Bad bedfellows.

Partnership announcements describe shared ambition. Operating failures reveal separate economics.

A startup forms strategic relationships because it cannot own every capability. A factory makes the garment. A battery supplier makes the vehicle possible. An automaker makes an infrastructure network useful. A distributor reaches customers the startup cannot. The dependency is rational.

It becomes a bad-bedfellows failure when:

1. the counterparty controls part of the startup’s core customer promise;
2. the outcome is less valuable, more costly or lower priority to the counterparty;
3. capacity and quality are assumed from reputation or intent rather than proved;
4. integration, exclusivity or volume hardens before performance is observable; and
5. replacement takes longer than the company can survive.

The sequence is:

core promise → external control → different payoff → dependence hardens → performance miss → strategic captivity.

Thomas Eisenmann uses “bad bedfellows” broadly within a multi-stakeholder theory of startup failure ^{[1] [2]}. This manual uses a narrower backlog boundary: bilateral partners, suppliers, distributors and strategic dependencies. It excludes a dominant platform applying general rules across an ecosystem, and it excludes a single customer using concentration power. Those require different controls.

The counterparty does not need to be malicious. In most failures it is behaving rationally. A large manufacturer prioritizes profitable volume. A channel partner sells the easiest product. A strategic investor preserves its option. A supplier near insolvency protects immediate cash. The startup fails because its contract and operating design assume shared purpose where only conditional overlap exists.

The basic rule is brutal: you may outsource production, distribution or infrastructure; you cannot outsource ownership of the customer consequence.

That means leadership must know how the partner makes money, what causes it to deprioritize the relationship, which performance data arrives before failure, who can intervene and how the promise continues after termination.

## The logo creates false proof

A recognized partner lends legitimacy. Customers, candidates and investors infer diligence, capacity and future distribution. The startup begins counting the relationship as an asset before real work passes through it.

This is the strategic-name halo. The brand may prove that a contract was worth signing. It does not prove that the startup is material to the partner, that operations can meet the specification or that executive enthusiasm has reached the people who allocate capacity.

The first proof is not an announcement. It is a completed operating cycle under representative load.

## The payoff is asymmetric

Write the relationship from each income statement.

For the startup, the dependency may determine survival. For the partner, it may be experimental revenue, utilization for an idle line, a public-relations story or an option on a future market. When conditions tighten, each side protects what is material to itself.

Alignment language hides this asymmetry:

• “strategic” may mean optional;
• “preferred” may not mean prioritized;
• “exclusive” may bind only the startup;
• “joint go-to-market” may have no seller quota;
• “dedicated capacity” may disappear behind a larger order; and
• “best efforts” may have no observable threshold.

A good contract cannot make a small company important. It can expose the difference early and preserve routes when incentives diverge.

## Integration creates a private switching tax

Dependency accumulates through tooling, certification, data models, training, customer commitments, inventory and workflow. None is fully represented by a termination clause.

The true switching time is:

decision + data extraction + alternative qualification + technical change + regulatory or customer approval + inventory transition + ramp.

If this exceeds the cash or service-continuity clock, the startup is captive even when it can legally terminate tomorrow.

## Failure is visible to the customer first

Weak dependencies have poor observability. The supplier reports shipment, not defect distribution. The distributor reports aggregate pipeline, not stage-level buyer evidence. The partner controls customer data. The startup learns of failure through churn, complaint or missed revenue.

Service-level agreements are useful only when measures arrive early enough to change behavior. A quarterly penalty does not protect a weekly customer promise.

## Exclusivity arrives before repeatability

Exclusivity can fund dedicated investment and prevent opportunism. Granted before quality, volume and economics are proved, it removes the learning alternatives that reveal whether the chosen partner is good.

The correct sequence is:

pilot → measured performance → bounded tranche → replacement test → conditional exclusivity.

Startups often reverse it because the partner makes exclusivity a condition of the announcement or investment. The company trades operational option value for social proof.

Create a dependency exposure map for every counterparty that can stop product availability, regulatory permission, customer access, delivery or more than one operating cycle of revenue.

## Promise share

What exact customer outcome sits outside the company? Avoid spend percentages. A cheap authentication service can control all access. A single certified component can stop an expensive product.

Classify the dependency:

• component: one replaceable input;
• capacity: volume or geography requires the party;
• permission: certification, license or approval flows through it;
• distribution: the party controls customer access;
• system: the product works only with the partner’s installed base; or
• identity: the offer’s credibility depends on the association.

## Incentive alignment

Record the counterparty’s revenue, gross margin, strategic benefit, resource cost and downside. Estimate how important the startup is relative to alternatives. Identify the event that would rationally change priority: larger customer, capacity shortage, new leadership, capital pressure, acquisition or product strategy change.

