Vendor Portfolio Design by Body of Work

Two bodies of work, two seats each, and a different reason for the pair in each: in customer-facing servicing the two seats give the body both continuity and a comparator; in transaction services each seat is judged on the automation path, and at least one must be able to automate what its own agents handle.

Vendor portfolio design by body of work is a design position on the shape an outsourced supply base should take once an organization has consolidated a long tail of providers. It holds that the durable structure is a small portfolio partitioned by body of work — customer-facing servicing on one side, asynchronous fulfillment and transaction processing on the other — with two providers in each body. The position explains why the partition follows the work rather than the organization, why two providers per body rather than one or three, and the midterm program that must follow consolidation before a data-driven placement function can be trusted. Front-Office and Back-Office Outsourcing Are Different Bodies of Work establishes that the two bodies carry different outsourcing risks and cannot be priced on one risk model. This page is about how many providers each body should hold, and what the portfolio has to build once the count is settled. It is narrower than Sourcing Strategy Under Cost Pressure, which sets the horizons of a cost program, and than Consolidate, Then Automate: Back-Office Fulfillment, which owns the automation arc for fulfillment work; this page covers provider count and calibration, and touches the automation path only as the criterion a back-office seat is selected on.

The two-by-two

The portfolio has two dimensions and four cells: two bodies of work, each delivered by two providers. Two facts from the companion page carry the count argument: fulfillment work is comparable on outcome measures far sooner than servicing work, and one commercial shape does not fit both.

The partition is deliberately not by region, brand, channel, or client segment. Those are properties of how the organization happens to be arranged, and a portfolio indexed on them has to be rebuilt every time the arrangement changes. Sourcing Design Axes: Node and Client Ownership states the principle — index the model on properties of the work, never on properties of the organization — and the test that goes with it; this page applies it to the provider count. A provider that serves three brands in two regions still serves one body, and the portfolio does not notice when a brand moves between leaders.

Seats within a body

The back-office path a supplier is judged on: break the body of work into its steps, automate what is documented well enough to be checked, and move the supplier's own workforce from doing the work to quality and oversight of the automation. At every step the cost and brittleness of the automation is weighed against keeping a person in the middle of the process.

The reason for two seats differs between the bodies of work, and so does what each seat is judged on.

In customer-facing servicing the two providers are not chosen for complementary specialties. Both are selected on the same properties: language coverage, breadth of markets served, delivery footprint and the scale to absorb variance, because in customer-facing servicing the supplier is the estate's relief valve, the component that absorbs the variance the fixed nodes are too expensive or too steady to carry (The Vendor as Relief Valve). The pair exists for continuity and for competition, the two things one provider cannot supply, argued below. Neither reason requires the two providers to differ in kind; both require them to be comparable.

In transaction services the logic inverts. Asynchronous fulfillment (ticketing, refunds, the processing that follows a booking or a change) is the body of work most open to deeper automation and, in time, agentic handling (Front-Office and Back-Office Outsourcing Are Different Bodies of Work); the arc itself is set out at Consolidate, Then Automate: Back-Office Fulfillment. A supplier in this body is therefore judged less on the rate for the work than on its path to stop doing the work. That path has three steps. First, the body of work is broken down into its pieces and the steps inside each piece: what arrives, what is decided, what is issued (Process Decomposition (L0–L3)). Second, the steps that are specified and measurable are automated one at a time, each passing the tests at The Agentic Handover Gate before its execution moves away from people. Third, the supplier's own workforce transitions from performing the work to quality and oversight of the automation, the role described at The Agent Overseer. At least one of the two seats must be able to carry out that path on the work its own agents handle; a seat that can only perform the work is a seat that will be priced out by the seat that can automate it.

The path is not free and not always right. Each step is a decision rather than a foregone conclusion, and the case against automating a step is made on the same record as the case for it. A step that breaks on every exception is not automated; it is deferred to whoever catches the exception, usually at higher cost and lower visibility than before. A person in the middle is slower and costs more per transaction but absorbs the exceptions the automation cannot. The judgment is made step by step on the transaction record, which is the measurement the back office already has. It is a separate judgment from the handover gate, which tests whether a step is ready to move and not whether moving it is worth the cost; a step can pass all five tests and still be left with a person. A back-office supplier's process discipline and its automation path are therefore one strength rather than two: the property that makes a process safe to place with a supplier is the property that makes it a candidate for automation.

