and destruction rules, and the digitisation may not be legally recognised yet. Digitise as a copy with integrity proof first; treat destruction as a separate, documented decision.
Chronolipi publishes what we measure. Nothing here is legal advice; record classes have their own statutes, and your own counsel should read your inventory.
Keep reading
Evidence & law
Why your court records need a digital backup
A missing file does not just delay a hearing. It collapses evidence, extends undertrials, and ends a litigant’s case on procedure. Every lawyer readin…
Compliance
The DPDP Act and your paper records
Everyone read the DPDP Act as a software problem. It is mostly a storage problem — and in India, most of the personal data you are responsible for is …