ERP Implementation Plan: A Realistic 90 Days
Most failed rollouts in Kenya were not badly built. They were badly sequenced — a system switched on before anyone decided who owns it, what counts as the truth on day one, and which three numbers would prove it was working. A ninety-day plan that assumes nothing about your budget and everything about your attention.
The question every Kenyan business asks a vendor is "how long does implementation take?", and the honest answer is that the software part takes about a week. What takes ninety days is the organisation deciding things it has been comfortably avoiding: which of the four stock figures in the building is the real one, whether the storekeeper is allowed to write off damage on their own, and what happens to the WhatsApp group everybody currently uses to approve purchases.
That is why implementations fail in ways that look technical and are not. Nobody ever says "we could not configure the warehouse". They say the staff went back to the old book, the stock figures never matched, or the person who understood it left. Those are governance outcomes wearing an IT costume.
The four decisions that have to happen before anything is configured
Take these in order. Each one closes a question that will otherwise reopen at the worst possible moment — usually during the first stock-take after go-live, in front of everyone.
-
Who owns it, by name
Not the MD, who will not log in daily, and not "IT", who do not know why a delivery was short. The owner is whoever will be embarrassed if the numbers are wrong — usually operations or finance. They need two half-days a week for the first six weeks, and that time has to come off something else, in writing.
-
What is true on day one
You are about to declare an opening position: this much stock, these suppliers, these customers owe this. Whatever you declare becomes the baseline every future variance is measured against. Declaring it carelessly does not save time; it converts a one-week counting job into a six-month argument about whether the system or the store is wrong.
-
Which modules, in which order
Almost nobody should switch on everything at once, and almost every vendor is happy to let you. Pick the module that fixes the thing currently costing you money, run it until it is boring, then add the next. Boring is the completion criterion.
-
The three numbers that prove it worked
Decide them now, while you are still honest, because after go-live everyone becomes invested in the project having succeeded. Good candidates: stock count variance, value of approved-but-unreceived purchase orders, and days to close the month.
A rollout has failed the moment the old spreadsheet is still being maintained "just for now". That is not a transition — it is two systems, and one of them will win by attrition.
The ninety days, week by week
This shape assumes an SME of ten to eighty staff with one or two locations, which is where most of this market sits. Bigger and multi-branch, stretch each phase rather than compressing the plan; smaller and single-location, you can genuinely do it in six weeks.
A ninety-day rollout that actually finishes
The week-5 count is the part people try to skip, and skipping it is the single most reliable way to be still arguing about stock figures in month eight. It is one bad week that buys you two good years. Note also that weeks 6–9 deliberately ask people to do the work twice — this is the only period where that is correct, and it must have a stated end date or it becomes permanent.
What is genuinely built for this, and where you are on your own
Loading a business into a new system is the least glamorous part of any product and the first place vendor claims go soft. Here is our own position, stated precisely enough to argue with.
What AWRA OpsHub does today
- Thirteen downloadable import templates — items, customers, vendors, assets, asset custodians, warehouses, locations, stock adjustments, stock transfers, requisition lines, RFQ lines, chart of accounts and opening balances — each with a sample row and the required columns marked.
- Row import working for eight of them: items, customers, vendors, assets, asset custodians, warehouses, locations and chart of accounts. Uploads under 256 KB report created / updated / skipped counts inline; larger files are queued and you are notified when they finish.
- A separate HR import path for employees and attendance that genuinely previews before it commits, so you can see what a file will do before it does it.
- Custom fields you have defined are added to the relevant import templates automatically, so your own columns are loadable rather than a manual pass afterwards.
- Enforced blind counting on stock count plans, which is what makes a week-5 opening count worth trusting rather than a copy of what the storekeeper expected.
What it does not do
- The other five templates are download-only — stock adjustments, stock transfers, requisition lines, RFQ lines and opening balances give you the layout, and there is no importer behind them. They touch records with approval and posting rules, and a bulk load that bypasses those would be worse than typing.
- No preview or dry run on the eight that do import. You get counts after the fact, not a diff before it.
- No rollback on an import. A bad file is undone by correcting records, not by pressing undo.
- No manual journal entry anywhere, which combined with the point above means opening financial balances cannot be loaded as balances at all — they have to arrive as real documents (unpaid invoices, unpaid bills, a stock check-in) or not at all.
- No migration tooling for any specific competing product. Nobody here has a QuickBooks reader.
The opening-balances limitation is the one worth pausing on, because it changes your plan rather than annoying you: it means your go-live date should sit at the start of a period, and it means your accountant should expect to enter the outstanding invoices and bills as documents. Ask any vendor the same question in the same words — "can I load a trial balance, and if so what posts it?" — and watch how long the answer takes.
