Modernize Outdated Internal Tools
There is a database on a desktop computer in the back office, and everyone knows not to touch it. It was built eleven years ago by someone who has since moved to another company, or another island. It prints the delivery sheets, tracks who owes what, and sends a nightly email through a macro nobody has read in a decade. If it stops, the business notices by about 9 a.m.
If you own a tool like that, whether it lives in Microsoft Access, FileMaker, an Excel workbook full of VBA, or some mix, you are holding two things at once. It is an asset, because it encodes years of how your business really works. It is also a liability, because it has a single point of failure, no documentation and an aging platform under it.
The temptation is to replace it all in one heroic project. That is usually the wrong move. This article is about the quieter alternative, which is to understand it, secure its data, and replace it a slice at a time while the old tool keeps earning its keep.
Why these tools survive, and why that is not stupid
Before you plan its retirement, give the tool its due. It survived because it fits. The person who built it understood the work, changed it every time the work changed, and never needed a requirements meeting. Replacement projects fail when they underestimate this and assume the old tool is just a messy database. Often it is a living specification of your business rules, written in forms, queries and macros.
It also helps to be plain about the real risks, because they differ.
| Risk | What it looks like | How likely it is to hurt |
|---|---|---|
| Key person | Only one person understands it | High in small teams |
| Platform age | Runs on an old operating system or file format | Rises each year |
| Single file | One shared file on one machine, with no tested backup | High, and quiet until it fails |
| Concurrency | Several users corrupt the file by editing at once | Common with file-based databases |
| Security | No real logins, no audit of changes, data copied to USB drives | Depends on the data held |
| Compatibility | Breaks after an Office or operating system update | Sudden, and often on a bad day |
Notice that most of these are about how the tool is housed, not about whether the logic is good. That distinction shapes the plan.
Step one, take an inventory before you touch anything
You cannot replace what you have not described. Spend a week, not a quarter, building an honest inventory.
List every function, not just the screens
Sit with the people who use it and ask them to show you a normal Tuesday. Write down each thing it does, including the invisible ones.
- The reports it prints, and who reads them.
- The numbers it calculates, such as late fees, commissions or taxes, and the rule behind each.
- The emails, exports and files it sends out, and to whom.
- The data it receives, and from where.
- The scheduled things, like a macro that runs at midnight.
- The workarounds, such as the column everyone uses for notes that actually means something.
The invisible jobs are what get lost in a migration. Nobody complains about the nightly backup export to the accountant until it stops.
Find out who depends on it
Make a short table of users, what they do in the tool, and how often. Include people who never open it but depend on its output, like the person who receives the weekly email. Ask each of them what would happen if it were down for one day, and for one week. Those answers tell you where the pain is, and where you can afford a pause.
Record where it lives and how it runs
Write down the file location, the machine, the version of the software, the licenses, who has admin rights, and how backups work. Then test whether the backup can be restored. A backup that has never been restored is a hope, not a control.
Capture the rules in plain language
For each calculation or decision the tool makes, write the rule in a sentence a non-programmer can check. "Customers on net 30 terms get a reminder on day 25 and a late fee of 1.5 percent after day 35." That sentence is more valuable than the code, because it survives any platform change. Then compare your sentence to what the code really does. You will find surprises, and some of them will be bugs the business has been living with.
Step two, make the data safe before you move any logic
If you do one thing this month, make the data safe. The data is the part you cannot rebuild.
- Make a dated copy of the whole file, and store it in a second location.
- Export each table to a plain, open format such as CSV, and keep the exports.
- Document each table and field, what it means, and which fields are keys.
- Check for problems like duplicate records, orphaned rows that refer to something deleted, and fields that hold two different kinds of information.
Move the data to a real database
A common and low-risk first move is to keep the old front end, meaning the forms and reports people know, while moving the tables to a proper server database. Access, for example, can link to tables stored in a server database. FileMaker can connect to external data sources. This gives you real backups, real user logins, safe concurrent access and room to grow, without changing what users see on Monday.
It also creates a clean seam. Once the data lives in a database with a defined structure, anything new, such as a web page, a report tool or an integration, can read and write the same records.
Watch for data traps
- Dates stored as text, which sort and compare wrongly.
- A field holding "Smith, John" in one row and "John Smith" in another.
- Numbers with hidden formatting.
- Calculated values that were typed in by hand and no longer match their inputs.
- Character encoding, which matters for accents and the letter n with a tilde in a bilingual business.
Fix the data issues in the new location rather than in place, so you keep the original untouched as evidence.
Step three, replace by slices with the strangler pattern
The strangler pattern is named after a vine that grows around a tree until it can stand on its own. You build the new system around the old one, move one function at a time to it, and retire the old parts only when nothing depends on them.
