One Vendor or Three? How to Price the Interface Cost of Splitting an Outsourced Project

"Three separate quotes come in NT$380K cheaper than one"

A 65-person supplement brand needed a new website, a membership system, and a warehouse integration. The single-vendor quote was NT$2.18M. Split across three vendors — front end NT$680K, membership back end NT$920K, warehouse integration NT$200K — it came to NT$1.8M. The procurement lead brought us that NT$380K gap. We asked three questions first: how many interfaces will exist between those three vendors, who is responsible for making them line up, and how many hours a week can that person actually give? He thought for three seconds and said "probably Chen in IT, but he has other things." That NT$380K is gone by about month seven. This piece turns "interface risk" from a vague worry into a number you can put in a spreadsheet. (All figures are New Taiwan Dollar range estimates.)

Myths worth breaking

• Myth 1: "Splitting is always cheaper." Splitting saves on the quote and pays in coordination. The saving is one-off; the coordination cost recurs every year.
• Myth 2: "Splitting prevents vendor lock-in." Right instinct, wrong execution. Cutting one system down the middle locks you into two vendors at once, with nobody accountable for the whole result.
• Myth 3: "Just agree the interface clearly and it's fine." "Agreed clearly" is not a meeting, it is a deliverable. An interface with no API schema, data dictionary, error codes, or response-time commitment will produce "that's not our side" the week you go live.
• Myth 4: "More vendors means checks and balances, so better quality." Checks and balances require a referee. Without a technically credible integration owner on the client side, multiple vendors just turn disputes into an infinite loop.
• Myth 5: "The prime contractor subcontracts anyway, so I'll go direct and save." You save the prime's markup (typically 12–20% of the quote) and buy 100% of the integration risk. That trade is good for companies with a dedicated PM and a disaster for companies without one.

The core framework: interface cost and the three-question split test

First, count interfaces. Let n = the number of parties that must integrate with each other, including the client's own IT team — the one most people forget:

Interfaces I = n × (n − 1) ÷ 2

Then price each interface per year. Our working range: each interface consumes roughly 1.5 hours a week of client-side coordination (at a loaded hourly rate of about NT$600, that is roughly NT$47K a year), plus 2–3 accountability disputes a year, each stalling things 3–7 working days at NT$15K–40K in delay and rework. That is NT$80K–150K per interface per year.

Decision rule: the quote saving from splitting must exceed annual interface cost × expected system lifespan in years. In the opening case, n went from 2 to 4, so I went from 1 to 6 — an added NT$400K–750K a year. A NT$380K saving does not survive year one.

The three-question split test (split only if all three are yes):

• ☐ Does this scope have a testable delivery boundary (a real API contract, file format, or table-level boundary)?
• ☐ Is there a technically credible integration owner on the client side with 8+ hours a week available?
• ☐ Is the annual quote saving larger than the added annual interface cost?

Three company profiles, three different answers

• A. 18-person e-commerce brand, greenfield, no IT staff: force n down to 2. Use off-the-shelf payment and logistics services at published rates (for example the ECPay official rate page) and bundle the rest. Splitting is a pure loss here.
• B. 120-person manufacturer with an incumbent ERP vendor: n is naturally 3 (ERP vendor, new system vendor, in-house IT). The interface cost is unavoidable, so the job is to document it and write it into the contracts rather than pretend it isn't there. Budget NT$250K–450K a year explicitly.
• C. 300-person group with a 5-person IT team and multiple projects a year: splitting across 3–4 vendors works, but only if in-house IT is the integration owner and owns a shared acceptance environment and CI pipeline. Cross-boundary metrics such as front-end performance should be accepted against an objective standard — write the web.dev LCP definition into the SLA instead of letting each vendor define "fast."

The hidden cost list for split delivery

• Integration owner hours: 8–12 hours a week, roughly NT$250K–370K a year — a line item most companies never budget.
• Duplicated PM fees: each vendor bakes in 8–15% project management; split three ways and you pay it three times.
• Duplicated environments and infrastructure: separate staging per vendor, NT$30K–80K a year (usage-based cloud tiers such as Vercel's published pricing make this easy to model).
• Cross-boundary debugging: issues that span two vendors take roughly 2.5–4× longer to localize than single-vendor issues.
• Legal and contracting: NT$15K–40K per additional vendor for review, NDAs, and security clauses.
• Account and security governance: NT$20K–50K per vendor per year for access reviews, key rotation, and offboarding.
• Release-cadence gaps: misaligned schedules cost an average of 1–2 idle weeks per quarter.

