Newest-buyer COD failure
Buyer share is not order share. Cohort order frequency is missing.
01 / THE CASH-ON-DELIVERY DILEMMA
SUPPLIED CASE → DERIVED ARITHMETICAN ORDER
ISN’T A SALE.
UNTIL IT
ARRIVES.
THE PRINCIPLE / KEEP CASH ACCESSIBLE
Cash on delivery gives millions of buyers a way in. But a checkout is only a promise. The sale happens at the doorstep.
The recommendation: a courier-acknowledged handover and an owned recovery path. Test selected serviceable NCR and Mindanao lanes before committing to scale.
FICTIONAL CASE FIGURES.PLANNING INPUTS TO VALIDATE.
29%
OF COD ORDERS FAIL TO
COMPLETE DELIVERY.
30.45m
FAILED COD ORDERS
EVERY 30-DAY MONTH
₱3.35bn
MONTHLY PLATFORM
FAILURE COST BURDEN
₱1.52bn / REVERSE LOGISTICS
₱1.83bn / ALLOCATED OVERHEAD
Allocated overhead can release capacity.
It is not automatically cash savings.
Source: supplied case and cost exhibit ↗. 5m orders/day × 70% COD × 29% failure × 30 days. All operating figures are fictional; intervention effects on this site are assumptions to test.
CHECKOUT → DISPATCH → DOORSTEP
THE DECISION STARTS BEFORE DISPATCH.
Separate exhibit views, not an observed funnel.
All figures are fictional. Click a ticket to inspect its source.
02 / THE DIAGNOSIS
DIFFERENT FRICTIONS.
ONE PRIMARY TEST.
Lead with the handover loop. Use geography and carrier differences to choose an operational test; use apparel’s burden to select the supporting confidence test.
Category burden / month
Raw exhibit allocation. Assumes category failure rates refer to COD. Burden = orders × COD share × failure × ₱110. It is not fully addressable savings.
A DIFFERENT REGIONAL LENS
Geography changes the problem.
Absolute burden stays substantial.
NCR has an 18% supplied failure rate; Mindanao has 35%. Their raw-exhibit monthly burdens are ₱653m and ₱766m. Select serviceable lanes in both to test urban and provincial operating fit; this is a pilot choice, not an observed joined population.
Creator-live failure in campaigns
At equal 100k COD orders, the descriptive 6 pp gap is 6,000 failures / ₱660k burden. It does not prove a campaign effect or cohort size.
XYZ Express COD failure
At 3.2 days, it beats faster A&B's 29%. Reliability merits a lane-level test.
Source: case exhibits 3.2, 3.3, 3.5, 3.6 and 3.7. Descriptive differences are not causal intervention effects. Regional and category allocations do not reconcile with headline totals.
THE STRATEGIC CHOICE
Protect delivered GMV and buyer access.
Test contribution after actual intervention costs.
03 / THE RECOMMENDATION
KEEP COD.
CLOSE THE LOOP.
Lead with a preference the courier acknowledges and a failed attempt someone owns. Use one existing 3PL in selected serviceable NCR and Mindanao lanes. Keep standard COD service available.
WHY THIS FIRST?
Choose the experiment.
Make the uncertainty visible.
Handover
Specific carrier actions and named recovery ownership preserve COD access. First verify actionability, cost and enrollment.
Apparel confidence
The largest raw category burden supports a separate recap experiment. Its effect on buyer commitment is unknown.
Prepaid
Address COD dependence, but approve safe payment conversion and incentive economics before development or enrollment.
The exhibits locate burden; they do not identify the largest preventable cause. Handover is the preferred controllable hypothesis, conditional on carrier feasibility. Different modeled break-even values are not a causal ranking.
OPTIONAL BUYER INPUT→
COURIER-ACKNOWLEDGED ACTION→
OWNED RECOVERY↗
01
LEAD TEST / A HANDOVER BOTH SIDES CAN ACT ON
A preference.
A supported action.
A named owner.
