A bill is only worth what it can survive.

The whole system: the mobile capture application, the OCR pipeline that reads a meter face and scores its own confidence, the approval workflow that separates capture from validation, the billing engine and its single published formula, the arrears and payment tracking, the reconciliation against the utility bill, and the month-close that freezes the books. He leads it as Head of Product & Platforms at BijliBachao.pk.

That the deliverable is not the invoice. Any spreadsheet can produce an invoice; what a building manager actually lacks is the ability to prove one. So the system was built around evidence — a photograph retained behind every reading, a formula shown to the person being billed, and a month that can be frozen — and the invoice falls out of it. Getting that ordering right is the product decision; everything downstream is ordinary engineering.

Because a reading that the same person can both take and approve is not evidence of anything. An operator in TenantBill can capture and nothing else — no rates, no billing, no approval — and a manager reviews each reading against its own photograph and last month’s figure, requesting an audited retake rather than editing a number silently. Without that separation the photographs would be decoration.

Because the raw captures contained real tenant names, shop trading names, meter serial numbers, the operator’s email address and, on one screen, the property owner’s monthly profit and loss. The tenants are small shopkeepers who never agreed to appear on a portfolio. The screens are shown because they are real and running; the people in them are not the evidence, and neither is the client’s margin.

Very few, and that is recent. Read-only queries against the production database on 25 August 2026 replaced the page’s earlier figures, several of which were wrong — the scale twice over, and the optical-character-recognition accuracy twice over. Four outcome percentages that had circulated in marketing material were traced to a document listing them as typical results for a hypothetical property, with no baseline and no measurement anywhere, and were dropped rather than softened. Two billing-cycle claims failed the same check and were dropped too. No speed figure is stated for the automatic reading because no timing telemetry is stored, and no offline-capture count is given because synced readings cannot be distinguished after the fact. What remains is measured.

A rooftop with a customer. BijliBachao sells a service it calls Solar Share System — described in its own catalogue as a first in Pakistan — which is commercial solar EPC bundled with TenantBill: the owner installs the array and each tenant gets a photo-verified bill crediting solar units against grid units. A landlord cannot sell rooftop generation to their own tenants on the strength of a spreadsheet, because the tenants have to accept the bill. The billing layer is the precondition for that service existing, not a feature of it.

The exact category — tenant billing software — was worth roughly USD 455 to 460 million in 2025 and is growing 8 to 9 per cent a year, with Mordor Intelligence and 360iResearch landing independently within about one per cent of each other. It sits inside a sub-metering solutions layer of around USD 3.8 billion and an electric sub-meter hardware market an order of magnitude larger again. The meters are a big market; the software that makes their readings defensible is the small, fast, difficult part. Asia-Pacific is the fastest-growing region for the category at about 7.8 per cent a year, and the driver analysts name for its commercial segment is tenant billing itself. Every figure, its confidence grade, and what was searched for and not found is recorded in the repository at docs/research/tenantbill-market.md.

Industry estimates put revenue leakage in manually billed buildings at 3 to 8 per cent of total energy cost — consumption that happens but is never billed, through meter-reading gaps, calculation errors, and tenants underpaying without being challenged. That figure comes from a vendor of billing software rather than from independent research, so it is directional and this page attributes it wherever it appears. What makes it worth citing is that the causes named alongside it read as a specification for what TenantBill does: human error in manual reading, spreadsheets missing edge cases, no timestamped record to show a tenant who disputes a bill, and common-area energy as the single biggest source of argument.

The shape is general: field staff capturing data on a phone in places with no signal, an automated read of that capture that has to declare its own confidence, an approval path where the capturer is not the approver, a computation that has to be explainable to the person it charges, and a periodic close that makes the record immutable. That is stock audit, clinical data capture, insurance inspection and field logistics. The electricity is incidental.