The Nine Hidden Costs of an Implementation
Nine costs that are real, predictable, and absent from every proposal — because seven of them are yours rather than the vendor's. Each with a rough shilling figure and the moment it lands.
The word "hidden" is unfair to most vendors. These costs are not concealed — they are simply not the vendor's to quote, which means nobody quotes them and everybody is surprised by them. Seven of the nine below are your own hours. The remaining two are cash, and they arrive at predictable moments.
Listing them is worth about half the value of reading this, because a cost you have named stops being a surprise and becomes a line in a budget.
The nine, with when they land
| Cost | Rough figure | When it lands |
|---|---|---|
| 1. Your internal owner's time | KES 70,000 – 140,000 | Weeks 1–13. The largest single item. |
| 2. Data cleaning | KES 10,000 – 20,000 | Weeks 1–2, before anything else can proceed. |
| 3. The counting Saturday | KES 25,000 – 40,000 | Week 5. Includes overtime and a lost weekend. |
| 4. Everyone's training hours | KES 40,000 – 80,000 | Spread across weeks 3–13, invisibly. |
| 5. The parallel-run double work | KES 15,000 – 30,000 | Weeks 6–10. Real hours doing the job twice. |
| 6. Barcode hardware | KES 40,000 – 100,000 | Optional, and usually decided in week 4. |
| 7. Your accountant's extra hours | KES 20,000 – 60,000 | Go-live month, entering open invoices and bills. |
| 8. The second module's mini-implementation | KES 40,000 – 70,000 | Year 2. Feels like a surprise and is a certainty. |
| 9. Standing attention, forever | KES 24,000 – 36,000 a year | From month 4 onward, permanently. |
The four worth expanding on
-
Training hours are invisible because they are distributed
Forty staff at two hours each is eighty hours, and nobody experiences it as a cost because it arrives twenty minutes at a time. It is the second-largest item on the list and the one nobody has ever put in a budget. Count it, and the case for short role-specific sessions over an all-day workshop gets easier to make.
-
The parallel run is deliberate double work
For about four weeks somebody keeps the old book as well as the new system. This is correct — it is how you find differences while they are small — and it is real hours. Budgeting it makes the stop date easier to defend, because an unbudgeted cost feels like something to end early and a budgeted one feels like something that finishes on schedule.
-
Your accountant's go-live month
Somebody has to enter the open customer invoices and unpaid supplier bills individually, and your accountant is usually the right person because they know which of those receivables are actually collectable. Twenty to sixty thousand shillings, once, in the go-live month, and worth every shilling because the alternative is an ageing report that means nothing.
-
The second module always feels unplanned
You go live on inventory, it works, and six months later somebody says "we should do procurement". That is a success rather than a surprise — but it carries its own owner time, its own configuration decisions and its own training. Put a figure in year two now, and the conversation becomes a decision rather than an argument about scope creep.
The two vendor-side costs to ask about explicitly
Two questions with expensive wrong answers
Are all modules included, or licensed individually?
What you will hear
"You get everything you need."
What to establish
Get a written quote for the modules you will need in year three, not the ones you are starting with. Per-module licensing is common and it is where a cheap-looking quote becomes an expensive contract — and the increase arrives at the exact moment you have no leverage, because you are already committed.
What does an extra user cost mid-term?
What you will hear
The tier price rather than the marginal price.
What to establish
Growth is priced at the margin. Ask what one more seat costs in month eight, and whether adding it forces a tier change. A system where two extra people mean a doubled subscription is a system with a growth penalty built in.
Three costs people budget for and rarely incur
- Custom development. Almost always proposed in month two and almost never needed. Most requirements that feel bespoke turn out to be a configuration nobody found, a report nobody ran, or a process that should change rather than be automated. Wait three months before commissioning anything.
- Data migration services. Vendors offer them and they cannot do the part that matters, because deciding which of two similar item records survives requires knowing your business. Paying for import execution is fine; paying for judgement you have to supply anyway is not.
- Extra storage or bandwidth. Real for businesses generating thousands of attachments a month, and effectively zero for everybody else. Do not provision against a volume you have never produced.
The most expensive hidden cost is not on the list
A rollout that fails and has to be repeated. Nothing on this page is more than about KES 140,000; a failed first attempt costs all nine items twice, plus the credibility to ask staff to change how they work a second time. That is why why implementations fail is the cheapest thing to read before spending any of this.
Our take
Write all nine into the budget with figures against them, even rough ones, and the surprises stop being surprises. The two largest — owner time and distributed training hours — are both yours, both invisible, and both worth naming out loud so the trade-off is made deliberately. Then ask any vendor exactly two things: are modules licensed individually, and what does one extra user cost in month eight.
Budget the whole thing, not the subscription
We would rather walk through the nine costs against your actual [headcount](/glossary/headcount) and item count than quote you a monthly figure that covers a third of what the year will take.
See plans & pricingFrequently asked questions
What are the hidden costs of an ERP implementation?
Nine, of which seven are your own hours: your internal owner's time, data cleaning, the counting day, everyone's distributed training hours, the deliberate double work of a parallel run, your accountant's go-live month, the second module's mini-implementation in year two, standing attention forever, and optionally barcode hardware. They are not concealed by vendors — they are simply not the vendor's to quote.
Which hidden cost is biggest?
The internal owner's time, at roughly KES 70,000–140,000 across a rollout. Second is training hours, which is invisible because it is distributed — forty staff at two hours each is eighty hours that nobody experiences as a cost because it arrives twenty minutes at a time. Naming that one makes the case for short role-specific sessions much easier to win.
Should we pay a vendor for data migration?
For executing imports, sometimes. For the judgement, no — deciding which of two similar item records survives requires knowing your own business, and no vendor can do it. Paying for execution while supplying the judgement is reasonable; paying for a service that hands the hard part back to you is not.
Will we need custom development?
Almost certainly not in the first three months, though it will be proposed in month two. Most requirements that feel bespoke turn out to be a configuration nobody found, a report nobody ran, or a process that should change rather than be automated. Wait a quarter before commissioning anything and most of the list disappears.
What is the most expensive mistake?
A rollout that fails and has to be repeated. Every individual cost here is under about KES 140,000; a failed first attempt costs all of them twice and also spends the credibility you need to ask staff to change how they work a second time. That second currency is the one you cannot re-budget.
