Support and Evolution

You inherited software nobody documented

In this article

Nori.mx7 min read

final_v3system.execonfig.inidata.mdb12 years on the floorThe two-hour testIs the source code there?Is it in version control?Does it build today?Does it run on a copy of the DB?If #1 fails, it is an installation

Almost every plant on the border has one. A program that has been running for eight or twelve years, that nobody wants to touch, that does something the corporate ERP does not do, and that was written by someone who no longer works here.

While it runs, it is not an issue. It stops being a non-issue the day somebody asks whether one thing can be changed, and the honest answer is that nobody knows.

What you actually inherited

You did not inherit a program. You inherited three separate things, and it pays to keep them apart because they get fixed separately.

The code. It may or may not exist. That is the first thing to find out, and it is more common than you would think for nothing to exist but the executable installed on one machine.

The knowledge. Why the process works this way, what that field means, why there is an exception for one customer. That is almost never written down. It lives in the head of the person on the floor who has used it longest, and that person does not know they know it.

The responsibility. Somebody has to answer the phone when it goes down. Today that somebody is usually whoever uses it, improvising, and that is what actually costs money.

The risk is not that it crashes

A crash is the fear people name, but it is rarely what does the most damage. An old system that is stable and lightly loaded does not crash much.

What actually happens is that the business runs out of room to move. The customer asks for a new label and it cannot be done. A compliance requirement changes and it cannot be done. You add a shift and the report stops adding up. Every reasonable decision runs into “that system cannot be modified,” and the plant starts organizing itself around the limitation instead of around what makes sense.

That is the cost, and it shows up on no invoice.

The two-hour test

Before quoting anything, there is one test that answers most of the question. It needs no budget and takes an afternoon: try to build it from scratch on a clean machine.

  • Does the complete source code exist, or only the installed program?
  • Is it in version control, or in a folder called final_v3?
  • Does it still compile today, with tools you can actually download?
  • Does it start against a copy of the database without touching production?

If all four answers are yes, you have maintainable software, however ugly and undocumented. If the first one fails, you do not have software: you have an installation, and what comes next is not maintenance, it is reconstruction.

Most of the cases we see in Juarez split exactly there, and almost nobody has checked before asking for a quote.

Five questions to take inventory

What follows is boring, and it is what saves the project. Five questions, in writing:

  1. Where does it run? A server in the IT room, a PC under a desk, a cloud account someone pays for with a personal card. All three exist, and the third brings the most surprises.
  2. What does it depend on? A Windows version that no longer gets patches, a database engine out of support, a scanner with a proprietary driver, an Excel file somebody drops in a shared folder every morning.
  3. Who holds the keys? Administrator passwords, domain access, the account the certificate is registered under. If the answer is one person, the risk has a first and last name.
  4. What goes down with it? Usually more than it looks: the label, the report, the file that feeds another system. If a number comes out of there and ends up in a customs filing or an IMMEX report, this stopped being an IT problem and became a compliance one, which is what we covered in what Annex 24 asks of your inventory system.
  5. Is there a backup, and has anyone restored it? A backup nobody ever restored is not a backup. It is a large folder.

Rescue or replace

The decision is not made by looking at the code. It is made with three criteria, best answered in this order:

Is the data any good? If the history is clean and complete, it is worth far more than the program. An ugly system with good data can be rescued; a pretty system with dirty data is useless the day it ships.

Is the process still the right one? Sometimes the system is rigid because a ten-year-old process got frozen inside it. Then there is nothing to rescue: the work is to redesign the process and decide on software afterwards, never the other way around.

Does the platform have a way forward? A language with a community and people who still write it lets you continue. A discontinued tool with a file format only it can open does not. This one is technical, and it is the only one of the three that is.

Rescuing usually costs a quarter to a third of rebuilding, and it delivers much sooner. Replacing is justified when the process changed, when the platform has no way forward, or when the system is the bottleneck for three areas at once.

If this system is one of several open fronts, the order matters, and there is a way to decide it that is not whoever shouts loudest.

Week one, whatever you decide

Some things get done regardless, even if the real decision takes months. None of them needs budget approval:

  • Get the credentials out of one person’s head and into somewhere the responsible area can recover them without him.
  • Restore a backup, even on a borrowed machine, and write down how long it took. That number is your real recovery time, and odds are nobody knows it.
  • Copy the code into a repository, exactly as it is, without cleaning it up. Before improving anything, make sure it exists somewhere it cannot be deleted from.
  • Document the critical path only: what happens from the order arriving to the label coming out. One page. Not the whole system, which never gets finished and so never gets started.
  • Make failures visible. If the program fails and nobody finds out until an operator says so, the first fix is not code: it is getting the failure onto a screen the same day.

With that, the system stops being a bomb and becomes an open item. That is a change of category, and it takes a week.

What to demand from a support contract

When it is time to hire someone, four things matter more than price:

  1. Who maintains what, in writing. Operating system, database, application. The gap between two providers is where long outages live.
  2. Response time and resolution time, separately. They are different, and only the second one matters to the plant.
  3. What happens if the provider disappears. Code, credentials and documentation have to sit on your side from month one, not at the end.
  4. Evolution hours included. Support that only puts out fires leaves the system just as rigid next year. The difference between maintaining and evolving is that the second one gives back room to move.

What not to do

Do not rebuild it because it is ugly. A system that has been on the floor for twelve years contains hundreds of small decisions somebody made for a reason, and half those reasons are no longer remembered by anyone. Rebuilding it from memory means rediscovering them one at a time, in production.

And do not leave it alone just because it still works. The question is not whether it works. It is whether it can change when the business needs it to, which is the same question as the ERP that does not reach the floor.

It is worth checking whether the same thing happens outside the plant. If the undocumented system worries you, the customer base living in one rep’s inbox is the same exposure wearing a different face: company knowledge kept where the company cannot reach it.

And there is a test with a date on it. If your attendance runs on that system, in January 2027 the workday record changes requirements and somebody will have to modify the calculation. If there is nobody, you know how alive it is.

If you want to know where you stand, you do not need to hire anybody: run the two-hour test and answer the five questions. In one afternoon you will know whether what you have is software or an installation, and that changes everything that comes after.

Related service

Support and Evolution

Launching the software is the beginning. Keeping it alive is the work.

See the service

Articles on this service

Copy 1Office serverCopy 2External driveCopy 3OffsiteRestore testFile: invoices-2026.xlsxCopy used: 3, offsiteOn another machineIt opened in full

Support and Evolution

A backup is worthless until you restore it

Almost every company backs up and almost none tests a restore. What to copy, where to keep it, how to test it and who has to own it.

7 min read

NOVEMBERSunday12:00 → 1:00Juárez and El PasoScans1:40Pallet A1st pass1:10Pallet B2nd passDowntime−45 min1:50 → 1:05actual: 15 minBy time: B before ANight shift23:00 → 7:00Clock says8 hWorked9 h

Support and Evolution

On Sunday, 1 a.m. happens twice

This Sunday Juárez sets its clocks back an hour. What breaks that night in time clocks, reports and servers, and what to check before it happens.

7 min read

Plant serverYour applicationSQL Server 2016Windows Server 2016Last patchJanuary 12, 20271Upgrade2022 or 2025, app on top2Extended updatesThrough 2030, via Azure Arc3IsolateOff the general network

Support and Evolution

Windows Server 2016 runs out of patches in January

Windows Server 2016 gets its last patch on January 12, 2027. What end of support really means, the three ways out, and why the hard part is the app that runs on it.

8 min read