Back to Articles|Published on November 18, 2026|30 min read

Houseblend Article

ASC 606 Returns Refund Liability in NetSuite: DTC Guide

Explainer on the ASC 606 returns refund liability in NetSuite for DTC brands: return asset, expected value reserve, a worked 2026 true-up example with journal entries, cohort curves and reconciliation tests.

By Houseblend

ASC 606 Returns Refund Liability in NetSuite: DTC Guide

Summary

  1. 01ASC 606 and IFRS 15 recognize revenue for goods expected to stay with customers, a refund liability for expected refunds, and a separate asset for the right to recover returned goods.
  2. 02Estimate returns using the method that better predicts consideration, apply the significant-reversal constraint, and update the estimate each reporting period using the brand’s own cohort history.
  3. 03A true-up adjusts revenue for changes in the refund liability and cost of goods sold for changes in the return asset. In the hypothetical January example, both increased by different amounts.
  4. 04In the proposed NetSuite design, the operational workflow records actual returns and a period-end journal brings the refund liability and return asset to their required balances.
  5. 05At each close, backtest estimates against closed cohorts, check that expired cohorts have zero balances, and tie the liability and asset rollforwards to the general ledger.
Inside this article
  1. 01Executive Summary
  2. 02Introduction and Background
  3. 03The ASC 606 Returns Model and Its IFRS 15 Counterpart
  4. 04Estimating the Reserve: Method, Constraint and Cohort Curves
  5. 05Worked Example and the Month-End True-Up
  6. 06NetSuite Design: Where the Reserve Lives
  7. 07Reconciliation Tests and Audit Evidence
  8. 08Data Analysis and Evidence
  9. 09Implications and Future Directions
  10. 10Frequently Asked Questions (FAQs)
  11. 11Conclusion

Executive Summary

Under ASC 606, a customer's right to return a product is not a separate performance obligation: it reduces the estimated transaction price [1]. A seller that grants a return right recognizes revenue only for goods it expects to keep, a refund liability for the consideration it expects to hand back, and a separate return asset (with a matching cost-of-sales adjustment) for the right to recover the goods [2]. The standard's own example, reproduced by PwC, sells 1,000 games at $50 with a $10 unit cost and 6% expected returns, giving revenue of $47,000, a refund liability of $3,000 and a return asset of $600 [1]. IFRS 15 uses the same three-part structure for IFRS reporters (Source: standards.aasb.gov.au), so the model below applies to both frameworks, with one wording difference in the constraint threshold.

Returns are variable consideration, so the estimate uses either the expected value or the most likely amount method, and PwC states that the choice "is not a policy choice" but follows whichever better predicts the consideration [3]. The refund liability is updated at the end of each reporting period [2], and changes in the transaction price land in revenue in the period of change [2]. The return asset is presented separately from the refund liability, not netted against it [2], and a change in the asset flows through cost of sales, which is why a true-up moves revenue and cost of goods sold by different amounts.

This report rebuilds the PwC structure as a labeled hypothetical December cohort with a restocking fee and a share of unsellable returns, then shows how a January true-up moves revenue and cost of goods sold by different amounts. The worked figures appear in the example section and are illustrative only. Published return-rate surveys differ by method and by year (for example, Appriss Retail and Deloitte report 13.21% of 2024 sales [4], and the Data Analysis section compares them with the NRF and Happy Returns survey), so none is a substitute for a brand's own cohort history.

On NetSuite, Oracle's documentation describes a return authorization as a non-posting record [5], with a credit memo that reduces sales income [5]. The pages reviewed for this report do not describe a built-in returns reserve, so the reserve is best designed as a controlled period-end journal that adjusts the refund liability and return asset to required balances, reconciled to Shopify refund exports. This is configuration and workflow design, not a platform shortcoming. The report is accounting education, not advice: confirm policies with your auditor or a qualified accountant.

$28,800Refund liability at the December sale in the hypothetical jacket cohort
$9,180Return asset at the December sale in the hypothetical jacket cohort
$1,935Gross profit increase from the January true-up in the hypothetical jacket cohort
26.8%Share of final refunds that arrived in the sales month across the illustrative closed cohorts

Introduction and Background

Most NetSuite returns content covers the operational flow: return authorization (RMA), receipt, credit memo, refund. That flow records returns when they happen. ASC 606 asks for something different: an estimate, at the time of sale, of the returns that have not yet happened. A brand that books returns only when refunds post has a cutoff problem. Consider the hypothetical example developed later in this report: 2,000 units sold in December for $160,000. A cash-basis ledger reports $160,000 of December revenue. A reserve-based ledger reports $131,200, because the brand expects to refund about $28,800 of that amount after year-end. The January refunds then reduce a period that already carries the reserve, instead of being the first place the returns show up.