Ask the partner to validate the model. Polite disagreement is useful evidence.

## Capacity and quality proof

Demand operating evidence at the required specification and volume. Audit where appropriate. Inspect exception behavior, not only happy-path output. Who pays for rework? How quickly does a defect reach decision-makers? Does quality worsen when volume rises?

## Observability

List the leading data the startup receives directly:

• capacity allocation;
• queue and cycle time;
• defect or return rate;
• inventory and supplier health;
• seller activity and pipeline stage;
• service availability;
• financial or solvency signals; and
• change notices.

If data is controlled entirely by the party being measured, add an independent verification path.

## Substitutability and switching time

Name a qualified alternative. “Many vendors exist” is not enough. Has one signed confidentiality terms, received the specification, produced a sample, passed security or regulatory review and priced the work?

Measure switching time through customer continuity, not contract exit.

## Failure tail

What survives termination? Inventory, warranties, customer support, data retention, tooling, licenses, public claims, stranded employees and prepaid cash can outlive the relationship. The startup owns this tail unless rights and resources say otherwise.

Score the dependency:

• Green: incentives understood, performance proved, leading data visible and alternative qualified.
• Amber: one material uncertainty exists, exposure capped and correction funded.
• Red: the core promise depends on unproved capacity or replacement exceeds the failure clock.
• Black: exclusivity or insolvency risk prevents a survivable switch.

Quincy Apparel set out to improve fit for professional women using a sizing system closer to men’s suiting. Eisenmann’s case describes founders with customer insight and strong early interest but little apparel-manufacturing experience. Their small, unusual orders were difficult for factories, which had more attractive customers and limited reason to prioritize Quincy. The company also raised less capital than planned and faced pressure to grow ^[1].

This is why the bad-bedfellows frame matters. The factories were not villains. Their economics differed.

Quincy needed flexible production, unusual sizing, reliable quality and small initial runs. A factory generally benefits from predictable, larger orders and standard work. The more Quincy needed the supplier to behave like a product-development partner, the less the relationship resembled ordinary contract manufacturing.

The founders’ lack of deep manufacturing experience compounded the mismatch. They were less able to estimate iteration time, recognize factory constraints and distinguish a supplier problem from a design-for-manufacture problem. Capital pressure reduced the room to learn. Investor growth expectations increased the volume of an unproved operating system.

The case contains several causes and should remain that way. Underfunding alone is incomplete. Inexperienced founders alone is incomplete. A supplier miss alone is incomplete. The failure emerges from interaction:

novel product system + low-volume buyer + conventional factory economics + limited manufacturing judgment + growth pressure.

The alternative was not necessarily vertical integration. Eisenmann suggests that tighter alignment—potentially including an equity relationship with a manufacturer—might have changed the partner’s payoff ^[1]. Other routes include a specialist development factory, paid pilot capacity, narrower assortment or a manufacturing leader with authority to redesign the product around production facts.

The decisive artifact would have been a factory dependency warrant: required tolerances, paid development work, representative pilot, capacity reservation, defect and rework economics, change-control process, leading production data, tooling ownership and a qualified second source. The warrant would reveal whether the factory was a vendor or a co-developer. Quincy appears to have needed the latter.

## Fisker and A123: a critical component meets supplier failure

Fisker Automotive and A123 Systems entered a long-term battery supply agreement with exclusivity and defined re-sourcing provisions ^[3]. The relationship was load-bearing: no high-voltage battery, no vehicle.

A123 later disclosed substantial costs associated with defective battery modules and entered bankruptcy proceedings ^[4]. A later bankruptcy-court record concerning Fisker states that A123’s production stop left Fisker without high-voltage batteries and that Fisker ceased production ^[5].

Fisker had other serious problems, including capital needs, vehicle issues and market execution. The case does not establish that A123 alone caused its failure. It establishes a hard dependency fact: the company’s product promise crossed through one financially distressed critical supplier, and replacement could not arrive inside the operating clock.

Contractual re-sourcing rights were necessary but insufficient. A legal right does not create validated tooling, certification, software integration, inventory and production yield at another supplier. The usable control is a warm second source or enough cash and time to qualify one.

The case also shows correlated startup risk. Two young companies depended on each other while both required capital and technical execution. The partnership did not diversify risk; it joined the failure surfaces. Counterparty diligence should therefore test not only product quality but liquidity, financing dependence, customer concentration and ability to fund a recall or ramp.

## Better Place and Renault: a joined system with thin ecosystem participation

Better Place built battery-switching infrastructure for electric vehicles. Renault developed the Fluence Z.E. with a switchable battery, making the vehicle and station network a joined proposition ^[6]. The idea required several complements to arrive together: compatible cars, station density, consumer adoption, capital and regulatory support.

