Defect Rate / ODR — weekly analysis of top partners
Order fulfilment quality in Stores: what Order Defect Rate is made of, which partners drive it, where it is deteriorating, where it is improving, and how Ukraine compares with Bolt's top-3 Stores countries by volume.
Key findings
Market-level ODR looks moderate, but the trend is negative and the problem sits with a handful of grocery partners.
MWB excl. VARUS — LOKO + Rukavychka
Volume-weighted Order Defect Rate of the two live MWB chains excluding VARUS. This is not a simple average of the two brands — it is the share of defective orders across LOKO and Rukavychka combined.
Weekly ODR: MWB excl. VARUS vs market and brands
% of orders · LOKO + Rukavychka weighted · source: Databricks
What ODR consists of
Order Defect Rate = share of delivered orders with at least one item defect that impacted the customer: quantity, weight or price. Replacement is tracked as a separate metric.
Quantity defect Main driver
Quantity changed with customer impact: the item was removed or reduced (typically out-of-stock).
Weight defect Secondary
Weighted items: actual weight differs from what was ordered, with customer impact.
Price defect Not an issue in UA
Item price increased after the order was placed.
Item defects by type — weekly volume
Number of defective line items · source: Databricks (weekly refresh)
ODR vs Replacement Rate — UA market level
% of delivered orders · source: Databricks (store_*, dish items)
Weekly ODR by key brand
Three behaviour groups: persistently high (VARUS), degrading (KOPIYKA, SANTIM), and improving (TAISTRA). LOKO and HOP HEY are the low-ODR reference at large volume.
Order Defect Rate by brand, % of orders
12 weeks · source: Databricks weekly Order Defect Rate per brand
Top partners — all metrics
One Databricks methodology for both tables: item-level metrics (A) and the full top-15 by volume (B), including partners with zero ODR.
A. Key partners (item-level metrics + weekly ODR)
| Brand | Items | ODR avg | ODR last | Δ 12 weeks | Qty % | Item repl % | Weight % | Order repl % | Status |
|---|
Qty / Item repl / Weight are item-level (share of line items). ODR and Order repl are order-level (share of orders). That is why Order repl is always higher than Item repl.
B. Top-15 UA Stores partners by order volume (Databricks)
| Brand | Orders 12w | ODR % | Qty % | Weight % | Price % | Replacement % |
|---|
Single ODR definition = qty ∪ weight ∪ price at order level. 7 of 15 partners sit at 0% ODR — mostly alcohol and beverage formats with a narrow assortment and high availability.
Problems, progress and actions by partner
Prioritised by impact on market ODR: volume × defect level × trend direction.
| Partner | Problem | What to do |
|---|
Where the problem is: partner → category → SKU
Interactive drill-down for the 8 partners with the largest impact. Select a partner and category to identify specific SKUs and the dominant signal.
Categories contribution = share of all partner orders with an ODR defect from this category
| Category | Orders | ODR contribution | Qty % | Repl % | Weight % | Price % | Signal |
|---|
Problem SKUs up to 30 SKUs with the most affected orders
| SKU / item | Category | Items | Affected orders | Qty % | Repl % | Weight % | Price % | Signal |
|---|
OOS signal is an analytical proxy, not a confirmed inventory status: quantity adjustment with eater impact and/or replacement. SKU rates use active items as denominator.
VARUS store drill-down
Stores with the highest combined quantity-adjustment and replacement rates.
Worst VARUS stores by qty + replacement
| Store | Items | Qty % | Repl % |
|---|
Benchmark: top-3 countries + Ukraine
The top-3 countries by Stores order volume over the same 12 weeks. ODR is computed identically (qty ∪ weight ∪ price at order level) so the comparison is valid. For each country: top-5 partners by volume and their average ODR.
ODR by country: market vs top partners
% of orders · "excl Bolt Market" = without own 1P, which has ODR ≈ 0 and drags the average down
UA Order Replacement Rate by segment ENT / MM / SMB · country-level reconciliation
Share of delivered Stores orders with at least one replacement. TOTAL is the volume-weighted result across all segments, not the Enterprise rate.
| Segment | Orders | Share of UA Stores | Orders with replacement | Order Replacement Rate |
|---|
How to read this
Action plan and targets
Sequence matters: clean up the metric first, then push on operations.