Five questions that separate a real implementer from a demo
Ask these in the second meeting, not the first
Who does the data cleaning — you or us?
What you will hear
"We'll help you with it."
How to read it
That means you. Which is correct and normal, but it needs to be in the plan with a name and a number of days against it, or it will happen at 11pm the night before go-live.
What is the go-live date, and what is the date we stop the old system?
What you will hear
One date, confidently. Sometimes not the second one at all.
How to read it
The second date is the one that matters. A rollout with no stop date for the old process is a rollout that will run both forever.
Can I load my opening trial balance?
What you will hear
Anything from "yes, of course" to a long pause.
How to read it
Very few SME systems in this market can, and the honest ones say so and tell you what to do instead. A confident yes deserves a follow-up: what record does it create, and can I see it on screen?
Show me an import failing.
What you will hear
Reluctance, then a demo of it working.
How to read it
You will spend more time on bad rows than good ones. How the system reports a broken row is more predictive of your first fortnight than any feature on the brochure.
What happens if we want to leave in two years?
What you will hear
"You can export everything."
How to read it
Ask what "everything" means, whether attachments come with it, and whether it is one button or a support ticket. Ask us too — our own answer is that it is manual work, and we would rather tell you now.
The mistake almost everyone makes
They start with the module that is most broken instead of the module that is most contained. Procurement is usually the most broken thing in a Kenyan SME — it is where the money leaves and where nobody owns the process, as the procurement guide argues at length. But procurement touches suppliers, budgets, approvers, receiving and finance all at once, which makes it a poor first module and an excellent second one.
Inventory is contained. It touches the store, the storekeeper and a count. If it goes badly you find out in a week and the damage is a bad count. If procurement goes badly you find out in month three, and the damage is that finance no longer trusts the system, which is unrecoverable inside a year.
The one exception
If your actual crisis is that you cannot tell what you have committed to spend — orders placed and invisible until the invoice arrives — then start with procurement and accept the harder rollout. Sequencing rules exist to protect you from a slow failure; they do not apply when you are already in a fast one.
Our take
Name the internal owner and free up their time before you sign anything, put a full physical count in week five and do not negotiate it away, run one module in parallel with a stated stop date, and pick your three success numbers while you are still capable of being disappointed. Do that and the software choice matters much less than anyone selling you software will admit.
Plan the rollout before you pick the system
We will walk through the ninety days against your actual constraints — how many locations, who could own it, what state the item list is in — and tell you plainly which parts we would not attempt in your first quarter.
See plans & pricingFrequently asked questions
How long does an ERP implementation take for a Kenyan SME?
Configuration takes about a week. Getting to the point where the numbers are trusted without a manual check takes around ninety days for a business of ten to eighty staff on one or two sites — and most of that time is decisions and data cleaning rather than software work. Single-location businesses with a clean item list can do it in six weeks. Multi-branch groups should stretch the phases rather than compress the plan.
Should we run the old system in parallel?
Yes, for one module, for about four weeks, with a stated stop date announced in advance. Parallel running is how you find the differences while they are still small. It becomes the failure mode when it has no end date — then you have two systems, staff choose the one with less friction, and the new one loses.
Can we import our existing data?
Items, customers, vendors, assets, asset custodians, warehouses, locations and the chart of accounts import from a template. Five further templates — stock adjustments, stock transfers, requisition lines, RFQ lines and opening balances — download so you can see the layout but have no importer behind them, because they touch records with approval and posting rules. Employees and attendance have their own import that previews before committing.
Can we load our opening trial balance?
No. There is no manual journal entry in the system and the opening-balances template is download-only, so opening financial positions have to arrive as real documents — outstanding customer invoices, unpaid supplier bills, a stock check-in — rather than as balances. Plan your go-live for the start of a period and budget your accountant's time for entering the open items.
Which module should we start with?
Inventory, in almost every case, because it is contained: it touches the store and a count, and a bad start is visible within a week. Procurement is usually the more broken process but it touches suppliers, budgets, approvers, receiving and finance simultaneously, which makes it a better second module. The exception is if your live crisis is invisible committed spend — then start with procurement and accept the harder rollout.
What does the internal owner actually do?
They decide the ambiguous things — whether a storekeeper can write off damage alone, what the reorder point on a slow-moving item should be, which of two duplicate item records survives. Roughly two half-days a week for the first six weeks, and that time has to be taken off something else in writing. Rollouts without a named owner do not fail loudly; they just quietly stay half-done.
How do we know it worked?
Pick three numbers before go-live and record their starting values: stock count variance as a percentage of counted value, the shilling value of purchase orders approved but not fully received, and the number of days it takes to close a month. Choose them early — after go-live everyone becomes invested in the project having succeeded, and the definition of success drifts to fit the result.