Choose the first slice carefully
A good first slice has these qualities.
- It is small enough to finish in a few weeks.
- It is visible and painful, so people feel the benefit.
- It has clear rules and few dependencies.
- It can be checked against the old result.
Often that is a read-only report or a lookup screen. Reporting is a natural first choice, because it changes no data, so a mistake costs little. Data entry screens come later, after the new data layer has proven itself.
Route work through a single front door
As slices move, users should not need to decide which system to use. Give them one entry point, such as a simple internal web page with links, so the old and new parts feel like one tool. Behind it, each function either calls the old tool or the new one.
Keep one source of truth for each piece of data
The hardest moment in a strangler migration is when both systems can write the same data. If both can edit a customer's address, you will eventually have two versions. For each table, decide which system owns it at each stage. The other system reads from it, or is read-only for that data, until ownership moves.
Decide what not to rebuild
A replacement is a chance to drop features. Ask of each function whether anyone still uses it. The inventory from step one, plus a look at usage over a month or two, will reveal screens nobody has opened since the last administration. Retiring unused features is the cheapest work in the project.
Step four, run in parallel before you cut over
Parallel running means operating the old and new systems at the same time on the same real work, and comparing their results. It is the safest way to prove the new tool is right, and it is also tiring, so plan it with a clear start and end.
What to compare
- Totals, such as the amount invoiced for the week.
- Counts, such as the number of orders shipped.
- Individual records, where you sample a set of cases and compare line by line.
- Edge cases, such as returns, partial shipments, credit memos and customers with unusual terms.
An illustrative example
This example is hypothetical. A small distributor replaces a FileMaker order and invoice tool. During a four-week parallel run, staff enter each order in both systems. At the end of each week, someone compares invoice totals and counts. In week one, the totals differ by a small amount on 6 of 120 invoices. Investigation shows the old tool rounded tax per line item while the new one rounded on the invoice total. Nobody had written that rule down, and the accountant had been relying on the old behavior. They decide which rounding the business wants, document it, and fix the new system. By week four, the totals match on all invoices, and the team agrees to cut over.
The lesson is that parallel running does not mainly find software bugs. It finds the rules nobody knew they had.
Define cutover and rollback in advance
Write down the exit criteria. For example, "Four consecutive weeks of matching totals and no open critical issues." Choose a cutover time with low activity, and not the week of month-end close. Prepare a rollback plan, which means knowing exactly how to go back to the old tool, and for how long you will keep it available in read-only form. Keep the old system and its data for a set retention period after cutover. Records retention rules vary by industry and jurisdiction, so confirm yours with your accountant or counsel.
Choosing what to build on
Do not pick a platform first. Pick it after you know your rules, your data and your first slice. Then compare options against your real constraints.
| Option | Fits when | Watch for |
|---|---|---|
| Stay on the same platform, upgraded | The tool works and the risk is mainly housing and backup | You may still be tied to a shrinking skill pool |
| Low-code platform | The workflows are form and approval shaped | License costs that grow with users, and export options |
| Off-the-shelf software | Your process is standard for your industry | Configuration limits, and migration effort |
| Custom web application | Your process is distinctive and central to the business | Ongoing maintenance, and a need for an owner |
For each, ask about data export. Can you get all your data and your definitions out, in a usable format, if you leave? Who owns the code or configuration? What does the vendor do if they are acquired or retire the product? The same questions you ask of your old tool now apply to the new one.
Questions for any vendor or developer you hire
- Will you start with a discovery of what the current tool does, and will I get that document?
- How will you move and verify my data, and can I see a reconciliation?
- Can you work in slices so I get value before the end?
- What do I own at the end, including code, data model and documentation?
- Who supports it after launch, and what happens if you are unavailable?
Resilience notes for Puerto Rico
If your tool lives on one computer in one office, think about hurricane season. Power loss, flooding or a damaged machine can take the whole business logic offline with it. Moving the data to a hosted database or to a service with off-site backups is a resilience improvement even before any feature work. Also think about how staff will reach the tool if the office is closed. Whatever you build, write a short recovery note that covers where the data lives, who can restore it, and how to reach the people who can.
What to do on Monday morning
- Make a dated copy of the file and store it somewhere other than the same computer.
- Try to restore that copy on a different machine, to prove it works.
- Ask two users to show you a normal day in the tool, and write down what they do.
- List the rules in plain sentences, starting with anything involving money.
- Name the one function you would replace first, and why.
You do not have to decide the destination today. You only have to stop depending on luck. If you want someone to go through the inventory with you, Ozcorp InfraOps helps small teams in Puerto Rico and the United States move old internal tools forward without stopping the work that depends on them. I am glad to talk it through.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.