Offer optional receiver, landmark and carrier-supported preferences before the route cutoff. Confirm that a designated courier or dispatcher has accepted the supported action. If a delivery attempt fails, verify its reason and route the next serviceable recovery to an exception owner.
No response keeps standard service. A preference is not a nationwide exact-time promise. Missing input or acknowledgment never triggers an automatic cancellation or COD ban.
ONE PARCEL / PROPOSED DESIGN DEMONSTRATION
A preference becomes
an accountable action.
Illustrative screens for parcel PH-001. Maria is a fictional authorized receiver. This is a proposed design, not an observed app or live carrier connection.
Choose.
- Authorized receiver
- Maria · nominated by buyer
- Landmark
- Blue gate
Optional input before cutoff.
Accepted.
Dispatcher accepts the supported receiver instruction before cutoff. The rider sees it in the manifest.
ACCEPTED BEFORE CUTOFFOwn the outcome.
If an attempt fails, verify the reason with a buyer confirm/dispute path. A named lane exception owner selects a supported recovery.
Record final delivery, settlement and actual costs.
Accepted path: a supported preference reaches an operator. It does not guarantee delivery.
What counts as acknowledgment and who owns it?
The manifest/driver view must carry the order ID, preference version, supported option, route/date, cutoff, accepted/fallback state and designated operator. An API receipt alone does not count. 3PL ops resolves unaccepted preferences before cutoff; product shows the standard fallback when an option is unavailable. Carrier and route reassignments are tracked.
The country operations lead is accountable. Product owns preference/fallback, 3PL ops owns acknowledgment and exception routing, seller ops owns the apparel recap, payments owns funding/refund information, and analytics/finance owns assignment and realized contribution.
ONE REAL PRECEDENT / AN EXPLICIT LOCAL ADAPTATION
DHL delivery controls,
adapted to Philippine COD.
DHL Malaysia Express ↗ offers date, recipient and location changes. Borrow buyer control; offer only options the participating Philippine 3PL can execute, with a traceable acknowledgment and fallback.
Express coverage, pricing and infrastructure do not establish local COD feasibility or a failure-reduction percentage. TikTok already documents verification and reattempts ↗; the improvement is the closed action loop, not another generic reminder.
02
SUPPORTING CONTROLLED TEST / APPAREL
Buy with confidence.
For a separate eligible apparel cohort, show a short size/variant, total-price and delivery recap. Resolve uncertainty before seller handover. Track early cancellations separately from rescued sales.
Apparel’s ₱1.189bn raw-exhibit burden justifies the cohort. It does not prove recoverability or a buyer × route × region intersection.
03
SUPPORTING CONTROLLED TEST / WALLET-READY BUYERS
Make first prepaid use useful.
Propose a pre-dispatch prepaid option for eligible unshipped COD orders, with funding/refund clarity and a capped ₱5 test. Do not release prepaid development or enroll its cohort until product/payments approve safe conversion; this is not an observed app feature. Confirm settlement and the carrier’s cash-due flag before release. Failed conversion keeps COD.
The 20% incremental switch target and PHP5 incentive are hypotheses. If feasibility fails, remove the cohort and recost/reapprove allocation, budget and impact. Measure conversion against control and include rewards to buyers who would already have prepaid. No partner subsidy is assumed. Shopee’s doorstep QR ↗ remains payment on delivery. GCash funding fees ↗ describe buyer friction, not the assumed merchant fee.
Defer: COD bans, blanket faster delivery, a new locker network and voucher-threshold redesign. Voucher redemption, campaign exposure and settlement data are missing. Keep the pilot focused on a measurable operating change.
All effects, costs and eligible volumes are planning inputs. No firsthand app observation or interviews are claimed. Download the full recommendation ↗.
ONE ORDER / FOUR STAKEHOLDERS
WHAT CHANGES
ON THE OTHER SIDE?
Move between the buyer, seller, courier and platform. Compare the case problem with a proposed response. These are design hypotheses, not measured outcomes.
01 / BUYERDRAG / SWIPE / USE THE ARROWS
04 / THE DECISION LAB
WHAT MUST
THE TEST PROVE?
