Sourcing Design Axes: Node and Client Ownership
Sourcing design axes is a framework holding that a sourcing decision in a service operation is determined by two variables — what kind of work it is and who owns the client relationship — and that location is a consequence of those choices rather than an independent one. Two further properties, eligibility and employment relationship, constrain and price the resulting cell but are not themselves decisions.
The framework exists to solve two recurring failures. The first is that sourcing programmes indexed on location drift toward rate comparison — not because quality is ignored, but because rate and speed to seat can be settled in the room while the consequences for capability only appear after a ramp. Under a deadline the fast variable wins, and the quality conversation arrives two quarters later attached to a supplier rather than to the decision. The second is that sourcing architectures indexed on organisational structure have to be rebuilt at every reorganisation, since the structure of a service business changes more often than the structure of its work.
Two properties of the framework matter as much as the axes themselves. How far work can be decomposed is set by what was sold, so an estate is permanently heterogeneous rather than converging on a single optimised design. And quality enters the decision as a gate rather than as a trade, which requires a common measurement instrument even where the standards being held to differ.
For the decomposition of work into nodes, see Service Chain Decomposition and Node Sourcing. For why the flexibility consequences of a sourcing decision matter as much as the cost consequences, see Chaining and Flexibility Design.
Why two axes and not three
Location presents itself as the obvious sourcing variable and is the one most programmes begin with. Treating it as a decision axis is an error, for a specific reason: once the node and the ownership model are fixed, location is largely determined by eligibility, cost, timezone and language coverage. It is where a decision is optimised, not where it is made.
Promoting location to an axis has a predictable consequence, and it is one of timing rather than of rigour. Of the four variables, location and its attendant rate are the only ones that can be settled at the moment of decision; node fitness, elasticity and chainability all reveal themselves after work has been placed and a ramp has run. A model that treats location as a choice therefore selects, structurally, for the variable with the shortest feedback loop — regardless of how carefully quality was discussed beforehand. Demoting it to a consequence is what forces the slower variables into the decision while they can still change it.
A practical test distinguishes the two cases: if the same country can host two different sourcing outcomes, location is not the variable doing the work. In estates with both an integrated servicing centre and a support centre in the same country, this is directly observable.
Axis one — the node
The node is the type of work, defined by its position in the transaction lifecycle and by its proficiency and consequence profile. In a decomposed service chain the nodes are typically:
- Agentic intake — triage, identification, data collection, research and pre-staging
- Transaction authority — the scarce human judgment that verifies and executes the consequential act
- Asynchronous fulfilment — non-customer-facing completion work, deferred by nature and variance-absorbing
Node determines the planning method and the proficiency requirement. It does not determine who is accountable to the client.
Axis two — client ownership
The ownership model records who is accountable for the client relationship. Three recur, and they are distinguished by where accountability sits rather than by where people sit.
| Model | Description | Client accountability |
|---|---|---|
| Integrated ownership | A servicing unit handles all channels end to end and owns the relationship within it — commonly a follow-the-sun or 24/7 construct | Inside the servicing unit |
| Supported ownership | The relationship is owned in-market or onshore; a service centre handles defined channels behind that ownership | Remains with the owning team |
| Delegated execution | Work is executed with minimal or no client contact and no relationship accountability | None |
Ownership determines the escalation path, the governance construct, and what can honestly be promised to a client. It does not determine the planning method.
This axis is frequently left implicit, and leaving it implicit is what produces service commitments no delivery model can support — responsiveness promised by a party that holds neither the capacity nor the authority to deliver it.
The design surface
A sourcing decision is a cell across both axes. Neither axis alone can site work.
- Node alone produces conclusions such as "fulfilment should be low cost" with no answer on who is accountable when it fails.
- Ownership alone produces "move it to the support centre" with no answer on which parts of the work can move — which is how whole-queue migration happens, sending the complex work with the simple work to the location carrying the longest learning curve.