Why two per body

One provider has no comparator; three or more have no volume. Two is the smallest number that produces a comparison.

The provider count is a choice among three failure profiles.

One provider concentrates the risk of a whole body of work in one counterparty and, less obviously, removes the comparator. A single provider is benchmarked only against itself, so a drift in quality or price has no reference point until it is large enough to be visible without one. There is no relief valve inside the body: a site loss has nowhere to overflow except the captive nodes the portfolio was designed to protect. A surge that reaches every client of a shared provider at once — a failure The Vendor as Relief Valve treats as design logic rather than tested finding — has nowhere to go at all. Renewal is negotiated against a transition the buyer has never rehearsed.

Three or more providers restore competition but pay for it several times over. Each additional provider carries its own governance calendar, rubric calibration, and contract. Volumes per provider thin, and the penalty falls on each provider's committed base and contestable share alike. A small provider's committed base sits close to the floor below which it cannot sustain supervision and a training pipeline, and the contestable share it can win is small. Pooling Architecture in Service Workforces derives its dedication penalty per account; the same partition arithmetic applies across providers, an extension this page makes. The multisourcing literature, written for information-technology services, treats coordination across providers as a cost that multiplies with the number of parties and as a governance problem still largely open.

Two providers is the smallest count at which a comparator exists on the same work, continuity exists inside the body, and governance remains affordable. Its failure modes follow from the structure: competitive bidding that drives price below the level at which either provider invests in the work, coordination overhead wherever their processes interlock, and quality that diverges when the two are scored on different frameworks. Each is a governance defect rather than a property of the number two, and the midterm program below is largely what prevents them.

The literature supports the direction of this argument more than its precision. Multisourcing has been studied as the emergence of a supplier portfolio rather than a set of bilateral contracts, and the single- versus multisourcing choice has been shown, on a large dataset of information-technology contracts, to be material to contract outcomes. None of it derives an optimal provider count per body of work, and the study closest to the question points the other way. It finds that committing to a handful of strategic partners may prevent a firm from discovering new suppliers or supply regions, and recommends more intensive multisourcing where supply markets change rapidly. Two answers apply: labor-based servicing supply markets move more slowly than the offshore professional-services markets that study observed, and discovery can run through a standing market scan and periodic tender rather than through live providers. The ceiling is on providers delivering work, not on providers the buyer knows. The practitioner evidence on governance cost is survey-grade: a 2010 advisory presentation cited research in which most companies surveyed reported losing about a quarter of contract value to poor governance, and the underlying study is not separately identifiable. The frequently repeated figure that firms spending two percent or less of contract value on governance lose ten percent or more could not be traced to a primary source and is not relied on here. The industry trend does show direction: in a 2021 advisory-firm survey, 44 percent of enterprises planned to consolidate their provider portfolio. Two-per-body is the deliberate stopping point on that path rather than its endpoint. Consolidate, Then Automate: Back-Office Fulfillment sets two or more suppliers as the floor for consolidated fulfillment work without arguing the count; this page supplies the argument, and the ceiling.

The midterm program

The program that follows consolidation, as a ladder. One rubric with one sampling instrument, and one work-type taxonomy, gate everything above them.

Consolidation produces a portfolio that can be calibrated; it does not produce the calibration. Between the last provider exit and the first allocation made on measured performance sits a program of eight rungs, scoped to labor sourcing; the automation of the work itself belongs to Consolidate, Then Automate: Back-Office Fulfillment, which the seat criterion above draws on and does not replace.

