Split-payment statistics for ecommerce order checks
https://checkoutcloser.com/guides/split-payment-statistics-for-ecommerce-order-checks/
## Quick answer
In the public Olist data, 2,961 of 99,440 orders with payment data had multiple payment records, or 2.98%. Of those orders, 2,672 included a voucher. A recovery reply needs the full order-level payment picture before it asks a shopper to pay again. (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
The useful question is how the payment rows belong to the order. A voucher and a card payment can together fund one purchase, while repeated rows of the same type need their own interpretation. The analysis separates these cases and shows what an operator should check before explaining a short card amount. Its figures describe Olist’s historical records, rather than the payment status of a current customer.
## Key Takeaways
A reanalysis of the public Olist dataset found 2,961 orders with multiple payment records among 99,440 orders with payment data, or 2.98% (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
Of the 2,961 Olist orders with multiple payment records, 2,672 included a voucher, or 90.24% (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
The Olist reanalysis found 715 multiple-record orders with only one distinct payment type, representing 24.15% of the multiple-record group (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
Exactly two payment records appeared in 2,382 of Olist’s 2,961 multiple-record orders, or 80.45% (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
## How can one order have several payment records?
A shopper says they used a gift voucher and paid the balance by card. A support view showing one card entry may give the agent only part of the transaction. The useful check starts with the order ID and brings together every payment record attached to it.
Figure description: Split-payment statistics: 2,961 orders with multiple payment records divide into 2,246 mixed-method and 715 same-method orders, with the method combinations directly labeled.
Figure 1: Several payment rows create different investigation paths. Source: Olist public dataset, original allocation analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026. Original analysis of the cited source. Source periods and calculation limits appear in the text.
The public Olist marketplace data contains 99,440 distinct orders with payment records. Most have one row. Another 2,961 have several. Those orders would appear more than once if an analysis counted payment rows as completed purchases. The same mismatch matters in a conversation when the agent has to explain what was paid and what still needs checking. (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
## How often do vouchers appear in the multiple-record group?
Vouchers appear in 2,672 of the 2,961 orders with several payment records, or 90.24%. That is a strong reason to look for a voucher allocation when a card amount seems smaller than the order total. It is still a reason to inspect the record. The figure alone leaves the payment status and any later reversal unresolved. (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
Some combinations contain a voucher and another payment type. Others contain several records of the same type. The latter group accounts for 715 orders, or 24.15% of multiple-record orders. A rule that treats two records as two different payment methods would describe these orders incorrectly. (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
## What should an explanation say about amounts and status?
Consider an order for BRL120 paid with a BRL20 voucher and a BRL100 card payment. A useful explanation states the full total, the voucher applied and the card amount. The remaining balance depends on the verified status of those records. The example shows how to organise the explanation; its amounts are worked inputs.
The calculation should use a stable order key and preserve the sequence number, method, amount and current status of each payment. For a live store, the processor record supplies the status. Olist supplies a historical payment allocation, so the analysis cannot settle whether a present-day shopper owes money. An ambiguous record needs a payment check before a further collection request.
## What should be checked before repeated records are added?
The replay found zero repeated payment sequence numbers within orders. That makes the sequence field a useful consistency check in this export. A different export may repeat a row because it records retries, webhooks or status updates. The correct deduplication key follows the payment system’s record definition.
Most multiple-record orders had exactly two records, 2,382 of 2,961, or 80.45%. The longest allocation had 29 records. An interface that displays only the first two entries could still leave part of an order unexplained. A compact total can sit above a complete expandable allocation. (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
## How does this change a checkout recovery reply?
Use the study to decide what information the agent must retrieve. It needs the whole order allocation before it explains a short card charge, an applied voucher or a possible remaining balance. Where the live processor has confirmed payment, the reply can explain the split clearly. Where status is uncertain, it can name the specific check being made.
Keep the benchmark’s boundary visible when citing it. These are recorded Brazilian marketplace orders from 2016 to 2018. The 2.98% figure describes multiple payment records in that dataset. A store can measure its own rate using the same order-level grouping, then inspect its own allocation and status rules. (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
## Which source figures support the findings?
The table retains the source measure and its population. Independent results in the next table use the stated formula; their precision follows the recorded inputs.