Start with the lead handover test. Change the paid-completion gain or operating cost and see whether it clears break-even. The basis is treated eligible original-COD orders—not a month or an asserted national population.
One operating intervention. All assigned eligible orders count, including nonresponders and missing acknowledgments.
TRANSLATE THE HURDLE /29% PLANNING BASELINE
A 3pp gain means 3 extra paid completions per 100 assigned orders: the equivalent of recovering about 10% of otherwise failed orders. Base operating break-even at 1.2pp requires about 4%, before setup. These are required-equivalent shares, not measured recoverability; use actual cohort baselines.
MODELLED CONTRIBUTION / EVALUATED COHORT
After programme costs, assumed incremental payment fees and extra-delivery reserve. Before ₱2.5m one-time setup, tax and other unmodeled costs. Allocated overhead is excluded.
additional paid completed orders
operating break-even / completion gain
incremental prepaid orders
allocated overhead shown separately. It is not counted as cash contribution.
WHY ₱1.50 / ORDER?
₱0.20 in-app events + ₱0.90 expected exception support (10% × 3 minutes × ₱180/hour) + ₱0.40 carrier coordination. These are planning cost targets; meter SMS, extra escalation and carrier charges.
BREAK-EVEN IS A DECISION RULE
Base handover: ₱1.50 ÷ (₱50 reverse + ₱80 commission − ₱5 reserve) = 1.20 pp. With no realizable reverse saving, it becomes 2.00 pp. Use actual net commission and charges in the pilot.
Open equations, cohort definitions and setup recovery
Denominator: all eligible original-COD orders assigned to the treatment arm, whether or not the buyer responds or carrier accepts. A 100k randomized pilot with 50% treatment has only 50k treated orders. The optional pilot mix allocates those treated orders 60% handover, 20% confidence and 20% prepaid offers. This is a proposed resource plan, not observed market composition or measured reach.
Outcomes: handover/confidence recovery = all assigned eligible orders × the absolute gain in paid-completion rate. Cancellations, unresolved orders and suppressed dispatches are not recovered sales. Prepaid recovery = offered orders × incremental conversion × (29% planning baseline − migrated final failure). Nonresponders and missing acknowledgments remain in intention-to-treat analysis. First-attempt failure is not terminal failure; pending outcomes do not count as saved.
Contribution: recoveries × realizable reverse charge + recovered commission − programme cost − incremental merchant fees − extra-delivery reserve. Commission assumes 10% of ₱800 handover/prepaid AOV and ₱450 confidence AOV. Early cancellations earn no modeled commission. Reverse savings apply only before the carrier charge becomes unavoidable. Actual finance evaluation includes net commission after refunds/returns, support, incentives and all carrier costs.
Overhead: confidence/handover recoveries × ₱60 plus migrated orders × 29% × ₱60. It is displayed separately and excluded from contribution. Actual cohort baseline, basket value and costs must replace common planning inputs.
Setup: ₱2.5m is separate. At the base ₱2.25 handover margin per treated order, recovery needs about 1.11m such orders; a six-month hurdle needs at least 185,186 treated orders/month. These are volume thresholds to verify, not a national reach claim. Scale uses conservative measured margin and actual setup.
Downside/base/upside settings are decision stresses, not confidence intervals or forecasts. Download reproducible Python model ↗ · Download authoritative scenarios ↗
FIXED DECISION STRESSES /100k TREATED ORDERS
Same target.
Different economics.
These fixed examples complement the adjustable lab. They are illustrative input combinations, not probabilities or a forecast. All exclude setup and allocated-overhead cash benefits.
| Stress | Gain | Cost/order | Recovered AOV | Reverse saving | Contribution |
|---|
THE INVESTMENT DECISION
FUND THE TEST.
EARN THE SCALE.
~₱2.6m
Estimated learning investment; ₱2.61m ceiling. The underlying rates/effort are planning estimates. Detailed conditional spend: ₱2.6044m = ₱2.5m setup + ₱104,400 variable cost/reserve. The full conditional design has 100k randomized orders, only 50k treated. Before build, approve carrier capability, commercial costs and enrollment. Prepaid requires separate payments approval before development/enrollment. If a cohort is infeasible, recost/reapprove the design and its outcomes.