P0 · this week
- VARUS: remove bags from quantity defect (reclassify the SKUs or exclude from the KPI)
- VARUS: deep-dive the worst stores in the table below
- KOPIYKA + SANTIM: weekly ODR stand-up with the AM
- Treat price/weight volume spikes on the chart as possible data-quality artefacts
P1 · 2–4 weeks
- CAFE RYNOK: audit the 100%-replacement categories (Sushi rolls, Combo meals) — likely a mapping issue
- RODYNNA KOVBASKA: scale calibration and unit rules (qty 19.5%, weight 5.9%)
- Daily out-of-stock sync for high-velocity beverages at VARUS and KOPIYKA
- Document the LOKO / TAISTRA processes as a playbook and roll it out across grocery
Targets
- UA Stores ODR: back below 16% (early-June level)
- VARUS ODR: <35% within 8 weeks
- KOPIYKA and SANTIM: <28%
- Market order replacement rate: <15%
- No scaling of new partners while ODR stays above 20% for 4 weeks
Glossary: what each metric means and how to improve it
Expand a metric to see its definition, how it is calculated, and which actions move it.
Order Defect Rate (ODR)
Definition. Share of delivered orders containing at least one line item with a defect that impacted the customer: quantity change, weight change or price increase. Technical adjustments without customer impact are excluded.
Formula. orders with (quantity defect ∪ weight defect ∪ price defect) ÷ delivered orders.
How to improve: real-time stock availability (OOS sync), hiding unavailable SKUs, picker training, cleaning service items out of the catalog (bags, fees), stock buffer on top SKUs.
Item Quantity Adjustment Defect Rate
Definition. Share of active line items where quantity was changed with customer impact — usually the item was removed or reduced because it was not on the shelf.
Why it matters. It is 66.5% of all defects in UA, making it the main lever on ODR.
How to improve: inventory accuracy on high-turnover SKUs, automatic hiding of out-of-stock items, minimum-stock thresholds, dedicated control over beverages and fresh categories.
Item Replacement Rate / Order Replacement Rate
Definition. Item-level: share of line items replaced by a different SKU. Order-level: share of orders with at least one replacement. The order-level figure is always higher, because a single replacement flags the whole order.
Nuance. Replacement is not part of ODR. A partner can have low ODR and high replacement (HOP HEY: ODR 5.1%, item repl 15.8%) — the customer still receives something other than what was ordered.
How to improve: an approved substitution matrix with in-app customer confirmation, and verifying that "replacement" is not a menu artefact (combos, modifiers, weighted items).
Item Weighted Adjustment Defect Rate
Definition. Share of weighted line items where the actual weight deviated from the ordered weight with customer impact — under-weighing or overcharging.
How to improve: in-store scale calibration, unambiguous units in the menu (100 g, 1 kg), a photo of the expected portion, and a system cap on acceptable deviation.
Item Price Adjustment Defect Rate
Definition. Share of line items whose price increased after the order was placed.
Status in UA. Effectively 0% at partner level. In Romania it is 15.3% and the main ODR component (Carrefour 58.6%).
How to improve: POS ↔ Bolt price synchronisation, blocking price increases after checkout, alerts on price mismatches.
Status thresholds used in this report
Critical — average ODR ≥30%, or material degradation at meaningful volume.
High — ODR 20–30%, or unstable dynamics at low volume.
Watch / improving — trend is improving, or a specific pattern (high replacement at moderate ODR).
Good — ODR below 10% at stable volume.
Methodology and limitations
Period: 12 completed weeks.
Filters: delivered orders, dish items, store_* vertical, country from
dim_provider_v2.country_code.
Source: Databricks main.ng_delivery.dim_basket_item_delivery ×
dim_provider_v2. The report refreshes every Monday at 10:00 Kyiv time.
Limitations. The "top-5 partner average ODR" is a simple average, not volume-weighted. Quantity item-rate uses all dish items with eater impact in the numerator and active items in the denominator (Looker-style). Weight/price volumes on the chart use all matching adjustment flags, so data-quality spikes are visible separately from ODR.