Source measure | Value and unit | Population or denominator
Orders with payment records | 99,440 orders (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | Unique order IDs with at least one payment record
Orders with multiple payment records | 2,961 orders (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | Unique order IDs with more than one payment row
Multiple-record orders including a voucher | 2,672 orders (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | Orders with more than one payment row
Multiple-record orders with one payment type | 715 orders (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | Orders with more than one payment row
Repeated payment sequence numbers | 0 records (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | Duplicate sequence identifiers within the same order
Orders with exactly two payment records | 2,382 orders (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | Unique order IDs with payment data
### How were the additional comparisons calculated?
These are calculations from the source inputs. Each retains its original unit and comparison population.
Calculation | Result | Formula
Orders with multiple payment records, percent | 2.98 % (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | 2961 / 99440 * 100
Voucher share of multiple-record orders | 90.24 % (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | 2672 / 2961 * 100
One-type share of multiple-record orders | 24.15 % (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | 715 / 2961 * 100
Exactly two records among multiple-record orders | 80.45 % (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | 2382 / 2961 * 100
Single-record orders, percent | 97.02 % (Olist data, independent analysis (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026) | 96479 / 99440 * 100
## Which payment pattern should the agent investigate first?
The 2,961 multiple-record orders divide into 2,246 with more than one payment type and 715 with only one type. The mixed-method group contains 2,245 card-plus-voucher orders and one card-plus-debit order. The one-type group contains 427 voucher-only, 287 credit-card-only and one debit-card-only order. These are classifications of the recorded allocation, not evidence of how a payment interface was presented. (Olist data, original allocation decomposition (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
That decomposition gives three different support jobs. A card-plus-voucher order needs an explanation of how the total was allocated. A voucher-only order with several rows needs all voucher entries gathered before a missing amount is inferred. Repeated credit-card rows need their sequence and status checked, because a row count cannot establish whether the records are separate charges, installments, retries or another accounting representation.
### Why is voucher presence not the same as using two methods?
Of the 2,672 multiple-record orders containing vouchers, 427 use vouchers as their only recorded type, 15.98% of that voucher-containing group. Calling all voucher-containing records card-and-voucher splits would overstate that exact combination by 427 orders. Conversely, the one mixed card-and-debit order shows why a voucher flag cannot identify every mixed-method allocation. (Olist data, original allocation decomposition (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
A compact allocation view can therefore show order total, sum of valid payment entries and the distinct method set as separate fields. Expanding the view should reveal every sequence and its amount. Keep the full method set even when the first visible entry is a card. An order with three voucher rows still has one payment type; an order with one voucher and one card has two.
### How much would a payment-row order count overstate purchases?
The original payment file contains 103,886 rows for 99,440 orders. Counting each row as a purchase would add 4,446 apparent purchases, a 4.47% overstatement relative to distinct order IDs. That error is larger than the 2.98% share of orders with multiple records because some orders contribute more than one extra row. The two measures answer different questions. (Olist data, original row-count reconstruction (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce), Nov 2026).
Use a three-record known-answer case when validating an integration: an order for BRL 120 has voucher entries of BRL 10 and BRL 10 plus a BRL 100 card entry. The correct purchase count is one, the payment-record count is three, the distinct-type count is two and the allocation sum is BRL 120. A duplicate sequence should fail the import check instead of silently becoming another payment. These worked inputs test the transformation; they do not describe an additional Olist observation.
Download the allocation decomposition (https://checkoutcloser.com/guides/split-payment-statistics-for-ecommerce-order-checks/data/allocation-decomposition.csv) and run the calculations (https://checkoutcloser.com/guides/split-payment-statistics-for-ecommerce-order-checks/analysis.py). A production reply still needs the provider's current status and reversal history. The exported method pattern selects the investigation, not the amount to collect.
## Methodology and Sources
Count one order once. Separate number of payment records from number of distinct payment types and verify payment_sequential uniqueness. Historical Brazilian marketplace data, not a representative DTC merchant sample. Recorded final states cannot identify authorization, refund, fraud or abandonment causes. No counterfactual or causal recovery effect is inferred. All available dataset dates are retained. Multiple records are not proof of duplicate charges, declined retries or a second purchase. Derived percentages are rounded to two decimal places.