The standard's logic is that until the return period lapses the seller has made an uncertain number of sales [6]. The Boards' Basis for Conclusions says that recognizing the return asset separately from the refund liability "provides greater transparency" [6]. For controllers at Shopify, Amazon and wholesale-mix brands, the consequences are concrete: period-end revenue, gross margin, inventory and the balance sheet all depend on the reserve.

This report is written for controllers, VP Finance and CFOs at DTC and e-commerce brands that run or are evaluating NetSuite. It does not teach button-clicking. It provides a worked reference with numbers a reader can copy into a workbook, and it covers the following questions:

  • What does ASC 606 require? The model, the paragraph map, and the IFRS 15 counterpart.
  • How is the reserve estimated? Expected value versus most likely amount, the constraint, and cohort return curves.
  • What are the journal entries? At sale, at actual return, at month-end true-up, and at window expiry.
  • How is it built in NetSuite? Account setup, reserve journals, subsidiaries, currency and audit evidence.
  • How is it tested? Reserve-versus-actual backtests, window roll-off and liability rollforwards.

Two caveats apply. The example figures are illustrative, and no industry return rate is asserted beyond the sourced survey figures in the data section. The ASC 606 paragraph text cited here is read from the Financial Accounting Standards Board (FASB) amendments document as issued in 2014, so confirm current wording in the Codification before relying on a paragraph number.

The ASC 606 Returns Model and Its IFRS 15 Counterpart

The model in one table

Table 1 below maps each element of the model to its US GAAP paragraph and its IFRS 15 counterpart. The IFRS text is read from the Australian Accounting Standards Board (AASB) adoption of IFRS 15, which carries the IFRS 15 wording (paragraph numbers should be checked against the IFRS Foundation text); the IFRS Foundation's own page confirms the standard's January 2018 effective date [7].

ElementUS GAAP (ASC 606)IFRS 15
RevenueOnly for goods not expected to be returned [1]Revenue in the amount of consideration to which the entity expects to be entitled (B21) (Source: standards.aasb.gov.au)
Refund liabilityRecognized when consideration is received and a refund is expected (606-10-32-10) [2]Same obligation (paragraph 55) (Source: standards.aasb.gov.au)
Return assetAn asset, with a cost-of-sales adjustment, for the right to recover products (606-10-55-23) [2]. Measured at former carrying amount less expected recovery costs, including potential decreases in value (55-27) [2]Same asset (B21(c)) (Source: standards.aasb.gov.au); initial measurement by reference to former carrying amount (B25) (Source: standards.aasb.gov.au)
PresentationAsset presented separately from the refund liability (55-27) [2]Asset presented separately (B25) (Source: standards.aasb.gov.au)
ExchangesSame type, quality, condition and price are not returns (55-28) [2]Same (B26) (Source: standards.aasb.gov.au)
Defective goodsEvaluated under the warranty guidance (55-29) [2]Same (B27) (Source: standards.aasb.gov.au)
Constraint thresholdInclude revenue only if it is probable that a significant reversal will not occur [1]Highly probable that a significant reversal will not occur (paragraph 56) (Source: standards.aasb.gov.au)
ReassessmentUpdate the estimated transaction price each reporting period (32-14) [2]Update at the end of each reporting period (paragraph 59) (Source: standards.aasb.gov.au)

Table 1 shows that the two frameworks are structurally aligned on returns. The visible wording difference is the constraint threshold: "probable" under US GAAP, which PwC describes as generally a 75% threshold [3], versus "highly probable" under IFRS 15. The difference matters mainly for new lines with thin history. PwC's guidance for IFRS 15 retailers adds that the converged standard requires the refund obligation and the returned-goods asset to be shown gross [8]. IFRS paragraph numbers are read from the AASB text; confirm against the IFRS Foundation text before citing them.

