What ERP actually costs over three years

The assumptions, stated

None of these vendors publish list pricing, so the figures come from quotes and engagements we have seen rather than from a rate card. They are directional, they will not match your quote exactly, and publishing them is the only way the output means anything.

• Subscription is 36 months of a monthly figure built from a platform base, a per-user rate where the vendor charges one, and a per-entity fee beyond the first.
• Implementation is a multiple of first-year licence with a floor — 1.35× for NetSuite, 1.1× for Sage Intacct, 0.95× for Acumatica, 1.25× for Dynamics, 0.85× for Odoo, 0.25× for us. Those multiples are the single most contested number here and the one worth challenging your own quotes against.
• Integration scales with the number of systems you connect, at a per-vendor baseline for five connections.
• Discounts are not modelled. Most mid-market deals land 10–25% below first quote, and roughly evenly across vendors, so it moves the totals more than the ranking.

What this deliberately excludes

Your internal time, which is usually the largest line of all. A nine-month implementation consumes a controller for most of a year plus a testing week from the whole finance team. At a loaded cost of $180K, half a controller for nine months is around $67K that never appears in any vendor comparison — and it is real money whether or not anyone books it.

We leave it out because we would have to guess at your salaries and your project shape, and a made-up number attached to a real model discredits both. Estimate it separately: elapsed months × 0.4 × loaded controller cost is a defensible starting point.

Also excluded: data migration beyond standard scope, custom development, training beyond what is bundled, third-party add-ons from a vendor marketplace, and the cost of the failure case — which for roughly seventy percent of implementations means missing the original business case and spending again to fix it.

How to use the output

Not as a ranking. The useful move is to watch what happens to the ordering as you move the user slider. Below roughly 60 users the per-seat products are competitive; above 150 they separate sharply, and that divergence is the most reliable thing this model tells you.

The second useful move is to take the implementation multiple to whoever is quoting you. A partner proposing an implementation at 0.4× first-year licence is either scoping something much smaller than you think or has not finished scoping. Both are worth knowing before you sign.