The source period is 2016 to 2018, all available Olist records. Independent calculations were completed in November 2026 and use the formulas shown above.
## Glossary
Order ID: The key used to join records belonging to an order.
Payment record: One row describing a recorded payment amount or method.
Freight: The shipping amount recorded with an item.
Matched order: An order found in both files used for a comparison.
Denominator: The population used to calculate a share or rate.
Recorded final state: The status present in the historical export.
Payable balance: The amount currently owed after relevant payments and adjustments.
Reconciliation: A check that related financial records agree under stated rules.
## Sources
Olist data: Brazilian E-Commerce Public Dataset by Olist (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce)
Check the source figures
The data file retains source labels, values, populations and limitations for this analysis.
Quick answer
In the public Olist data, 2,961 of 99,440 orders with payment data had multiple payment records, or 2.98%. Of those orders, 2,672 included a voucher. A recovery reply needs the full order-level payment picture before it asks a shopper to pay again. (Olist data, independent analysis, Nov 2026).
The useful question is how the payment rows belong to the order. A voucher and a card payment can together fund one purchase, while repeated rows of the same type need their own interpretation. The analysis separates these cases and shows what an operator should check before explaining a short card amount. Its figures describe Olist’s historical records, rather than the payment status of a current customer.
A reanalysis of the public Olist dataset found 2,961 orders with multiple payment records among 99,440 orders with payment data, or 2.98% (Olist data, independent analysis, Nov 2026).
Of the 2,961 Olist orders with multiple payment records, 2,672 included a voucher, or 90.24% (Olist data, independent analysis, Nov 2026).
The Olist reanalysis found 715 multiple-record orders with only one distinct payment type, representing 24.15% of the multiple-record group (Olist data, independent analysis, Nov 2026).
Exactly two payment records appeared in 2,382 of Olist’s 2,961 multiple-record orders, or 80.45% (Olist data, independent analysis, Nov 2026).
A shopper says they used a gift voucher and paid the balance by card. A support view showing one card entry may give the agent only part of the transaction. The useful check starts with the order ID and brings together every payment record attached to it.
Figure 1: Several payment rows create different investigation paths. Source: Olist public dataset, original allocation analysis, Nov 2026. Original analysis of the cited source. Source periods and calculation limits appear in the text.
The public Olist marketplace data contains 99,440 distinct orders with payment records. Most have one row. Another 2,961 have several. Those orders would appear more than once if an analysis counted payment rows as completed purchases. The same mismatch matters in a conversation when the agent has to explain what was paid and what still needs checking. (Olist data, independent analysis, Nov 2026).
Vouchers appear in 2,672 of the 2,961 orders with several payment records, or 90.24%. That is a strong reason to look for a voucher allocation when a card amount seems smaller than the order total. It is still a reason to inspect the record. The figure alone leaves the payment status and any later reversal unresolved. (Olist data, independent analysis, Nov 2026).
Some combinations contain a voucher and another payment type. Others contain several records of the same type. The latter group accounts for 715 orders, or 24.15% of multiple-record orders. A rule that treats two records as two different payment methods would describe these orders incorrectly. (Olist data, independent analysis, Nov 2026).
Consider an order for BRL120 paid with a BRL20 voucher and a BRL100 card payment. A useful explanation states the full total, the voucher applied and the card amount. The remaining balance depends on the verified status of those records. The example shows how to organise the explanation; its amounts are worked inputs.
The calculation should use a stable order key and preserve the sequence number, method, amount and current status of each payment. For a live store, the processor record supplies the status. Olist supplies a historical payment allocation, so the analysis cannot settle whether a present-day shopper owes money. An ambiguous record needs a payment check before a further collection request.
The replay found zero repeated payment sequence numbers within orders. That makes the sequence field a useful consistency check in this export. A different export may repeat a row because it records retries, webhooks or status updates. The correct deduplication key follows the payment system’s record definition.
