How to Build an Internal Admin Panel in 2-3 Weeks

Most Indian SMBs run their back office on a spreadsheet, a WhatsApp group and one person who knows the process. The internal admin panel keeps getting pushed back because it never feels urgent next to customer-facing work. The cost is quiet: ops re-keys the same order three times, and nobody can say who approved last week's refund. Here is the order of work that gets a usable panel live in weeks instead of quarters.

Scope it around three workflows, not your whole schema

An admin panel should launch with the two or three manual workflows that burn the most hours today, not with a screen for every table in your database. Spend a week sitting with operations and log every task they do by hand — anything involving copy-paste between tools, a WhatsApp approval, or a screenshot pasted into a spreadsheet is a workflow worth building. Common first candidates are order corrections, refund and credit-note approvals, vendor or KYC document review, stock adjustments and GST invoice fixes. Everything else can wait: bulk imports, report builders, one-off exports. If a screen would be used less than once a week, it belongs in release two. Putting it in release one is how a six-week project becomes a six-month one.

Pick boring building blocks and buy the scaffolding

Choose a stack your team can already debug at 11pm, and generate rather than hand-write the parts that are not your differentiator. In practice that means Postgres, one backend framework your developers know well, and a CRUD-friendly admin layer on top: Django admin, Laravel Filament, or React Admin and Refine against your existing API. Reuse Google Workspace or your existing SSO for login instead of writing a password system. Skip the bespoke design system — a component library is fine for internal users who will live in this thing daily for years. If you don't have a front-end bench, custom web application development is usually better outsourced than stretched across a product team that is already behind.

Model roles and permissions before you build screens

Roles should live in the database as data, not as if-statements scattered through views. Most SMB panels need three or four: support (read plus narrow writes), operations manager (approvals), finance (money-touching actions and exports), and one admin who can grant access. Define permissions as named actions — refund.approve, invoice.edit, customer.export — and map roles to actions in a table, then enforce the check in the API layer rather than only in the UI. Keeping the mapping in the database means you can change who approves refunds without a redeploy. This is roughly half a day of work upfront, and it prevents months of arguments after the first dispute.

Build the audit log on day one

An audit log is far cheaper to build before the first write endpoint exists than after, and it is usually what gets finance and compliance to sign off. Log the actor, the action, the entity touched, a before-and-after diff, the timestamp and the IP address, into an append-only table nobody edits through the panel. That one table answers DPDP Act data-access requests, settles arguments about who changed a price, and gives you a trail when GST reconciliations don't match. In our experience, retrofitting an audit log across 40 existing endpoints takes two to three weeks; writing the same log entry as each endpoint is built takes two to three days.

Budget 15 working days for a first usable release

A realistic first release is about 15 working days for one competent full-stack developer, assuming the workflows are scoped and access to existing data is ready. A typical split:

• Days 1-3: workflow interviews, data model review, permission map, one-page architecture note.
• Days 4-6: SSO, roles, the audit table, base layout and navigation.
• Days 7-10: read-only list and detail screens for the three priority workflows — search and filters first, no write actions yet.
• Days 11-13: write actions and approvals, each with a confirmation step and an audit entry.
• Days 14-15: CSV exports finance will actually accept, UAT with two ops staff, and a rollback plan.

Add 30-50% if the source data sits across spreadsheets, a legacy ERP and a WhatsApp inbox, because cleaning and joining it is usually the real work. A short architecture review before day one often pays for itself there.

Where to start

Start by logging a week of manual ops tasks and circling the three that cost the most hours — that list is your v1 scope, and everything else is a distraction until it ships. If you would rather not pull your product team off the roadmap to build it, Maysan Labs builds internal tools and admin panels for Indian SMBs, typically three to five weeks end to end including data cleanup.