Contemporaneous reporting described limited vehicle sales, heavy infrastructure spending and little participation from other automakers ^{[7] [8]}. Renault was a real partner and produced a compatible vehicle. The strategic dependency problem was that one automaker and one model could not by themselves create a broad interoperable ecosystem quickly enough to repay network costs.

Neither side had identical economics. Better Place needed high utilization across its installed network. Renault sold vehicles across a broad portfolio and could not make adoption occur by commitment alone. The startup controlled infrastructure but not the range, pace or customer appeal of compatible vehicles.

The failure mechanism is not “never partner with a large company.” It is to treat a bilateral agreement as proof of a multilateral market. Better Place needed milestone gates tied to vehicle availability, binding fleet demand, station utilization and additional manufacturer participation before expanding the infrastructure footprint.

The dependency warrant would have named the system minimum: compatible vehicles in market, committed volume, geographic density, customer switching behavior, additional automaker path and the capital required until utilization. Without those joined thresholds, each partner could fulfill its narrow promise while the whole proposition still failed.

The decision event is a second miss against a load-bearing service or supply commitment.

## Branch one: protect the announcement

Leadership describes the miss as temporary and preserves scope. The public partnership becomes an identity commitment. Customer exposure grows while operating evidence weakens.

## Branch two: escalate socially

The founder calls the partner’s senior sponsor. Executive attention can unblock one event. It cannot replace daily capacity, process or economics. If the relationship works only through escalation, the operating contract does not work.

## Branch three: grant more exclusivity

The startup offers volume, geography or term to become more important. This can align investment after proof. Before proof, it increases captivity and removes comparison.

## Branch four: switch in panic

Leadership terminates before securing data, inventory, tooling, approvals and customer continuity. The new provider starts cold and the attempted cure produces another outage.

## Branch five: contain and qualify

The company caps new exposure, invokes data and service rights, protects customers, begins a parallel qualification and renegotiates economics from observed facts. Survival depends on replacement completing before the failure clock.

First, stop deepening the dependency. Pause exclusivity expansion, unsupported customer promises, new integrations and volume commitments. Preserve current service while creating options.

Issue a dependency warrant:

Customer promise. The exact outcome the counterparty controls.

Counterparty economics. Revenue, cost, resource burden, strategic value and event that changes priority.

Performance proof. Required quality and capacity, representative pilot, actuals and exceptions.

Leading information. Data frequency, source, audit and change notice.

Control rights. Service levels, remedies, step-in, tooling, data portability, subcontracting, business continuity and insolvency rights.

Replacement route. Named alternative, qualification work, cost, approvals and transition inventory.

Failure clock. Time until customers, cash or permission fail.

Decision date. Renegotiate, dual-source, internalize, narrow or exit.

Build the replacement before announcing it. Secure specifications, data exports, samples, security review, tooling access, certification and transition inventory. For a distributor, rebuild direct customer knowledge and pipeline evidence. For a strategic integration, document the interface and test degraded operation.

Renegotiate incentives, not sentiment. The partner may need a capacity reservation, minimum commitment, development fee, better forecast, equity, margin or narrower specification. The startup may need audit rights, credits, termination assistance or non-exclusivity. If the economics cannot support the promised behavior, stronger relationship language will not.

Create a joint operating cadence below the executive sponsor. Name owners for forecast, quality, incidents, change control and escalation. Review leading data. Require a written recovery plan after each material miss and an automatic strategic review after the second.

Protect customers. Decide which promises can still be met, which require a controlled migration and which should stop. Do not blame the supplier publicly before legal and operational review. The customer contracted for the startup’s outcome.

Treat counterparty distress as a company incident. Monitor financing, layoffs, ownership change, litigation, recalls, delayed payments and unusual requests. Use counsel for insolvency, intellectual property, termination and step-in questions.

If no replacement fits the clock, narrow the company. Reduce assortment, geography, service level or customer segment until the remaining promise is internally controllable. A smaller truthful product is better than a broad promise held hostage by an external balance sheet.

Bad bedfellows are not bad companies. They are relationships whose actual payoffs cannot support the dependence placed on them.

The failure equation is:

core promise × incentive mismatch × low observability × switching time = strategic captivity.

Map every external party that can invalidate the customer outcome. Model its economics. Prove capacity under real load. Get leading information. Qualify a replacement. Make exclusivity conditional on performance and keep the failure tail funded.

Partnership can be the fastest route to market. It can also place the company’s survival inside another organization’s priority queue. The founders’ job is not to eliminate dependence. It is to keep dependence governable.

The customer will hold the startup responsible no matter whose factory, battery, seller or infrastructure failed. Design the relationship accordingly.