| Integrated ownership | Supported ownership | Delegated execution | |
|---|---|---|---|
| Agentic intake | Natural — the owning unit triages its own work | Natural, and usually the cheapest available improvement | Viable, but errors propagate into a unit with no relationship visibility |
| Transaction authority | Natural — this is the model's core | Ambiguous. Simple-transaction handling is client interaction, so the ownership claim needs to be explicit | Characteristic failure. Arbitraging the specialist buys speed to seat, not speed to proficiency |
| Asynchronous fulfilment | Cost leak. The most expensive ownership model performing the least differentiated work | Workable | Natural — and the node where cost arbitrage is genuinely safe |
Two cells account for most of the value lost. Integrated ownership performing asynchronous fulfilment carries all three nodes at the cost of the most expensive one — a leak invisible in a tier-only view, because nothing in that view names the fulfilment work separately. And delegated execution performing transaction authority is the classic arbitrage failure: capacity bought for speed to seat, which a supplier genuinely delivers, then judged months later on speed to proficiency, which is a property of the work and was never purchasable.[1]
Where ownership is set: offer, configuration, fulfilment
A sourcing architecture is downstream of a commercial one. Work reaches delivery with its ownership model, its constraints and its required standard already fixed, because all three were settled when the service was sold. Three stages sit behind every cell on the surface:
| Stage | What is decided | Who decides it |
|---|---|---|
| Offer | Service tier, responsiveness promised, any location or dedication commitment, price | Commercial |
| Configuration | How the promise becomes operational — entitlement, eligibility rules, routing criteria | Commercial and delivery jointly |
| Fulfilment | Node, employment relationship and location | Delivery |
The ownership axis is an output of the offer stage. Delivery records it and prices it; delivery does not choose it. That is why an undefined service offering leaves every cell with an undefined standard, and why the ownership axis is not a delivery decision dressed as one.
An offer should bind only client-facing work
Commitments made at the offer stage should reach the client-facing node and stop there. How an invoice is produced, a settlement processed or a queue managed are delivery decisions in which the client has no legitimate interest. Where offers extend into fulfilment work, a delivery decision becomes contractual and leaves the estate permanently — it can no longer be pooled, optimised or relocated without renegotiation. Most of the accumulated rigidity in a mature estate arrives this way, one deal at a time.
Price the flexibility, in whichever direction sells
A constraint on delivery is a product feature either way it is expressed: a premium where a client requires dedicated capacity, named sites or on-soil delivery, or a discount where a client permits unconstrained delivery. The mechanism is identical and only the sign differs. In practice the discount framing is materially easier to sell, because it presents flexibility as something the client gains and puts the choice in their hands at the point of sale rather than in a later renegotiation.
Which delivery model a client should get
Where the offer permits a choice rather than dictating one, three client properties place the client-facing work: profitability, lifetime value, and gating — whether the client is served from a shared pool or a designated one. Low profitability, low lifetime value and a shared gate is the natural case for delegated execution; any one of high profitability, high lifetime value or a dedicated gate argues for supported or integrated ownership. Work under an eligibility constraint sits outside the test, because the constraint decides before the economics do.
This is the piece most often missing. A model that derives the standard from what was sold, without a view on what should be sold to whom, can classify decisions but cannot improve them.
Inherited doctrines
Where a business has grown by acquisition, incoherence in the sourcing estate is usually not drift. Each acquired business arrives carrying its own sourcing doctrine — one may have restricted suppliers to back-office processing, another placed almost everything with suppliers and ran them competitively, a third relied on its own centres. Reconciliation rarely happens, because each doctrine was correct for the business that formed it. The result looks arbitrary but is a stratified record of several coherent designs.
The diagnostic move is to date each arrangement to the doctrine that produced it rather than treating the estate as one degraded design. It converts a blame conversation into an archaeological one and separates arrangements that are load-bearing from those merely inherited. Supplier multiplicity is often one such inheritance: it creates commercial tension but trades against continuity and accumulated proficiency, and where it was inherited rather than chosen it should be re-decided.
Decomposition depth is set by the offer
Decomposition is a means of driving efficiency. It is not a destination, and nothing in this framework implies that an estate should converge on fully decomposed work. Treating it as an end state produces two errors: applying decomposition where the offer forbids it, and reading contained delivery as immaturity when it is something that was deliberately sold.
How far a given piece of work can be decomposed is a property of the contract, and in a large book it varies across thousands of contracts simultaneously.
| Containment | What it is | Decomposition available |
|---|---|---|
| Contained | A dedicated team; work does not leave the boundary, whether for regulatory reasons or because dedication was sold | Only inside the boundary, at whatever scale the contained volume supports |
| Partially constrained | Constrained by channel, function or geography — a front and back office division, an onshore-voice commitment, a named-site requirement | Inside the permitted division |
| Unconstrained | No delivery constraint beyond standard terms | Full decomposition available |
All three coexist permanently. The mix shifts with the composition of the book, not with the maturity of the operation. A business that continues to sell dedicated delivery will continue to hold contained work indefinitely, and that is a commercial position rather than an operational deficiency. Any target state expressed as "all work decomposed" is describing a book that does not exist.