An honest counterweight — single-vendor has hidden costs too: weaker renewal leverage (5–15% annual increases are common), single-point dependency, and nobody internally who understands the system. That is why the answer is not "always bundle," it is "do the arithmetic first."

Bundle-or-split fit scorecard (10 dimensions)

Score each 0–3, for a maximum of 30. 24+: splitting across three or more vendors is viable. 16–23: split into two at most. Below 16: bundle.

• Testability of the delivery boundary (can a machine verify it?)
• Interface documentation maturity (schema, data dictionary, error codes)
• Integration owner's technical judgment and available hours
• Technology-stack compatibility across vendors
• Clarity of data ownership and access boundaries
• Existence of a client-owned shared acceptance environment
• Precision of incident accountability and response-time terms
• How well release cadences can be aligned
• Size of the annual quote saving from splitting
• Takeover-ability if any one vendor exits

How ScriptWalker fits — and where we don't

• Project (bundled delivery): for n = 2 situations, one accountable party. Best fit for profile A.
• Retainer as integration owner: when multiple vendors are unavoidable and there is no internal capacity, we write the interface contracts, run joint testing, and act as referee in disputes. Roughly NT$35K–80K a month.
• Advisory: interface contracts and acceptance criteria only, no code. Suited to profile B where vendors exist but integration has drifted.
• Full outsourcing: we take the whole scope and hand it over to client IT in stages.

We are the wrong fit when:

• You have already decided to split across five or more vendors, with no integration owner, and want us to take the smallest slice
• You will not stand up a shared acceptance environment that you own
• You want us to give unlimited warranty over other vendors' deliverables
• Total budget is under NT$300K but the work spans three vendors
• You want endorsement for a split architecture that is already signed off and not open to recalculation

90-day startup playbook

• Weeks 1–2: draw the interface map, fix n and I, and put the interface cost into the total budget.
• Weeks 3–4: produce the interface contract — API schema, data dictionary, error codes, response times, and a change-notice period (10 working days is a good default) — and attach it to every vendor contract.
• Weeks 5–8: stand up a client-owned shared acceptance environment. Every vendor deploys into it. "It works on ours" stops being an argument.
• Weeks 9–12: run joint testing, then deliberately stage one cross-boundary failure to measure how fast accountability is assigned and repair happens.
• Day 90 review: three metrics — mean hours to localize cross-boundary issues, number of accountability disputes, and joint-test pass rate. Fail any one and reduce n.

Decision checklist

• ☐ I have calculated n and the interface count I
• ☐ I counted our own IT team inside n
• ☐ I know the expected lifespan of this system
• ☐ The split saving exceeds interface cost × lifespan
• ☐ A named integration owner exists with 8+ hours a week
• ☐ The interface contract is a document, not a meeting consensus
• ☐ A client-owned shared acceptance environment exists
• ☐ Contracts state how cross-boundary incidents are adjudicated
• ☐ Change-notice periods are aligned across vendors
• ☐ Data ownership and accounts are in the client's name
• ☐ Handover documentation exists in case any vendor exits
• ☐ An annual interface-cost line item is budgeted

FAQ

Where does NT$80K–150K per interface per year come from, and can I calculate my own?

You can, and you should. The formula is weekly coordination hours × loaded hourly rate × 52, plus annual disputes × cost per delay. Divide your PM's monthly salary by 150 for a rough hourly rate, then log two weeks of actual cross-vendor coordination time. Our figure is a common SME range, not a universal constant.

We already split across four vendors. Is it too late to fix?

No, but stop the bleeding first. Two priorities: stand up a client-owned shared acceptance environment, and retrofit the interface contract documentation. Those usually take 4–6 weeks and NT$150K–350K. Only then evaluate merging two vendors. Swapping vendors outright is almost always the most expensive option.

If the prime contractor subcontracts anyway, don't the same interface problems exist?

The interfaces exist; the accountability does not fragment. The difference is whether disputes are absorbed inside the prime or spent in three-party meetings on your calendar. Require prior notice of subcontracting and full prime liability for downstream delivery.

Do small projects (under NT$500K) need an interface contract?

If n = 2, a single page covering data formats and error handling is enough. Investment in interface contracts should scale with I: half a day for one interface, one to two weeks once you have three or more.

Run the numbers before you decide

We offer a free 30-minute interface cost calculation. Bring your quotes and scope, and we will work out n, I, and your annual interface cost together, then tell you whether that price gap survives year one. No bundling pitch — if you should split, we will say so.

• Email: [email protected]
• Phone: 0916-224-047
• LINE: @ufv9089p