A gate register somebody will actually keep

Every vehicle that brings material onto a site passes one point. Every vehicle that takes anything away passes the same point. The gate is the only complete census of physical movement a project has, and it is usually kept in a ledger that nobody reads until there is an argument.

The reason gate registers fail is almost never discipline. It is that they were designed by somebody sitting down, for somebody who is standing up.

What the gate is uniquely good for#

Three things nothing else on a site can do.

It sees everything. Stores see what reaches the store. Accounts see what is invoiced. The gate sees the vehicle, including the ones that turn round, the ones that go to the wrong site, and the ones nobody was expecting.

It fixes a time. A delivery has a moment, and that moment is often the only way to distinguish two similar loads from the same vendor on the same day.

It is the last point before commitment. After the barrier, material is on your site and getting it off again is a conversation. Before it, refusal is cheap.

What kills a gate register#

Too many fields. Every additional field reduces the completeness of all the others, because there is a fixed amount of time while a driver waits. A register with eighteen columns is filled in with four.

Fields that require lookups. Purchase order number, item code, vendor code. The person at the gate does not have these and cannot get them. Anything that requires leaving the barrier to look something up will be guessed or left blank.

Writing in conditions that do not permit writing. Rain, dark, one hand on a torch. A paper register in a shed twenty metres from the barrier is filled in afterwards from memory, which is the same as not having one.

No feedback. Nobody at the gate ever finds out whether what they wrote was useful. A record that disappears into a system that never answers back is a record whose quality decays, because there is nothing to maintain it against.

The fields that earn their place#

For an inbound vehicle: date and time, vehicle number, what it is carrying in ordinary words, who it is from, the delivery challan number if there is one, and who received it.

For an outbound vehicle: date and time, vehicle number, what is leaving, on whose authority, and where it is going.

The outbound half is the one that is usually thin, and it is the half with the security value. A gate that records arrivals meticulously and departures casually is watching the wrong direction.

Two of those fields do most of the work. Vehicle number is the handle that connects a gate entry to a store receipt to an invoice, and it is the one thing the person at the barrier can always read. Time separates otherwise identical loads.

Photograph rather than transcribe#

The single largest improvement available to most sites is to stop transcribing and start capturing.

A photograph of the challan, taken at the barrier, contains every field on it, takes two seconds, requires no lookups, works in the dark with a flash, and cannot be misheard. The transcription can happen later, by somebody at a desk, or by a machine — and if it happens badly, the original is still there.

We built our own capture around exactly this, because it was already happening: site staff were photographing delivery slips into WhatsApp without being asked to, because it is the fastest thing a person holding a phone can do. The software's job was to accept that rather than to ask for something else. There is a longer piece on that principle — WhatsApp is the interface, whether you designed for it or not.

Two warnings from doing it.

Machine-read text from a photograph is a suggestion, not a record. It gets digits wrong, and a challan number with a wrong digit is worse than no challan number because it looks usable. A person confirms.

And a photograph without the vehicle number and time attached is much less useful than it seems, because it cannot be placed. Capture those two as fields even when everything else is an image.

The expected-arrivals list, and how it goes wrong#

A good gate screen shows what is expected today. It lets the guard say yes, you are on the list or wait, you are not.

Ours was built to do that, and it had a defect worth describing because it is the sort that hides for a long time.

The list refused any load whose money had not settled, which is sensible. But an order paid in stages never wrote the field that meant this is paid — when a payment has a schedule, the stages are the payment record. So a load that had been paid for to the last rupee never reached the list, and a fully-paid lorry arrived at the barrier unannounced.

That is the exact event the screen exists to prevent. It had been prevented in every case except the ones paid in the way the biggest orders are paid. The full account is a fact that had no owner.

The general warning: an expected-arrivals list is only as good as its completeness, and a list that silently omits a category trains the guard to treat absence as meaningless. Then the list stops doing anything at all. What the barrier needs to be told before a vehicle arrives, and what happens when it has nothing, is why a lorry surprises the gate.

Design for the person, not the record#

Two details that sound trivial and decide whether the whole thing works.

Very large tabular numerals. A vehicle number read at arm's length in sunlight, entered on a phone by somebody wearing gloves.

Language and pictograms rather than prose. Gate staff on Indian sites may read neither English nor the language the rest of the application is in. A screen built on symbols, big numbers and the local script works for everybody; a screen built on English labels works for the people who designed it.

We kept our gate screens deliberately untouched during a site-wide typography change for exactly this reason — they had never leaned on English typographic convention in the first place, so they had nothing to lose. That story is in capitals are a contrast, not a volume.

The short version#

The gate sees everything, once, and only briefly.

Cut the fields to what can be captured in the time a driver will wait. Prefer a photograph to a transcription. Make vehicle number and time non-negotiable. Watch the outbound direction as carefully as the inbound one.

And make sure whatever list the guard is checking against is actually complete, because an incomplete list is quickly treated as no list.