The practical consequence is that decomposition depth becomes another attribute of the work — alongside node, ownership and eligibility — rather than a programme with a completion date. It belongs in the same attribute set that portfolios filter on.
Which constraints are priceable
Two kinds, and only one is negotiable.
External constraints — sovereignty, citizenship, clearance, regulatory location terms. Not available to be traded. Their capacity cost has to be absorbed and funded.
Commercial constraints — a dedication commitment, a named-team promise, a channel or site restriction agreed in the contract. These are product features, and they have a cost.
The second is where value routinely leaks. A dedicated team means the client has bought the forfeiture of pooling and chaining benefit for that work, and the forfeiture is measurable: lower achievable occupancy at the same service level, no access to estate capacity during a surge, and a separate ramp. It is very commonly given away rather than priced.
Making it visible does not require renegotiating anything. It requires that the cost of the constraint appear in the deal model at the point the commitment is made, so that dedication is sold as the premium feature it is rather than conceded as a term.
Eligibility — a constraint, not an axis
Some work cannot be delivered from some places: data sovereignty, citizenship or clearance requirements, regulatory location terms, contractual delivery commitments.
Eligibility is not a node and not an ownership model. It is a flag that shrinks the location set available to a cell, and it can attach to any cell on the surface. Because it constrains all nodes of the affected work, a restricted population does not produce a carve-out of one node — it forces a complete parallel chain inside the eligible boundary, carrying isolation, small-scale and extended-ramp penalties that compound. That chain must be planned as a standalone system, with its buffer reported separately so the cost appears as a consequence of the constraint rather than as inefficiency. The full treatment is in Service Chain Decomposition and Node Sourcing.
Eligibility also sets a floor on the addressable base of any cost programme. Volume that cannot move must be excluded before a savings target is accepted, because the reduction required from the remainder rises accordingly. Sizing that floor is a precondition for agreeing a target, not a detail of executing one.
Employment relationship — the pricing attribute
The employment relationship is whether delivery is performed by directly employed staff, by a captive or in-house centre, or by a contracted supplier. It prices the cell and determines whether capacity can be redirected once placed.
| Relationship | Chainability | Why |
|---|---|---|
| Directly employed, or captive centre | Highest | Common employer, systems, taxonomy and skill definitions — connecting into the wider estate is a configuration change |
| Contracted, dedicated | Moderate | Connections are possible but must be specified commercially, and rarely are |
| Contracted, shared or multi-client | Lowest | Capacity is not the buyer's to redirect; the chain terminates at the contract boundary |
The chainability column matters because limited, deliberately configured flexibility captures most of the benefit of full flexibility — roughly two capabilities per resource, connected into a single chain[2][3] — so the cost of connecting a delivery location into the wider estate is small, and the cost of discovering later that it cannot be connected is not.
Employment relationship prices the cell and constrains chainability. It does not predict quality. The opposite is widely assumed. Observed quality differences between an in-house centre and a supplier — or between two teams inside the same supplier — are generally explained by tenure stability and case mix rather than by who employs the staff; a hand-picked, low-attrition supplier team routinely outperforms a high-churn in-house one on the same work. Treat this attribute as a cost and flexibility variable, and look to composition when explaining quality.
Employment relationship is orthogonal to ownership: a captive centre may operate under integrated or supported ownership, and a contracted supplier may in principle hold either, though delegated execution is the common case.
Quality as a gate, and what must be standardised
Quality is not a third axis, for a different reason than location. It is a constraint on the cell, not a dimension of it. Each cell inherits a required standard from what was sold; delivery does not choose it. Where the service offering is undefined, every cell has an undefined standard — which in practice defaults to the highest one anyone remembers promising.
Cost is habitually measured per unit of input — an hour, a full-time equivalent, a contact — and quality per interaction, sampled. The two are not comparable, which is why the argument is rarely settled by either side's evidence. Both belong on the resolved case. Adopting that denominator also reveals how much of the apparent trade is a measurement artifact, since rework, repeat contact and escalation are quality failures that surface as cost once resolutions are counted. The full treatment, including why the resolved case is the only unit that survives arrangements which decompose work differently, is in Comparing Delivery Arrangements.
Standardise the instrument, vary the threshold
Standardising the target is wrong: holding every tier to one quality number either over-serves the lower tier or under-serves the higher one, and differentiated tiers are the point of a service architecture.
Standardising the instrument is essential: definition of a good outcome, sampling method, case-mix adjustment and scoring must be common, or no comparison across sites, ownership models or suppliers is valid and every cross-tier decision remains contestable.