Rung	What it buys	What it needs	Applies first to
One rubric, one sampling instrument per body	A quality score that means the same thing at both providers in a body	A buyer-owned rubric, buyer-designed sampling, cross-provider calibration sessions; a stated detectable difference, and a rule to decline the comparison where the smaller provider cannot support it (Sample Size and Detectable Difference in Quality Measurement)	Front office. Back office is comparable sooner on outcome measures — accuracy, cycle time, rework
Shared work-type taxonomy	A definition of "the same work" that precedes any claim that two providers are doing it	A process classification held by the buyer; the APQC Process Classification Framework is the standard reference	Both, before any comparison
Matched comparison: same work, case mix, tenure	Comparisons that separate the provider effect from the work-mix effect and the tenure effect	Case-mix adjustment of the kind Performance-Based Vendor Allocation Design specifies; tenure recorded per receiving team; comparisons restricted to matched work types. Tenure matching and same-work benchmarking are logical extensions of documented calibration practice, not named practices in the accessible literature	Front office, where the proficiency curve is steep (Speed to proficiency curve)
Outcome-based or gainshare pricing for transaction work	Provider revenue that tracks the buyer's outcome rather than the buyer's headcount	Agreed baselines, a verification method, an attribution rule; suited to transactional work, and most deals remain hybrid because attribution among provider, client, and technology is hard to settle	Back office
AI and model-ownership clauses	Protection of the buyer's data and of the buyer's share of automation gains	Data-isolation and no-training language; a rule for how automation-driven volume decline meets minimum commitments; a productivity-dividend split, as pricing moves toward productivity commitments and gainshare. Unsettled market practice	Both; back office first, where automation lands first
Continuity and relief-valve arrangements	The overflow path between the two providers in a body	A pre-agreed trigger, notice, and price for volume shift on disruption. Routing overflow to a partner under a threshold policy is established in the queueing literature; the contractual clause form between two providers of one buyer is not documented in the accessible literature	Front office, where a disruption is a same-day event
Exit and transition governance	A rehearsed path out for the providers being consolidated, and later for either of the two that remain	An inventory of contract terms and notice periods before proposals are issued; assessment, parallel run, hypercare, closure; knowledge transfer with the leaving provider still paid to cooperate	Both, at consolidation
Tiered governance cadence	Governance intensity matched to what each provider is for	Segmentation of providers into tiers, with scorecard frequency and business-review cadence scaled by tier; the segmentation principle is documented for technology vendors	Both, once the count is settled

Three comments on the table.

Comparability before consequence. The first three rungs exist so that the fourth and fifth can carry weight. An allocation launched without case-mix adjustment ranks the queues rather than the providers, and a provider that can show its queue was harder takes the mechanism's credibility with it — the failure Performance-Based Vendor Allocation Design describes.

The two bodies climb at different speeds. Back office is comparable on outcome measures with a thinner instrument, so it reaches outcome pricing and automation clauses first; front office needs the rubric, the sampling design, and the matched comparison first. One calendar for both bodies either holds the back office back or pushes the front office into consequences it cannot yet justify.

The AI rungs are being written now. The clause types are settled: no-training and data-isolation language is now among the clauses enterprise legal teams routinely redline into service agreements. The clause substance is not. Large contracts priced before agentic delivery changed provider economics are being reopened within twenty-four months of signature, with pricing moving from rate cards toward productivity commitments and gainshare. That renegotiation wave is documented in information-technology services; its extension to process outsourcing is an expectation, not a reported finding. No standard productivity-dividend split and no standard automation-triggered adjustment to minimum commitments has converged, so a buyer writing these clauses today is writing market practice rather than adopting it.

Sequencing

The sequence after consolidation: make the providers comparable before allocating on performance.

The rubric and the taxonomy precede any allocation on performance, because without them there is no defensible statement that two providers were compared on the same work. The contract and notice inventory precedes the consolidation proposal, because the cost and timing of every exit is set by terms already signed. The continuity arrangement precedes the first allocation move, because an irreversible volume shift is a bet rather than an allocation. Outcome pricing follows the taxonomy and a baseline.

The allocation mechanism itself is specified on Performance-Based Vendor Allocation Design. That page also makes the point this one depends on: volume reallocation alone is a weak incentive at any contestable share that preserves provider viability, and it has to be paired with a fee placed at risk against the same index. A two-provider body makes the arithmetic starker, since the contestable pool is split two ways and each committed base must stay above its viability floor. The incentive therefore lives in the fee and the strategic instrument in the volume.