Base pilot contribution: ₱100k before setup; −₱2.4m after setup. This investment buys evidence rather than repaying setup during the pilot.
- Handover: 30k treatment / 30k control
- Confidence: 10k treatment / 10k control
- Prepaid: 10k treatment / 10k control
- Weeks 1–2Before build: verify selected NCR/Mindanao lanes, actionable carrier acknowledgment, actual charges, workload, enrollment and payments feasibility. Check event definitions and power. Lock disjoint first-order cohorts before randomization.
- Weeks 3–4Implement one existing-3PL adapter. QA driver/dispatcher acknowledgment, supported options, standard fallback and owned recovery.
- Weeks 5–8Run the primary handover trial; start only approved supporting cohorts after instrumentation passes. An infeasible cohort changes sample, cost and impact together. Keep courier-route assignment stable; no daily crossover.
- Weeks 9–12Mature and read the outcomes. Use a common 30-day order window, extend unresolved late outcomes and follow 90-day repeat separately. No automatic rollout date.
ALLOCATION / A RESOURCE PLAN
60k handover.
20k confidence.
20k prepaid.
Assign handover routes first, then separate eligible apparel, then remaining wallet-ready offers. Enroll one first order per buyer; lock assignment for follow-up. Never remove missing preferences or acknowledgments after assignment.
POWER / POOLED HANDOVER TEST
300 stable clusters
per arm.
Verify that 600 stable clusters can enroll. Plan 100 first orders per courier-route cluster, not all route traffic. At 29% baseline, 1.5 pp paid-completion MDE, 5% two-sided alpha, 80% power and ICC 0.01, the equal-size approximation needs about 28k orders/arm; propose 30k.
Re-estimate baseline, ICC, cluster-size variation and delivered-value variance before assignment. Power is pooled across the two settings, not independently regional or a guarantee of an economic go decision.
Open the ₱2.5m setup planning ledger
| Work | Person-weeks | Loaded PHP/week | Budget PHP |
|---|---|---|---|
| Product / engineering | 16 | 75,000 | 1,200,000 |
| 3PL integration / QA | 12 | 50,000 | 600,000 |
| Launch / training | 12 | 25,000 | 300,000 |
| Evaluation | 4 | 50,000 | 200,000 |
| Contingency | — | — | 200,000 |
Fully loaded planning rates, not quotes. Launch/training excludes the per-exception support already charged as variable cost. No partner funding is assumed. Meter actual costs at the feasibility gate; rescope or stop if the economics fail.
Scale gate: the 95% lower confidence bound on realized contribution is positive; delivered value per eligible buyer/order rules out a predeclared material decline; complaints and wrongful restrictions meet predeclared limits. Verify six-month setup recovery using measured throughput and conservative margin. Extend or redesign if inconclusive.
Analytics/finance counts all assigned eligible orders using actual net commission, settlement, known refunds/returns and carrier charges over a common 30-day window. Disclose later liabilities and extend immature outcomes. No retention value is monetized. Full operating and evaluation plan ↗.
05 / OPEN THE CASE DATA
THE NUMBERS.
IN THE OPEN.
Transcribed from the original case images. Values are retained as supplied, including inconsistencies.
Read the reconciliation and assumption notes +
Inspect the ten original case images +
06 / THE EVIDENCE TRAIL
NO CLAIM
WITHOUT CONTEXT.
Public research checked 24 October 2026. Existence of a practice establishes precedent, not a transferable failure-reduction percentage.
REAL-WORLD PRECEDENTS / DRAG TO EXPLORE
Open the complete source register
What we do not know yet
No interviews or firsthand app walkthrough were available. There is no order-level dataset, cross-segment join, verified intervention uplift or supplier quote. The feasibility gate must establish lane enrollment, cohort baselines, actual charges and 3PL acknowledgment/recovery capacity. The trial must establish intention-to-treat final outcomes and realized contribution. The public BSP report is accessible through web extraction; local PDF archival was blocked.