One instrument, different thresholds by tier. That combination is what makes a tiered offering operable rather than rhetorical.
Quality enters as a gate, not a trade
Below the sold standard is not a cheaper option. It is a different product that was not sold. Options failing the standard are therefore excluded rather than scored lower, which removes the retrospective conversation in which a saving is banked and a service decline is discovered separately.
Above the gate, cost per resolved case decides — a measure that already absorbs the rework consequences of marginal quality.
A constraint that is rarely stated: the resolution of the quality instrument bounds the tier spacing that can be sold. Where the instrument cannot detect a difference smaller than several percentage points, two tiers separated by less than that cannot be demonstrated to differ, and the premium tier is a promise with no available evidence. Instrument resolution is therefore a constraint on the service architecture, not only on supplier management.
Resolution is set by observation count, not coverage
Measurement resolution is habitually discussed as a percentage sampled. It is in fact set by the absolute number of interactions scored, which is why complete monitoring transforms a large unit and barely moves a small one — and why a comparison is limited by whichever unit is measured least. The arithmetic, with worked tables and runnable code, is at Sample Size and Detectable Difference in Quality Measurement.
Vary the threshold by tier, never by location
The threshold may legitimately differ between service tiers, because tiers are sold as different products. It should never differ by delivery location.
A location-differentiated target states institutionally that delivery from one place is expected to be worse, and three consequences compound. It hands the commercial function a standing veto — no cost advantage answers the question of why a client should be placed somewhere carrying a lower promise. It degrades the aggregate mechanically as that location's share of volume grows, regardless of whether any unit's performance changed. And it contradicts a location-agnostic service promise, which most organisations with a distributed estate make in some form; the promise and the goal architecture cannot both be true.
Where units genuinely handle work of different difficulty, apply one target to case-mix-adjusted scores rather than different targets to raw ones. Different raw targets create a rational incentive to refuse complex work — the mirror image of the commercial veto.
Targets set before baselines
A target chosen before the instrument exists is a guess that will be defended as a commitment. Publish it as provisional, benchmark it externally, and withdraw it explicitly rather than quietly missing it.
Chain-level quality is the commitment
Chain-level quality is contractual — the outcome for the resolved case, whatever path it took. Node-level quality is diagnostic — it locates a failure, and reporting it as the commitment produces delivery that hits every node target while failing the customer. How this plays out when two arrangements decompose the work differently is set out in Comparing Delivery Arrangements.
Independence from organisational structure
The most consequential property of this framework is what it deliberately excludes.
Service businesses organise their client-facing books along dimensions that shift — region, client size, industry vertical, brand — and those dimensions frequently overlap rather than partition cleanly. A sourcing architecture indexed on any of them must be rebuilt whenever the organisation changes, which is more often than the work itself changes.
The principle: index the model on properties of the work, never on properties of the organisation.
| Stable — safe to build on | Volatile — never build on |
|---|---|
| Node · ownership model · eligibility · language · timezone · complexity · employment relationship | Region · client size or segment · vertical · which leader holds which book |
The implementation follows directly. Every unit of work carries attributes; a leader's portfolio is expressed as a filter over those attributes rather than as a fixed assignment. A reorganisation then changes the filters and leaves the architecture untouched. This is the same principle as attribute-based supply routing — where eligibility is computed rather than encoded in static queue membership — applied to ownership rather than to routing. See Next Generation Routing.
The test: if a reorganisation would require the sourcing architecture to change, the architecture was indexed on the wrong variables.
Capability ownership and outcome ownership
A governance dispute follows from this framework in every estate that mixes internal and contracted delivery: who owns supplier performance management. It persists because both obvious answers are wrong. If the sourcing function owns measurement and judgment, the business discounts its findings as self-interested. If each client-facing unit owns both, definitions diverge and no cross-tier decision can be defended.
Separate the two:
- Capability ownership — the sourcing architecture, commercial constructs, allocation, the measurement instrument and its integrity, and the ramp mechanics of onboarding, offboarding and scaling. These require consistency across the estate to be worth anything, so they sit with the function holding the supplier relationship.
- Outcome ownership — whether performance is acceptable for a given book of clients, and the escalation when it is not. This sits with the unit accountable to those clients.
Whoever owns the instrument should not own the verdict. One governance forum serves both. A forum per client-facing unit is the arrangement that quietly destroys comparability, because each will drift toward definitions that suit its own book and no cross-tier comparison survives the divergence.
A further reason to hold the instrument centrally: standard supplier contract forms can coordinate the staffing level while leaving delivered quality below the system optimum,[4] so quality shortfall is partly a property of the construct rather than evidence about the supplier — and diagnosing that needs comparison across suppliers on a common instrument.
