How to Prepare for a Software Implementation
Here is a pattern I have watched more than once. A company signs for new software, a kickoff call goes well, and everyone feels momentum. Six weeks later the vendor asks for the customer list and gets four spreadsheets that disagree with each other. Nobody can say who decides how a discount is approved. The two people who understand the old process are on vacation, and the go-live date is already in an email to the whole staff.
None of that is a software problem. It is a preparation problem, and preparation is the one part of an implementation that belongs entirely to you. The vendor can configure the product. They cannot decide your rules, clean your data, or persuade your team that the new way is better.
This guide is for the project sponsor, the person who signed or championed the purchase and who will answer for the result. Whether the system is a CRM, an accounting package, a scheduling platform or an ERP, the preparation work is much the same. I will lay it out roughly in the order it should happen, from before you sign to the day after launch.
Decide what done looks like before you start
Write a one-page statement of purpose. It should say what problem the software is solving, in terms a person outside the project would understand, and how you will know it has worked. Good examples are specific. "Quotes go out within one business day instead of four" is something you can check. "Better visibility" is not.
Add three more items to the page.
- What is in scope for the first release, and what is deliberately postponed. A first release that does fewer things well beats one that does everything badly.
- The date by which the old system or process will stop being used, or the date you will review whether it can.
- What you will not change. If payroll cannot move during the fiscal year end, say so now.
This page becomes the tiebreaker later, when someone proposes a clever addition in week five. You will be glad you have it.
Name the people, and give each one real time
Implementations stall most often because nobody had the hours. Name these roles explicitly. In a small company one person may hold several, but each should be written down.
| Role | What they decide or do | Typical time pressure |
|---|---|---|
| Sponsor | Resolves conflicts, protects the schedule, owns the outcome | A few hours a week |
| Project lead | Runs the plan day to day, talks to the vendor | Often a large share of a week |
| Process owners | Decide how their area works and sign off on it | Spikes at design and testing |
| Data owner | Decides which records are correct and what gets migrated | Heavy during cleanup |
| Technical contact | Handles accounts, access, email and network settings | Spikes at setup and cutover |
| Trainers or champions | Learn early and help colleagues | Rising near launch |
Then do the unglamorous thing and look at the calendar. If the project lead is also covering the front desk, the plan will slip. Either move some of their regular work to someone else for the duration, or lengthen the schedule honestly. In Puerto Rico, plan around known pressures such as hurricane season, holiday retail peaks and fiscal year ends. A cutover scheduled for a week when a storm warning is plausible deserves a fallback date.
Map the current process as it really runs
Vendors often ask for your process, and teams often describe the version in the manual. Walk through the real one. Sit with the person who does the work and watch. Note each step, who does it, which tools they touch, and where they leave the system to do something in a spreadsheet or on paper. Note the exceptions, such as the customer who pays by check, the supplier who invoices in a different currency or the job that needs a second approval.
The goal is not to document everything forever. It is to find two things. First, the rules that live only in someone's head, which the new system will need to be told. Second, the steps that exist only because of old limitations, which you may not need to carry forward. Moving a bad process onto new software gives you a bad process with a nicer screen.
Clean the data before it moves
Data cleanup is the preparation task that gets underestimated most, and it is where a sponsor's attention pays off. Migration software can copy bad records faithfully. It cannot tell you which of two conflicting customer addresses is right.
A workable sequence follows.
- Inventory the sources. List every place the relevant data lives, including spreadsheets, an old database, email folders and the personal notes of a long-serving employee.
- Decide what migrates. Not everything should. A five-year-old lead who never replied, or an inactive supplier with no transactions in a decade, may be better archived than loaded. Check record-keeping obligations with your adviser before deleting anything, because retention rules vary by industry and jurisdiction.
- Profile the data. Count records, count blanks in key fields, look for duplicates, and sort by the oddest values. A column of phone numbers will contain at least one entry that is plainly not a phone number.
- Choose matching keys. A matching key is the field used to decide that two records describe the same thing, such as a tax identifier, an email address or an account number. Agree on it with the data owner before merging.
- Fix at the source when possible. If you only correct the copy, the old system keeps producing the same errors until cutover.
- Do a trial load. Load a sample into a test environment and look at it with the people who know the real records. They will spot problems in minutes that a script cannot.
- Reconcile. After each trial, compare counts and totals between old and new. If the old system shows 1,204 active customers and the new one shows 1,190, find the fourteen.
A small hypothetical. A firm with 3,000 contact rows runs a duplicate check and finds that about 400 are repeats under slight spelling differences. Cleaning them before migration costs an afternoon of review. Cleaning them afterward means every sales report is wrong for months while salespeople argue about which record is real.