Five points that drive the entries

  • A return right is not a separate performance obligation. The promise to stand ready to accept returns is not accounted for apart from the refund obligation [2] (Source: standards.aasb.gov.au).
  • A refund liability is not a contract liability. PwC explains that a refund liability is an obligation to transfer cash rather than goods or services [1]. The FASB later removed a "contract liability" label from its own illustrative example for this reason [9].
  • Gross presentation. The standard has no netting permission for the asset and liability; PwC adds that whether a refund liability may be offset against another asset from the same contract follows the general offsetting guidance in ASC 210-20 [1]. That guidance requires, among other conditions, that each party owes the other determinable amounts [10], which a refund owed to a customer and inventory the seller expects back do not satisfy (this is an inference from the conditions, not a statement in the guidance). KPMG describes the result as a "grossed up" balance sheet for entities that previously netted return reserves [11].
  • Exchanges are not returns. A customer swapping a size or color for the same product at the same price does not trigger the returns model [1]. For a DTC apparel brand, this means exchange volume should be stripped from the return-rate numerator, and a size exchange processed as a refund plus a new order should be mapped back to an exchange in the data.
  • Defective goods follow warranty guidance. Returns of defective product in exchange for a working one are a warranty matter, not a right of return [2].

Restocking fees and partial refunds

Many DTC policies keep a fee or refund less than the full price. PwC's view is that a product sale with a restocking fee is no different than a partial return right [1], so the refund liability reflects only the portion of the transaction price expected to be refunded [1]. Expected costs to recover the goods reduce the carrying amount of the return asset [1]. Shopify return rules let a merchant set a restocking fee, but the platform's documentation states that return fees are not automatically deducted from refunds [12], so the refund data a controller reconciles to may not reflect the fee unless it is handled in the refund workflow.

Estimating the Reserve: Method, Constraint and Cohort Curves

Expected value or most likely amount

ASC 606 identifies a product sold with a right of return as variable consideration [2]. It offers two methods. The expected value method is a probability-weighted sum and may suit an entity with many similar contracts (Source: standards.aasb.gov.au). The most likely amount method picks the single most likely outcome and fits contracts with two possible outcomes [2]. A DTC brand with thousands of similar orders per month typically has the population that suits expected value, and the method is applied consistently within a contract [2]. Expected value can produce an amount that is not one of the possible outcomes, which is acceptable for a pool of orders but not for a single large wholesale order with a binary return outcome.

Public filers describe the choice in different words. Nike's fiscal 2026 annual report says it estimates the most likely amount it will be entitled to receive [13], while basing its estimates on historical rates among other inputs [13]. Warby Parker and Revolve describe estimates built on historical return patterns [14] and on historical experience and merchandise mix [15]. The labels differ, but the working input in each case is observed history, adjusted for expectations.

The constraint

After estimating, the constraint limits revenue to the amount for which a significant reversal is not probable. The assessment considers both the likelihood and the magnitude of a reversal [2], and PwC notes that it applies to the cumulative revenue recognized, not only the variable portion [3].

For a mature SKU family with a stable curve, the expected-value estimate and the constrained amount will usually coincide. For a new product line, a launch season or a new return policy, the sensible approach (a judgment, not a rule in the standard) is to widen the range of outcomes, test whether the downside return rate would reverse a significant amount of revenue, and record the more conservative figure. A Journal of Retailing meta-analysis of 21 papers found that scope leniency increases returns while time and exchange leniency reduces them [16], so policy changes can move the curve.

Return windows and cohort curves

A cohort return curve shows, for each sales month, how much of the eventual returns arrive in the sales month and in each following month. The window matters because it defines when the curve ends. Shopify lets merchants set a preset window or a custom number of days [12], and its return rules carry window overrides [12]. Amazon's seller help states that most items may be returned within 30 days of the estimated delivery date [17]. In the Appriss Retail and Deloitte survey of 150 North American retail executives and 1,000 consumers [4], limiting return windows to 30 days or less was among the measures executives named [4], while the NRF survey found retailers prioritizing extended windows for the 2025 holiday season [18]. A curve built on last year's window may not describe this year's cohort.

Table 2 below is a cohort return-curve template with illustrative numbers (Hypothetical Example). Each row is a closed sales cohort with refunds by month after the sale; the last row is an open cohort projected with a development factor.

Sales cohortGross salesRefunded in sales monthMonth +1Month +2Month +3Final refund rate
August$300,000$14,400$33,600$6,000$60018.2%
September$340,000$17,000$37,400$6,460$68018.1%
October$380,000$18,620$43,320$7,980$76018.6%
November (open)$400,000$19,600ProjectedProjectedProjected18.3% projected

The pooled share of final refunds that arrive in the sales month across the three closed cohorts is 26.8% ($50,020 of $186,820 in total refunds). Dividing November's $19,600 by 26.8% projects $73,204 in final refunds, or 18.3% of sales, so the required refund liability at the November close is $53,604 (the projection less the $19,600 already refunded). This chain-ladder style calculation rests on thin history: three cohorts, and any change in policy, mix or seasonality breaks the assumption that November behaves like August to October. Segment the template by channel (Shopify, Amazon, wholesale) and product family before using it for a reserve.