The end state

The portfolio is the supply side of a placement function. Sourcing Strategy Under Imperfect Data describes the two tracks by which an estate decides now while building the foundation for computed placement, and Placement Engine Architecture specifies the machinery that turns placement into a produced answer. Both assume a supply side whose pools can be compared, and a wide portfolio cannot supply that: with many providers on overlapping work, no rubric is shared, no volume is thick enough to measure, and no comparison survives challenge. Consolidation is what makes calibration possible; calibration is what makes placement on data possible; and the two-per-body portfolio is the smallest structure in which both hold. Allocating on performance without the midterm program produces the confident wrong answers the imperfect-data strategy exists to avoid.

Limitations

• The position assumes a body of work large enough to sustain two providers above their viability floors. Below that scale, one provider with a rehearsed exit may be the honest answer.
• Two-per-body is a design position with practitioner support; no study derives an optimal provider count.
• A small portfolio has a discovery cost. Committing to a few partners may prevent a firm from discovering new suppliers or supply regions, and the position accepts that cost in exchange for comparability.
• Several claims — tenure-matched comparison, cross-provider overflow clauses, automation-adjusted minimum commitments, and the per-provider reading of the dedication penalty — are stated as logical extensions of documented results, not as documented practice.
• The back-office seat criterion, the three-step automation path and the rule that at least one seat must be able to automate the work its own agents handle, is a design position of this page. No cited source establishes that a supplier able to automate its own work displaces one that cannot, and the claim that process discipline and automation readiness are one property rather than two is an inference from Process Decomposition (L0–L3), not a reported finding.
• Most of the cited literature is from information-technology outsourcing; its application to labor-based service delivery is by analogy.

Maturity Model Position

In the WFM Labs Maturity Model™, a portfolio with fixed contractual shares and a monthly scorecard per provider is characteristic of Level 2. Consolidation, the shared rubric, and the taxonomy are Level 3 work: the definitions converge and the portfolio becomes measurable before it becomes consequential. The two-per-body portfolio functions as the supply side of a placement function at Level 4, where the operation is an ecosystem of differentiated pools under a multi-objective governance layer and allocation is a planning instrument rather than a reporting artifact. Governing all labor — in-country, captive, contracted, and digital — as one adaptive system allocated dynamically as conditions change is the Level 5 form described on Vendor Operating Model Maturity.

See Also

• Front-Office and Back-Office Outsourcing Are Different Bodies of Work — the companion page: why the two bodies are governed, measured, and priced differently
• Sourcing Design Axes: Node and Client Ownership — the principle of indexing on properties of the work, and the reorganization test
• Performance-Based Vendor Allocation Design — the allocation mechanism, its case-mix adjustment and guardrails, and why volume must be paired with fee at risk
• The Vendor as Relief Valve — what a supplier is for in an estate with captive nodes; the continuity rung applies it between two providers
• Sourcing Strategy Under Imperfect Data — deciding now while building the data foundation the portfolio will feed
• Placement Engine Architecture — the machinery the calibrated portfolio supplies
• Sourcing Strategy Under Cost Pressure — the two-horizon program whose end state is a placement function
• Comparing Delivery Arrangements — the comparison key (work type, channel, decomposition, case mix) and the unit at which a labor source can be compared
• Sample Size and Detectable Difference in Quality Measurement — why the smaller provider sets the detectable difference, and when to decline the comparison
• Vendor Governance Placement — which governance functions sit centrally, in the line, or in neither
• Vendor Operating Model Maturity — the client-side maturity ladder for governing a multi-tier labor model
• Consolidate, Then Automate: Back-Office Fulfillment — consolidating fulfillment work under one process set and commercial shape, then automating it through the handover gate
• BPO and Vendor Management for WFM — the operating mechanics of the provider relationship
• Speed to proficiency curve — the tenure effect the matched comparison controls for
• Build vs Buy and Vendor Governance — the platform-side build, buy, and govern discipline that sits beside labor sourcing
• Business Process Outsourcing — the practice in overview