Operational Policies / Supply & Demand Trade-offs
April 2024 – August 2025
Finding Compromise in a 2-Sided Marketplace
Across three separate initiatives spanning promotions, compensation, and cancellations, I kept running into the same structural tension: what feels fair to a customer and what feels fair to a catering partner are rarely the same answer, and ezCater sits in the middle of both.
Rather than solve each policy question in isolation, this body of research built a repeatable pattern for resolving that tension: default to a standardized policy for consistency and trust, but leave enough room for partner-level judgment that no one feels like the system ignored their specific situation.
dyadic survey design
sentiment triangulation
competitive benchmarkin
situational sentiment testing
cross-functional recommendations
financial impact analysis
I was responsible for...
The Ask
Three separate policy questions all asked the same open question: when ezCater sits between a customer and a catering partner, who decides what's fair, and how much should be standardized versus left to individual judgment?
Approach
Each initiative started the same way: survey customers about what they expect, survey partners about what they're willing to offer or accommodate, then look for where the two groups agree, diverge, or land in the same place without ever comparing notes.
Ask customers what they expect,
ask partners what they're willing to offer
Lay the two side by side without letting either group see the other's answers. That method applied to Promotions, Compensation, and Cancellation policies. While converging answers defined policies, misalignment showed gaps in trust and understanding.
Promotions: what's a fair incentive, and who should fund it?
Testing high-fidelity designs for a new promotions flow surfaced a bigger question than the interface: what should we even call this, and whose money is on the table? Both partners and customers were skeptical; for cost-conscious partners they wanted clear ROI attribution and control. For Customers past experiences built resentment towards promotions with restrictive fine print, making transparency and simplicity essential.
-
Customers preferred "promos" or "deals", associating "promos" with professionalism and "deals" with a straightforward discount.
-
Strong alignment in dollar-for-dollar expectations were a strong signal that promotions were mutually-beneficial.
RECCOMENDATION
Lead with straightforward dollar/percent-off offers since they're familiar and easy to implement, explore value-add promotions as a second phase, and treat ezRewards as a lower-risk, high-trust alternative worth building out further.
Auto-compensation: when something goes wrong, who decides the fix?
ezCater's compensation process relied on a long manual handbook, leaving support agents to eyeball severity case by case. That inconsistency showed up on both sides: partners felt customers prematurely compensated, while customers experienced unpredictable outcomes for similar issues.
-
Partners were largely opposed to full automation (63%), wanting a review-and-dispute window.
-
Customers, responded positively to automation as long as a human was still reachable.
-
While partners measured severity in terms of dollar value, customers thought in terms of whether the event itself was salvageable.
RECCOMENDATION
A severity-based compensation matrix that gives partners a review window before payout, keeps a human escalation path always available, and reframes "auto-compensation" language carefully, since the word "auto" alone raised partner suspicion regardless of the actual policy.
Cancellation & modification: how do you set a policy without punishing normal life?
Only 12% of catering partners had a cancellation policy actually configured, in part because setting one required contacting support manually. The cost of that gap was concrete: $824K spent conceding cancellation fees in a single year and3M support-agent minutes on cancellations and modifications.
-
Partners preferred a standardized cutoff and fee ("I would prefer ezCater to be the bad guy") for consistency and reduced management overhead.
-
Customers were reflexively opposed to cancellation fees in the abstract but softened considerably once fees were framed around transparency, situational context, and protection against repeat offenders.
RECCOMENDATION
A standardized default cutoff and fee structure, configurable by partners through self-service tooling rather than a support ticket, with a retroactive fee-waiver option and clear upfront messaging at checkout so customers aren't surprised by a policy they never saw.
Results
The pattern underneath all three
What ties these together isn't the policy area, it's the resolution shape. In every case, the operationally sound answer wasn't "side with the customer" or "side with the partner." It was: set a sensible standardized default, make it visibly consistent, and preserve a configurable or human-reachable escape hatch for the cases that don't fit the default. That pattern, tested three separate times across three separate teams, became the de facto framework for how ezCater approaches trust and fairness in a two-sided marketplace.
Promotions:
-
Validated $/% discount guidance against closely-aligned customer and partner data
-
Fed directly into the 2026 pricing and rewards-redemption revisit
Auto-compensation:
-
Informed the severity-based compensation matrix piloted with support agents
-
Shaped the decision to preserve partner review windows and human escalation, avoiding a fully automated rollout that partner sentiment didn't support
Cancellation & modifiaction:
-
Directly informed ezCater's standardized cancellation policy
-
Cancellation-related support contacts fell 43% against the prior $824K payout and $2.7M cost baseline
So What?
Three teams brought me three different policy problems over sixteen months, and each one arrived looking like a negotiation between what customers wanted and what partners could tolerate. The real work wasn't picking a side. It was finding the shape of a policy, standardized enough to build trust, flexible enough to not feel punitive, that both sides could actually live with. That shape held up across promotions, compensation, and cancellations, three different teams, three different stakes, the same underlying answer.