Make the decisions that configuration will need
Software does not arrive ready to run your company. It arrives with choices. Someone must decide the following before the vendor can set it up.
- Statuses and stages. What are the steps a deal, an order or a patient moves through, and what ends each one?
- Roles and permissions. Who may see, edit, approve or delete what? Start narrow and widen as needed, because it is easier than the reverse.
- Naming and numbering. How are customers, invoices and jobs identified, and does the new scheme collide with the old one?
- Approvals and thresholds. What amounts need a second signature?
- Required fields. Which data must exist before a record can move forward?
- Integrations. Which other systems must exchange data, in which direction, and how often?
Write the decisions down, with the name of the person who made each. When something is disputed a month later, you will have the answer in a document and not in a recollection.
Ask the vendor the questions that expose risk
- Who from your side will be assigned, and will that person stay for the whole project?
- What do you need from us, with dates, and what happens to the schedule if we are late?
- What is out of scope in this proposal that customers commonly assume is included?
- How do I export all my data if I leave, and in what format?
- How will you handle migration, and who is responsible for data quality?
- What does support look like after launch, including hours and response expectations?
- What changes would trigger a change order, and how are they priced?
Good vendors answer directly and welcome the questions. Treat evasion as information.
Write acceptance tests in advance
Acceptance testing is the stage where your people confirm the system does what the business needs. It fails when it is improvised. Write the tests while you design, so that "done" is not negotiated at the end.
Take the real scenarios from your process map and turn each into a short script with a clear expected result. A few examples, in plain language.
- Create a quote for an existing customer, apply the standard discount, send it for approval, and confirm the approver receives it.
- Enter an invoice from a supplier with two line items and confirm it posts to the correct accounts.
- Register a new client using a Spanish-language name with an accent and confirm the record, the printed document and the email all show it correctly.
- Remove a user who has left and confirm they can no longer reach any data.
- Simulate the failure case, such as a lost internet connection mid-entry, and confirm what the user sees and what happens to the half-entered record.
Run the tests with real data and real users, not only the project lead. Record each result as pass or fail, keep a list of defects, and agree in advance which kinds block launch and which can follow shortly after.
Train for the work people actually do
Generic product tours are forgettable. Train by role and by task. A person in billing needs to know how to raise, correct and chase an invoice, and does not need the report builder. Use your own data in a practice environment, record short videos for the most common tasks, and write a one-page cheat sheet for each role. In a bilingual workplace, offer materials in both languages if your team works in both, because a confusing screen in a second language is a common cause of quiet workarounds.
Schedule training close enough to launch that people remember it, and far enough that questions can still change the setup. Identify a few champions, colleagues who learn early and answer questions at the next desk. They carry more credibility than any vendor trainer.
Manage the change as well as the system
People resist change for sensible reasons. The new tool may be slower for them even if faster for the company. Their old spreadsheet may encode skill they are proud of. They may worry, with some reason, that automation means fewer jobs. A sponsor who ignores these concerns will meet them as low adoption.
A few practices help.
- Explain the reason in plain terms, early, and more than once. Tell people what is changing for them and what is not.
- Ask the people who do the work to help design it. Their involvement improves the result and the mood.
- Be honest about the awkward first weeks and promise support rather than perfection.
- Set a clear date after which the old way stops, and keep it, so that nobody is running two systems indefinitely.
- Watch real usage after launch. If a field is empty on every record, find out why instead of scolding.
Plan the cutover and the fallback
Cutover is the moment work moves to the new system. Choose a time when volume is low. Freeze changes in the old system for a defined window, run the final data load, reconcile the totals and only then open the new system for business. Keep the old one readable for a set period so people can look things up.
Write a rollback plan before you need one. It should say what conditions would cause you to revert, who decides, and what happens to the work entered in the meantime. Hopefully you never use it. Having it makes everyone calmer.
A short readiness checklist
- The one-page purpose is written and agreed by the sponsor and the process owners.
- Named people have scheduled time, and their regular duties are covered.
- The current process has been observed, with exceptions noted.
- Source data is inventoried, profiled, deduplicated and reconciled after a trial load.
- Configuration decisions are written down with the decider's name.
- Acceptance tests exist, with real scenarios and an agreed definition of pass.
- Training is planned by role, with champions identified.
- The cutover date avoids known business peaks and storm risk, and a fallback date exists.
- A rollback plan and a post-launch support plan are written.
What to do Monday morning
Write the one-page purpose and send it to your process owners for comment. Then open the calendar and put real hours next to real names. Those two actions surface most of the conflicts that would otherwise appear in week six.
If you would like another pair of eyes on a readiness plan, or help profiling messy data before a migration, I am glad to talk it through. Ozcorp InfraOps works with small teams on exactly this kind of preparation, and a first conversation is free of charge.
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.