Total Cost of Ownership: Five Years, Not One Month
Five years, not one month. The costs that recur, the ones that arrive in year three, and the single largest item on most Kenyan SMEs' five-year total — which is not the subscription and not the implementation.
Total cost of ownership is a phrase from procurement textbooks and it earns its keep here, because the shape of software cost is unusual: heavily front-loaded in effort, lightly recurring in cash, and then punctuated by two or three moments in year three that nobody modelled. Comparing monthly subscriptions tells you almost nothing about which system will have cost you more by 2031.
The five components
Where the money and the hours actually go
Subscription
The only line item any vendor quotes, and typically 20–35% of a five-year total for an SME. Predictable, easy to compare, and the least decision-relevant number in the set.
Implementation effort
Twelve to fifteen days of your internal owner, plus data cleaning, plus a counting day, plus everyone's training hours. Front-loaded, mostly invisible, and usually larger than the first two years of subscription combined.
Standing attention
About two hours a month of somebody tending the thing — reading the exception reports, re-tuning thresholds, checking who has accumulated permissions. Small, permanent, and the item most often assumed to be zero.
Growth and drift
More users, more employees, a second location, a tier upgrade. Also the module you did not switch on at first, which arrives with its own miniature implementation.
The exit
Extracting your data, loading it somewhere else, running both for a month. Nobody budgets this, everybody eventually pays it, and how expensive it is was decided by questions you asked or did not ask at the start.
A five-year model
40-staff distributor, Pro plus two users, five years
Subscription is 30–45% of the low estimate and under a third of the high one. Which means a competitor at half our monthly price is not half the cost — at best they are 15% cheaper overall, and if their implementation is harder or their exit is worse, they are more expensive.
The item most people get wrong
Standing attention. Two hours a month sounds like nothing, and over five years it is a hundred and twenty hours — comparable to the entire implementation. It is also the item that determines whether the other four were worth spending, because a system nobody tends drifts: thresholds set in 2026 demand conditions still sitting there in 2029, reorder points that no longer reflect anything, permissions that only ever accumulated, a connector that stopped working in March.
The cost of not spending it does not show up as a software cost. It shows up as the business concluding, in year three, that the system is not giving them what they expected — which is true, and is not the system's fault.
Every operations system decays into its defaults unless someone is paid to notice. That is the recurring cost nobody puts in the model.
Where vendors differ most, in TCO terms
Five questions ranked by their effect on a five-year total
Notice that monthly price is not on this list. It is the number you will spend the most time comparing and the least affected by.
Are modules licensed separately?
Make them prove it: Get a quote for what you will need in year three, not year one.
What does implementation cost, and who does the work?
Make them prove it: A fee is honest; no fee means the work is yours. Both are fine — price them the same way.
What does an extra user cost mid-term?
Make them prove it: Ask the add-on price, not the tier price. Growth is priced at the margin.
Is the price in shillings or dollars?
Make them prove it: A dollar-denominated contract is a currency position you did not intend to take.
How hard is it to leave?
Make them prove it: One export button, or a support ticket? Do attachments come out?
The two costs that are genuinely optional
Hardware and integration. Both get treated as inevitable and both are choices.
Barcode scanners and a label printer are somewhere between KES 40,000 and KES 100,000 all in, and they pay back quickly in a store with more than a few hundred lines and not at all in a store with sixty. Integration work — anything needing a developer — is the item most likely to be bought speculatively and never used; the test is whether a named person has asked for it, not whether it would be nice.
One cost that is not optional and is usually forgotten
A queue worker has to be running for background jobs to happen — large imports, eTIMS filing, outbound connectors, scheduled tasks. If you self-host anything or manage your own infrastructure, that is a small standing operational responsibility rather than a one-off setup step, and a queue that silently stops is a set of features that silently stop with it.
Our take
Model five years, not one month, and expect the subscription to be under a third of it. The three items to budget deliberately are the year-one implementation effort, two hours a month of standing attention forever, and the exit. Then compare vendors on per-module licensing, implementation responsibility and currency — because those move a five-year total by hundreds of thousands, and the monthly headline moves it by tens.
Model the five years, not the month
Every module in the plan rather than licensed separately, add-on pricing at the margin so growth is predictable, shillings as the contract currency — and an honest statement of what implementation and exit will cost you in your own hours.
See plans & pricingFrequently asked questions
What is the total cost of ownership of an ERP for a Kenyan SME?
For a forty-staff business over five years, somewhere between KES 840,000 and KES 1.26 million all in. The subscription is 30–45% of the low estimate and under a third of the high one; the rest is year-one implementation effort, about two hours a month of standing attention, growth costs as you add users and modules, and the eventual exit.
Why is the monthly subscription not the main number?
Because it is under a third of the total and it is the only component every vendor states clearly. A competitor at half the monthly price is at best around 15% cheaper over five years, and if their implementation is harder or their exit is worse, they cost more. The differences that actually move a five-year total are per-module licensing, who does the implementation work, and contract currency.
What is "standing attention" and why does it cost anything?
About two hours a month of somebody tending the system: reading the exception reports, re-tuning thresholds, checking accumulated permissions, noticing that a connector stopped working. Over five years that is a hundred and twenty hours, comparable to the whole implementation. Skip it and the system decays into its defaults — which shows up in year three as the business concluding the software is not delivering, correctly, and not because of the software.
Should we budget for barcode hardware?
Only if your store justifies it. Scanners and a label printer come to roughly KES 40,000–100,000, which pays back quickly above a few hundred item lines and not at all below about sixty. It gets treated as inevitable and it is a choice — as is any integration work, where the test is whether a named person has asked for it rather than whether it sounds useful.
How do we budget for leaving a system?
Between KES 60,000 and KES 200,000 of effort in a fifth year, and how much depends almost entirely on questions asked at the start: is there one export button, do attachments come out, and who performs the extract. Nobody budgets this and most businesses eventually pay it. Asking those questions before signing is the cheapest way to reduce it.
