Software house, Tebet, South Jakarta

From commit to production in 15 minutes 29 seconds

Scroll and one release runs above: lint, unit tests, build, integration tests, staging, production. The first attempt fails at minute 3:12 and a fix commit follows.

Book an assessment See the services

A release every two weeksThe code is yoursNo night cover

Four numbers you can check

15:29 commit to production, the full pipeline

23 internal systems shipped since 2017

16 people, nothing subcontracted

7 months, the average system from nothing

Five shapes of work

And a list in the questions below of what we do not do at all.

A four week assessment

Two people read the database, the old code, and sit beside the people who use it every day. It ends in one thirty page document: a plan, a cost estimate with a range, and a list of things that should not be built at all. The fee is credited in full if the work continues.

An internal system built from nothing

Warehouse, claims, payroll, expense approval. Seven months on average with three or four people, and the first release to production lands in month two, not in the last month. After that, a release every two weeks.

Moving an old system

Usually off a shared Excel file or a Visual Basic application whose author retired seven years ago. Old and new run side by side for at least six weeks, with both sets of numbers compared daily, before the old one is switched off.

Integrations and data interfaces

Connecting a system to Accurate, SAP Business One, a fingerprint clock, or a bank. When the interface documentation is gone we read the traffic and write the specification again, and that alone usually takes two to three weeks.

Maintenance, working hours only

Fixes within one working day, monthly security updates, and one person reachable between 08.00 and 18.00. We do not offer round the clock cover, because sixteen people cannot promise it without lying.

Three systems, with the numbers before and after

Warehouse system, pharmaceutical distributor, Bekasi

Eleven warehouses, 240 users, and a stock count that took three full days with all deliveries halted. Records lived in four Excel files copied to each other by email every night.

The stock count fell to six hours and now runs without stopping deliveries. The gap between records and physical stock fell from 2.8 percent to 0.4 percent. Nine months, four people, and two warehouses refused to use it for the first six weeks.

20249 monthsPostgreSQL and React

Claims system, general insurer, Jakarta

Nineteen hundred claims a day arriving by email and re keyed by hand into two systems that did not speak to each other.

Claim handling fell from 4.2 days to 1.1 days. The re keying disappeared entirely. Seven months, three people, and one risk scoring module we recommended against building because the data was not there yet.

20257 monthsGo and PostgreSQL

Attendance and payroll, components plant, Karawang

Two thousand four hundred staff, three shifts, and a payroll calculated over two full days each month by four people with calculators and paper overtime sheets.

Payroll now finishes in four hours. Overtime disputes fell from 180 to 11 a month. Six months, three people, and the old clock hardware stayed because replacing it gave no benefit we could measure.

20236 monthsLaravel and MySQL

Two hundred and fourteen tests, run in front of you

Three shapes of engagement

	AssessmentRp96,000,000four weeks	BuildRp185,000,000per month	MaintenanceRp42,000,000per month
Two people, four full weeks	✓
Access to the database and the old code	✓
A plan and a cost estimate with a range	✓
A list of what should not be built	✓
Credited in full if building goes ahead	✓
Three to four people		✓
A release to production every two weeks		✓
Code and database yours from day one		✓
A weekly meeting with users, not with managers		✓
Four month minimum		✓
Fixes within one working day			✓
Monthly security updates			✓
One person reachable 08.00 to 18.00			✓
A monthly report of what broke and why			✓
New features not included			✓

Book an assessment

The same order for twenty three systems

Weeks one and two, reading
We read the database and sit beside the people using the old system. No code is written. What comes out of these two weeks is a list of the real workflows, which almost always differ from the ones management described.
Weeks three and four, one whole workflow
One workflow is built end to end, including printing, export, and the cases that fail. Not a prototype that looks good, but the most frightening part of the system, so that our estimate is tested first.
Month two, first release to production
Used by one small group of real users. If the first release lands after month two, the scope is too large and we cut it rather than add people.
Every two weeks, one release
The full pipeline takes 15 minutes 29 seconds and runs on every commit. On average one release in nine fails at integration tests and is fixed the same day. That number is in your monthly report.
Six weeks side by side
The old system is not switched off on launch day. Both run together for at least six weeks and the numbers are compared every morning. The old one goes dark after seven consecutive days with no discrepancy.
Handover, and one free month
Repository, database, an operations document, and two sessions for your team. For the month after that we answer questions at no charge, and after that you are free to go to anyone.

What gets asked before an assessment

Do we own the code

Yes, from the first commit. We work in a repository in your company's name, not ours. The database, the configuration, and the operations document are handed over at the end with no further conditions, and no part is locked behind a licence.

Why is there no round the clock cover

Because sixteen people cannot staff three shifts without someone answering the phone at three in the morning after a full working day. We offer 08.00 to 18.00 and build systems that recover on their own outside those hours. For work that genuinely needs night cover we name two other providers.

Could this be cheaper with freelancers

It could, and several of our clients tried that before coming here. We do not subcontract and we do not enter tenders won on lowest price. We decline projects under Rp150,000,000, not because they are uninteresting, but because below that we cannot build them the way we promise.

What if our existing system is a mess

That is three out of five jobs here. We read the code first during the four week assessment, then give one of three answers: it can be repaired, it should be rebuilt in stages, or it should not be touched at all. We have given the third answer four times.

Who does the work, and do the people change

The person you meet at the assessment is the person who writes the code. People stay on one project until it finishes, and a mid project change has happened twice since 2017, both times for a new baby.

What do you not do

Consumer apps, marketing sites, games, and anything to do with crypto. We also decline projects whose deadline was fixed before the scope was read, because that kind of deadline is always paid for with skipped tests.

Bring the one process that breaks most often

Jalan Tebet Barat Dalam Raya 28, South Jakarta

Monday to Friday 09.00 to 18.00. The first conversation is ninety minutes, at your office, at no charge.

Message on WhatsApp