Choosing an ERP is the most expensive software decision most Kenyan businesses make in a decade. Get it right and the next ten years of finance, HR, procurement and operations run smoothly. Get it wrong and you'll spend three years rebuilding around the limitations of a system you can't easily replace.
This is the five-step process we run with customers when they ask us "which ERP?" — including the questions that disqualify nine out of ten shortlists.
Why "which ERP?" is the wrong starting question
Most ERP selections start with vendor demos. That's already too late. Vendors will show you the shiny features that make their product look magical; they won't show you the workflows where their product is rigid, the integrations that don't exist, or the customers who quietly migrated off six months ago.
Start with your processes. Get clear on what you do, how you do it, and which parts of that are non-negotiable. Then go to the vendors with a defensible specification — not the other way round.
Step 1 — Map your actual processes
Document every meaningful workflow that ERP would touch: order-to-cash, procure-to-pay, hire-to-retire, plan-to-produce, period close. For each one, capture:
- The trigger — what kicks off the process?
- The actors — who does what, in what sequence?
- The decisions — where does someone with authority approve, route, or reject?
- The integrations — what other systems does it touch?
- The compliance touchpoints — KRA, NHIF, NSSF, OAG, regulator-specific?
This is the document you'll use to compare ERPs. It's also the document that exposes how much of your "process" is actually one person doing things from memory — which is its own problem.
Step 2 — Disqualify aggressively
Take your top-10 ERP candidate list and apply these disqualifiers. Most lists drop to two or three real contenders within an hour.
- No KRA iTax / eTIMS integration in the standard product. If you'd have to build the integration yourself, the vendor doesn't really sell in this market. Disqualify.
- No live customer in your industry, in Kenya, in the last 24 months. Reference customers from 2018 don't count. Disqualify.
- Refusal to share total 5-year TCO including services. If they only quote licence cost, the implementation cost is going to surprise you. Disqualify.
- Per-seat pricing that doesn't degrade gracefully. If your headcount grows 30% over five years, model what that does to the cost. If it scales linearly, your CFO will eventually have a problem with the deal. Disqualify.
- Customisation requires the vendor's professional services arm. Lock-in by design. Disqualify unless you're certain about the depth of partnership.
If you can't articulate the disqualifying answer for each candidate, you're not in selection — you're in shopping. Selection has rules.— David Baraza, Project Manager
Step 3 — Reference-check like an auditor
Every vendor will offer references. Half of those references are paid or coached. The references that matter are the ones you source — through industry contacts, LinkedIn, customer-event side conversations.
The questions to ask a real reference customer:
- How long did the implementation take vs. what was quoted?
- What's actually in production, what was descoped, and why?
- What integrations were harder than expected?
- If you were starting again tomorrow, what would you do differently?
- How responsive is the vendor's support team when something breaks?
Step 4 — Model 5-year TCO honestly
Add up: licence + implementation + customisation + integrations + training + ongoing support + cloud / hosting + internal team time. Multiply year-1 estimate by realistic growth and renewal escalation through year 5. Compare across candidates.
The cheap-licence option often turns out to be the most expensive system over five years. The "expensive" enterprise system sometimes wins on TCO because the implementation overhead is genuinely smaller. You won't know without modelling honestly.
Step 5 — Pilot before you sign the multi-year deal
Wherever the vendor allows it, pilot for one quarter on a single business unit or process before committing to enterprise-wide rollout. Most vendors will allow it; some will fight you on it. The fight itself is information.
The pilot tells you four things you can't learn from a demo: how the product feels in production, how the vendor responds when things break, how your own team adapts, and what the true integration overhead is.
A defensible answer
The selection framework above produces an ERP recommendation you can defend at board level — including the candidates you considered, the disqualifiers that ruled most out, the references you spoke to, the TCO you modelled, and the pilot you ran.
If your selection process can't reconstruct that audit trail, you're not selecting — you're guessing in a suit.