Implementing a Treasury Management System: A 20-Week Plan From Kickoff to Go-Live
Signed the TMS contract? This week-by-week plan sequences bank connectivity, ERP links, cash positioning and approvals around the long-lead items that most often push go-live back by a quarter.
What has to be true before week one
Your contract is signed and the vendor's project manager wants a kickoff date. Before you agree to one, confirm three things are in place. A treasury management system implementation that starts without them usually loses its first month to rework.
- A named bank list with priorities. Every bank, country and account in scope, ranked by payment volume and cash balance. Banks holding most of your liquidity go in wave one. The rest follow after go-live.
- A treasury owner with decision rights. One person who can approve the account structure, the approval matrix and the cutover date without convening a committee.
- IT and ERP capacity you've actually booked. Integration work typically needs an ERP developer or integration specialist for roughly a third to half of their time from week 5 to week 14. If that person is also on a finance transformation program, expect slippage and plan for it now.
The plan below assumes a mid-sized multinational: 8-15 banks, 60-200 accounts, 10-30 legal entities, one main ERP and a cloud TMS such as Kyriba, GTreasury, FIS Treasury and Risk Manager or ION Treasury. With more than 20 banks or several ERPs, stretch it to 26-32 weeks rather than compressing it.
The 20-week timeline at a glance
The phases overlap on purpose. Bank connectivity has the longest external lead time, so it starts while design is still being finalized.
| Weeks | Phase | Exit criteria | Where it usually slips |
|---|---|---|---|
| 1-3 | Mobilize and design | Signed-off design, bank list, connectivity method per bank | Account inventory turns up unknown accounts |
| 3-10 | Bank connectivity | Statements received from every wave-one bank in production format | Bank onboarding, KYC paperwork, certificates |
| 5-11 | ERP integration | Payment files and GL postings flowing in test | ERP developer availability, mapping disputes |
| 8-12 | Cash positioning | Daily position matches manual spreadsheet within agreed tolerance | Missing intraday data, forecast categories |
| 10-14 | Payment workflows | Approval matrix tested, segregation of duties signed off | Policy debates surfacing late |
| 14-18 | Testing and parallel run | Two to four weeks of clean parallel results | Test payments with banks, month-end clash |
| 19-20 | Cutover and hypercare | Go/no-go criteria met, legacy channels retired on schedule | Unretired legacy bank portals |
Weeks 1-3: Design decisions that are expensive to reverse
Use these three weeks to settle the decisions that ripple through everything else. Workshops should produce documents, not slide decks.
- Account and entity structure. Load a complete account inventory and reconcile it against bank confirmations. Expect to find dormant accounts and a few nobody can explain. Decide now whether to close them or bring them in.
- Connectivity method per bank. Assign each bank to SWIFT, host-to-host, EBICS, API or the vendor's own connectivity hub. This determines your onboarding lead times.
- Payment types in scope. Treasury payments only (intercompany, debt, FX settlement), or also AP runs from the ERP? Including AP can roughly double integration and testing effort. For a first go-live, keep it out unless AP consolidation is the reason you bought the system.
- Data residency check. Confirm where the vendor hosts and processes payment data. Rules such as GDPR in Europe, India's DPDP Act and central bank payment-data requirements, and Saudi Arabia's PDPL can affect hosting region and subprocessors. Have legal or compliance counsel confirm the position for each country rather than relying on the vendor's summary.
Exit criteria: a signed design document, a frozen wave-one bank list and a named owner for every open decision.
Weeks 3-10: Sequencing TMS bank connectivity by lead time
TMS bank connectivity is where most 20-week plans become 30-week plans. Your project team doesn't control the pace; each bank's onboarding desk does. Start the slowest channels first.
SWIFT for corporates
If several global banks are in scope, SWIFT through a service bureau, or the TMS vendor's own bureau, is usually the right call for a mid-sized group because you avoid running your own infrastructure. Allow roughly 8-12 weeks: SWIFT onboarding, a BIC for your group, then each bank's acceptance of your corporate setup under SCORE or bilateral agreements. Typical 2026 market ranges are $20,000-$70,000 a year for a bureau service plus per-bank setup, varying widely by contract and message volume. Confirm with vendors.
Host-to-host and EBICS
Regional and domestic banks often prefer host-to-host SFTP links, especially in India, the Gulf, Latin America and parts of Southeast Asia. In Germany, France, Austria and Switzerland, EBICS is common. Expect roughly 6-10 weeks per bank, mostly spent on file format specifications, IP whitelisting and key exchange. Local formats still matter: Brazilian banks may want CNAB files, and some Asian banks run their own ISO 20022 variants.
Bank APIs
APIs give real-time balances and faster payment status, but coverage is uneven by bank and country. Use them as a supplement for intraday visibility at banks with production-grade services, not as your only channel in wave one.
Red flags: a bank that hasn't assigned an onboarding contact by week 4, KYC or board-resolution paperwork still unsigned in week 5, or test files exchanged by email instead of through the real channel. Exit criteria: production-format prior-day statements (MT940 or camt.053) arriving automatically from every wave-one bank.