The same reasoning places asynchronous fulfilment work. That node is defined by the absence of client ownership, so a client-owning unit is the wrong home. What matters more than which function holds it is that it is held once — fragmenting it destroys the pooling gain that makes it inexpensive.
Failure modes
| Failure | What it looks like | Correction |
|---|---|---|
| Location as an axis | The decision collapses to rate comparison | Demote location to a consequence; decide node and ownership first |
| Ownership left implicit | Service commitments no delivery model can support | State the ownership model for every cell |
| Undecomposed migration | A whole queue moves; quality falls months later and is attributed to the receiving site | Make the node, not the queue, the migration unit |
| Eligibility treated as a carve-out | One node is separated; the rest of the constrained chain has no home | Design a complete parallel chain inside the eligible boundary |
| Org-indexed architecture | Every reorganisation triggers a sourcing redesign | Express portfolios as filters over work attributes |
| Instrument and verdict held together | Findings discounted, or comparability lost | Separate capability ownership from outcome ownership |
| Fulfilment fragmented by book | Pooling gain never materialises | Hold it once |
| Decomposition as a destination | Decomposition pushed where the offer forbids it; contained delivery read as immaturity | Treat decomposition depth as a property of the contract |
| Dedication given away | A commercial constraint conceded as a term rather than sold as a feature | Price the forfeited pooling and chaining benefit at the point of commitment |
| Offers binding non-client-facing work | Delivery decisions become contractual and leave the estate permanently | Bind commitments to the client-facing node and stop there |
| Quality traded rather than gated | A saving banked, a service decline discovered separately | Exclude below-standard options; decide on cost only above the gate |
| Node quality reported as the commitment | Every node target met, the customer still failed | Hold the commitment at chain level; treat node quality as diagnosis |
| Threshold varied by location | Commercial function declines placement; the aggregate declines as the location's share grows | Vary by tier only; adjust for case mix and apply one target |
| Constrained pool at estate occupancy | Service fails there first, because it has no chain to draw on | Plan it as a standalone system; report its buffer separately so the cost reads as a consequence of the constraint |
| Resolution discussed as coverage | Small units judged on noise; full monitoring expected to fix it, and it does not | Report observation count and detectable difference beside every score |
| Inherited doctrine read as drift | The estate looks arbitrary, so remediation targets the wrong arrangements | Date each arrangement to the doctrine that produced it |
| Target set before baseline | A guess defended as a commitment long after the evidence arrives | Publish as provisional; benchmark externally; withdraw explicitly rather than quietly missing it |
Maturity Model considerations
Maturity here is not how much of the estate has been decomposed. It is the capability to decompose where the offer permits, leave work contained where it does not, and know what each constraint costs.
- Levels 1–2. Sourcing decisions are location decisions evaluated on rate. Ownership is implicit, work moves as whole queues, and quality is measured per node with no chain-level view.
- Level 3. Work types and ownership models are named but decided together; the architecture tracks the organisation chart and delivery constraints are known without being priced.
- Level 4. Node and ownership are decided independently and location is derived. Decomposition depth is set per contract, eligibility carries its own capacity treatment, and quality operates as a tier-specific gate on a common instrument.
- Level 5. The estate is managed as a permanent mixture of contained and decomposed delivery. Commercial constraints carry an explicit price, portfolios are filters over work attributes, and instrument resolution is known and bounds what the business commits to selling.
See Also
- Service Chain Decomposition and Node Sourcing
- Comparing Delivery Arrangements
- Sample Size and Detectable Difference in Quality Measurement
- Chaining and Flexibility Design
- Supply Elasticity in Workforce Planning
- Conservation of Labor
- BPO and Vendor Management for WFM
- Vendor Operating Model Maturity
- Conservation of Labor
- Next Generation Routing
- Three-Pool Architecture
- Speed to Proficiency Curve
References
- ↑ Kim, Y., Krishnan, R., Argote, L. (2012). The learning curve of IT knowledge workers in a computing call center. Information Systems Research.
- ↑ Jordan, W.C., Graves, S.C. (1995). Principles on the benefits of manufacturing process flexibility. Management Science.
- ↑ Wallace, R.B., Whitt, W. (2005). A staffing algorithm for call centers with skill-based routing. Manufacturing & Service Operations Management 7(4), 276–294.
- ↑ Ren, Z.J., Zhou, Y.-P. (2008). Call center outsourcing: Coordinating staffing level and service quality. Management Science 54(2), 369–383.
