Choosing a TMS in 2026 — a 7-step framework

Buying a TMS in 2026 is harder than buying a TMS in 2020 — not because the products are worse, but because the category has fragmented. There are thirty serious vendors, each with a credible-sounding pitch, each with a slightly different opinion about what a TMS is for. The most common procurement failure is not picking the wrong vendor. It's running an evaluation that can't tell the vendors apart in the first place.

This guide is the framework we wish we'd had when we were on the other side of the table. It is structured as 7 steps, in the order they actually need to happen. Skipping steps is the single best predictor of a 12-month implementation that doesn't ship.

Step 1 — Write down the three workflows you want to fix first

Not "modernise our TMS". Not "real-time visibility". Three named workflows. Examples from real procurement docs we've seen work:

• "Cut the time between indent creation and carrier confirmation from 47 minutes to under 10."
• "Reduce monthly POD-to-invoice cycle from 21 days to 3."
• "Shrink dispute resolution from 3 weeks to same-day for ≥80% of disputes."

If you can't write down three of these, you're not ready to evaluate vendors. You're ready to do an internal workflow audit. That's a different (cheaper) project.

Step 2 — Map your data sources before you map vendors

Every TMS implementation breaks at the integration layer. The breakages are predictable once you know which sources you're integrating. Build a one-page map of: ERP, WMS, carrier portals, e-invoicing, banking, e-way / customs, plus any home-grown systems. Mark each as "API available", "file drop only", or "manual". The ratio determines your implementation cost more than any line item on the vendor's quote.

Step 3 — Define the modules you actually need (and the ones you don't)

A modern TMS is modular. Most procurement teams sign for everything because the bundle looks cheaper. The bundle is cheaper per-module and dramatically more expensive in change-management. Pick the 3–4 modules that map to your three workflows from Step 1. Add the rest later.

Step 4 — Score vendors on 12 capabilities, not 200

Most RFPs run to 200+ line items. Vendors get checkbox fatigue and tick everything as "yes, with customisation". You get back a matrix that doesn't separate them. The 12-capability scorecard that does separate them — and that we've seen work in production at three different shippers — is in the appendix. The short version: prioritise capabilities that map to your top three workflows. Nothing else gets scored.

Step 5 — Run a paid 30-day pilot, not a free POC

Free proofs-of-concept are theatre. Both sides know they're being evaluated. Both sides ship A-team engineers. The result generalises to nothing. A paid pilot — with a real scope, a real go/no-go, and a real fee — produces evidence you can actually use. It also filters out vendors who can't operate inside a fixed scope, which is a useful filter.

Step 6 — Pricing: hunt the hidden line items

The list price is the smallest part of the conversation. The five line items that actually drive total cost of ownership are: per-shipment fees above contracted volume, integration setup, custom report builds, premium support tiers, and price escalation clauses. Ask for each one in writing. Ask for the renewal terms. Ask whether decommissioning data export is in-scope (it usually isn't).

Step 7 — Pre-mortem the rollout before you sign

Block four hours with the project sponsor and your top two ops leads. Pretend the rollout has failed. Write down why. The list — usually 8–12 items — is your real pre-implementation risk register. Address the top three contractually (SLAs, escape clauses, named team continuity). Address the rest operationally.