Digitising a mandi business without losing the floor
The market starts at four in the morning, the staff are not office workers, and the network is unreliable. Software here has to survive conditions that break most business systems.
Engineering·22 August 2026·6 min read
Most business software assumes an office: a desk, a chair, a stable connection, and staff who spend their day in front of a screen. A wholesale market is the opposite of all four.
That is why so many mandi digitisation attempts stall, and it is why the design constraints matter more here than the feature list.
Move in this order
Trying to start at reporting is the classic error. Reports built on incomplete entry teach everyone that the system is wrong.
- Arrivals and lots. The spine of everything else, and easy to record at the point of unloading.
- Auction sale entry. Buyer and rate against a pre-created lot, two taps.
- Grower statements with deductions, which is the first thing that visibly improves relationships.
- Buyer receivables, which is where the money risk sits.
- Reporting and the daily close, once the data going in is reliable.
Design for the floor, not for the office
Big touch targets, short forms, defaults that match what usually happens, and no typing where a selection will do. Urdu labels where the staff read Urdu. Numbers formatted the way the market says them.
The measure is simple: can a lot sale be recorded in the time it takes to write it in the register. If not, it will not be recorded during the auction, and everything downstream reverts to slips.
Assume the connection will fail
In a covered market at five in the morning, mobile data is unreliable and shared wifi is worse.
So entry has to keep working when the network does not, and the reconciliation afterwards has to be automatic rather than a manual re-entry job. Any system that stops when connectivity does will be abandoned within a fortnight, correctly.
Run the register in parallel, briefly and deliberately
Two weeks, with a date agreed in advance for stopping. During that window, compare the two at the end of each day and investigate every difference, because each one is either a data entry problem you can fix or a workflow the software has not modelled.
Open ended parallel running is the failure mode. The register is faster for the person holding it, so if it never stops, it never stops being the real book.
Train the person who will teach the others
In most agent businesses there is one person who effectively runs the floor. Train them properly and thoroughly rather than everyone superficially. They will train the rest in the way the market actually communicates, which is by demonstration during a working morning.
Sabzi Mandi Software is built for these conditions: fast entry for arrivals, lots and auction sales, commission and deduction rules, grower statements, receivables and a daily close, deployed on premise with cloud backup so the floor does not depend on a perfect connection.
Digitising a mandi is not an IT project. It is a change to the busiest ninety minutes of the day, and it succeeds only if the software is faster than the register at exactly that moment.
Frequently asked questions
Will staff who are used to a register actually use software?
Yes, when the screens are in the language they think in, the fields are the ones they already write, and entry takes fewer actions than the register. Resistance in this environment is almost always a design problem, not an attitude problem. If entering a lot sale takes more than a few taps, the paper wins and it deserves to.
What happens when the internet drops at the auction?
That has to be assumed, not handled as an exception. In practice this means entry on a device that keeps working locally and reconciles when the connection returns, and a printed fallback for the few minutes where nothing electronic is available.
How long does it take to switch?
Plan a season, not a week. Start with arrivals and lot sales, add ledgers and statements once those are habit, and keep the register in parallel for a defined period. In a business where a day's error is a real loss, parallel running is worth the duplication for a short, agreed window.