Weeks 5-11: ERP treasury integration without a mapping war
ERP treasury integration runs in two directions, and teams routinely underestimate the second. Inbound to the TMS you need AP payment files or payment requests, open receivables if you forecast from them, and master data such as entities and GL accounts. Outbound to the ERP you need journal entries for bank fees, interest, FX deals, intercompany loans and cash movements.
Agree the GL mapping table early and get the controller's signature on it. Most integration delays aren't technical. They're accounting debates about which account a hedge revaluation hits, settled in week 10 when they should have been settled in week 5.
Choose the integration method by ERP. SAP S/4HANA and Oracle Fusion customers can often use vendor-supplied connectors. Microsoft Dynamics, NetSuite and regional ERPs commonly rely on scheduled flat files or middleware. Flat files are less elegant but faster to stabilize. For a first go-live, take the faster path.
Exit criteria: a full test cycle where ERP payment files enter the TMS, reach the bank test environment and post back correctly to the GL.
Weeks 8-12: Cash positioning setup that treasurers will trust
A cash position that doesn't match the spreadsheet won't get used. Judge your cash positioning setup against that spreadsheet, not against the vendor demo.
- Define the position cut-off time for each region, since an Asian close and a North American close may land on different value dates.
- Build transaction categorization rules from real bank descriptions. Expect partial automation at go-live, then tune the rules over the first months.
- Decide which intraday sources (MT942, camt.052 or API) feed the same-day view and which accounts stay on prior-day only.
- Keep forecast categories to a short list at first. Fifteen well-used categories beat sixty ignored ones.
Say a group with 120 accounts runs the TMS position alongside the spreadsheet for ten business days. If the variance stays within an agreed tolerance each day, often a fixed amount per entity, the position is ready. If not, trace every break to its source before moving on.
Weeks 10-14: Payment approval workflows and controls
This is where the TMS starts preventing fraud, so design it with internal audit in the room. Build the approval matrix by payment type, amount band, entity and currency. Require at least two approvers for treasury payments above a defined threshold, and make sure no one can both create a beneficiary and approve payments to it.
Settle these explicitly:
- Beneficiary creation and change controls, including call-back verification for new bank details.
- Authentication: hardware tokens, mobile authenticators or SSO with your identity provider.
- Whether payments are approved in the TMS only, or banks also require approval in their own portal. Double approval adds work, but some banks and regulators expect it for certain payment types.
- Sanctions screening: whether the TMS screens, the bank screens, or both.
Exit criteria: a segregation-of-duties review signed off by internal audit and every approval path tested with real user roles, not admin accounts.
Weeks 14-18: Testing and the parallel run
Run user acceptance testing for two weeks, then a parallel run of two to four weeks. During the parallel run, the legacy process stays authoritative while the TMS processes the same activity, and the team reconciles both daily.
Send low-value live test payments, sometimes called penny tests, through every bank, currency and payment type in wave one. Bank test environments often behave differently from production, so a clean test-environment result isn't proof.
Don't run the parallel run across fiscal year-end or a quarter-end close. If the calendar forces it, extend the run rather than pushing cutover into close week. After bank onboarding, this is among the most common slip points, and it's entirely in your control.
Weeks 19-20: The go/no-go rule and the first 30 days
Set the go/no-go criteria in writing before the parallel run starts, so nobody negotiates them in week 19. A workable rule: go live only if every wave-one bank has delivered statements for ten consecutive business days, all payment types have completed a live payment, the cash position has matched within tolerance, and no severity-one defects are open.
If one bank fails, go live without it and keep that bank on its legacy portal. Don't hold the whole program for one connection.
Plan hypercare for about 30 days with daily vendor check-ins. Then retire legacy bank portals and payment tokens on a fixed date. Teams that leave old channels open "just in case" end up running two processes indefinitely, and the control benefits of the TMS disappear. Your next step this week: send each wave-one bank a written onboarding request with your target connectivity date, because that clock starts only when they receive it.
Questions buyers ask
Is 20 weeks realistic for a treasury management system implementation?
For 8-15 banks, one ERP and a cloud TMS, yes, if bank onboarding starts in the first three weeks. Groups with more than 20 banks, several ERPs or AP payments in scope should plan for 26-32 weeks. The deciding factor is usually how quickly banks complete onboarding, not the software.
Should we use SWIFT, host-to-host or APIs for bank connectivity?
Most multinationals use a mix. SWIFT for corporates suits groups with several global banks, host-to-host or EBICS suits regional and domestic banks, and APIs add real-time visibility where banks offer production-grade services. Choose per bank based on what it supports and your payment volume there.
What does TMS bank connectivity typically cost?
As a typical 2026 market range, SWIFT service bureau fees often run $20,000-$70,000 a year, with per-bank setup fees on top, while host-to-host links may carry one-off setup charges from the bank. Prices vary significantly by region, volume and contract, so confirm directly with your TMS vendor, bureau and banks.
How long should the parallel run last?
Two to four weeks is common, covering at least one full payment cycle and ideally a month-end. Fix the go/no-go criteria in writing beforehand and extend the run rather than going live with unresolved reconciliation breaks.
Written by the Vaultgridplus editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.