Most multiple-record orders had exactly two records, 2,382 of 2,961, or 80.45%. The longest allocation had 29 records. An interface that displays only the first two entries could still leave part of an order unexplained. A compact total can sit above a complete expandable allocation. (Olist data, independent analysis, Nov 2026).
Use the study to decide what information the agent must retrieve. It needs the whole order allocation before it explains a short card charge, an applied voucher or a possible remaining balance. Where the live processor has confirmed payment, the reply can explain the split clearly. Where status is uncertain, it can name the specific check being made.
Keep the benchmark’s boundary visible when citing it. These are recorded Brazilian marketplace orders from 2016 to 2018. The 2.98% figure describes multiple payment records in that dataset. A store can measure its own rate using the same order-level grouping, then inspect its own allocation and status rules. (Olist data, independent analysis, Nov 2026).
The table retains the source measure and its population. Independent results in the next table use the stated formula; their precision follows the recorded inputs.
The 2,961 multiple-record orders divide into 2,246 with more than one payment type and 715 with only one type. The mixed-method group contains 2,245 card-plus-voucher orders and one card-plus-debit order. The one-type group contains 427 voucher-only, 287 credit-card-only and one debit-card-only order. These are classifications of the recorded allocation, not evidence of how a payment interface was presented. (Olist data, original allocation decomposition, Nov 2026).
That decomposition gives three different support jobs. A card-plus-voucher order needs an explanation of how the total was allocated. A voucher-only order with several rows needs all voucher entries gathered before a missing amount is inferred. Repeated credit-card rows need their sequence and status checked, because a row count cannot establish whether the records are separate charges, installments, retries or another accounting representation.
Of the 2,672 multiple-record orders containing vouchers, 427 use vouchers as their only recorded type, 15.98% of that voucher-containing group. Calling all voucher-containing records card-and-voucher splits would overstate that exact combination by 427 orders. Conversely, the one mixed card-and-debit order shows why a voucher flag cannot identify every mixed-method allocation. (Olist data, original allocation decomposition, Nov 2026).
A compact allocation view can therefore show order total, sum of valid payment entries and the distinct method set as separate fields. Expanding the view should reveal every sequence and its amount. Keep the full method set even when the first visible entry is a card. An order with three voucher rows still has one payment type; an order with one voucher and one card has two.
The original payment file contains 103,886 rows for 99,440 orders. Counting each row as a purchase would add 4,446 apparent purchases, a 4.47% overstatement relative to distinct order IDs. That error is larger than the 2.98% share of orders with multiple records because some orders contribute more than one extra row. The two measures answer different questions. (Olist data, original row-count reconstruction, Nov 2026).
Use a three-record known-answer case when validating an integration: an order for BRL 120 has voucher entries of BRL 10 and BRL 10 plus a BRL 100 card entry. The correct purchase count is one, the payment-record count is three, the distinct-type count is two and the allocation sum is BRL 120. A duplicate sequence should fail the import check instead of silently becoming another payment. These worked inputs test the transformation; they do not describe an additional Olist observation.
Download the allocation decomposition and run the calculations. A production reply still needs the provider's current status and reversal history. The exported method pattern selects the investigation, not the amount to collect.
Count one order once. Separate number of payment records from number of distinct payment types and verify payment_sequential uniqueness. Historical Brazilian marketplace data, not a representative DTC merchant sample. Recorded final states cannot identify authorization, refund, fraud or abandonment causes. No counterfactual or causal recovery effect is inferred. All available dataset dates are retained. Multiple records are not proof of duplicate charges, declined retries or a second purchase. Derived percentages are rounded to two decimal places.
The source period is 2016 to 2018, all available Olist records. Independent calculations were completed in November 2026 and use the formulas shown above.
Use the citation below when referring to this analysis. The original source and calculation scope accompany each finding.
Checkout Closer. "Split-payment statistics for ecommerce order checks." Checkout Closer, November 2026. https://checkoutcloser.com/guides/split-payment-statistics-for-ecommerce-order-checks/
<a href="https://checkoutcloser.com/guides/split-payment-statistics-for-ecommerce-order-checks/">Split-payment statistics for ecommerce order checks</a> (Checkout Closer, November 2026)