What history is enough for a new line

The standard does not set a minimum number of cohorts, and no public benchmark exists for the amount of history that suffices. The constraint factors include limited experience with similar contracts, and the practical response is to borrow a proxy curve (a similar category or price point), apply a wider range, document why, and replace the proxy as actual cohorts close. Forecasting research is mostly about individual orders rather than reserves: a systematic review analyzed 25 publications on e-commerce returns forecasting [19].

Figure 01
Illustrative final refund rates by sales cohort% of gross sales
Source: Hypothetical Example

The reserve moves the returns conversation from what refunds posted to what refunds will post.

Worked Example and the Month-End True-Up

Rebuilding the PwC example

PwC's Game Co example is the anchor. At the sale, Game Co recognizes revenue of $47,000 on 940 expected-to-stay games [1], a refund liability of $3,000 for the 60 expected returns, and an asset of $600 for the right to recover the 60 games at $10 each. Cost of sales is therefore $9,400 (940 games at $10), because the $600 recovery asset is carved out of the $10,000 inventory cost. The return asset is initially measured at the carrying amount of the goods at the time of sale [1], and PwC notes it can be immediately impaired in some circumstances, for example when returned goods are expected to have no value [1].

A DTC variant (Hypothetical Example)

The hypothetical brand, call it "Harbor Jacket Co", sells 2,000 jackets in December at $80 each ($160,000 gross), with a unit cost of $30 ($60,000). All numbers are invented for illustration; sales tax, shipping and payment fees are ignored.

  • Expected returns: 20% by units, or 400 units, with a 30-day window.
  • Restocking fee: 10% on every return, so the refund is $72 per unit and the expected refund liability is 400 x $72 = $28,800. The retained fee (400 x $8 = $3,200) stays in revenue at transfer of control, consistent with the treatment of a restocking fee as a partial return right described above.
  • Damaged returns: 15% of expected returns (60 units) are unsellable with no recovery value, so no return asset is recorded for them.
  • Recovery cost: $3 per resalable unit (inspection and relabeling), which reduces the asset per unit to $27, as described above.
  • Return asset: 340 resalable units x $27 = $9,180.
  • Revenue: $160,000 less $28,800 = $131,200. Cost of goods sold (COGS): $60,000 less $9,180 = $50,820.

Table 3 below walks the cohort through its four events. Row 3 is the month-end true-up: in January the brand cuts its estimate of total returns for the cohort from 400 to 360 units, with 310 returns already processed and 50 expected, of which 40 are resalable.

EventJournal entries (illustrative)P&L effect
1. Sale (December)Dr Cash or Shopify clearing 160,000; Cr Revenue 131,200; Cr Refund liability 28,800
Dr COGS 50,820; Dr Return asset 9,180; Cr Inventory 60,000
Revenue 131,200; COGS 50,820
2. Actual returns (January: 310 units, 265 resalable, 45 damaged)Dr Refund liability 22,320 (310 x 72); Cr Cash 22,320
Dr Inventory 7,950 (265 x 30); Cr Return asset 7,155 (265 x 27); Cr Accounts payable to returns processor 795
None: the reserve already absorbed these returns
3. Month-end true-up (January 31)Required liability 50 x 72 = 3,600 against book 6,480: Dr Refund liability 2,880; Cr Revenue 2,880
Required asset 40 x 27 = 1,080 against book 2,025: Dr COGS 945; Cr Return asset 945
Revenue +2,880; COGS +945
4. Window expiry (46 further returns: 38 resalable, 8 damaged)Dr Refund liability 3,312; Cr Cash 3,312
Dr Inventory 1,140; Cr Return asset 1,026; Cr Accounts payable 114
Release remaining liability: Dr Refund liability 288; Cr Revenue 288
Release remaining asset: Dr COGS 54; Cr Return asset 54
Revenue +288; COGS +54

Table 3 reconciles to zero. The refund liability goes 28,800 less 22,320, 2,880, 3,312 and 288, ending at nil; the return asset goes 9,180 less 7,155, 945, 1,026 and 54, also ending at nil. Final cohort revenue is $134,368 (131,200 + 2,880 + 288), which equals $160,000 less 356 actual returns at $72. Final COGS is $51,819 (50,820 + 945 + 54), which matches an independent build: 1,644 kept units at $30 ($49,320), 53 damaged units at $30 ($1,590) and $909 of recovery cost on 303 resalable units.

Why the true-up hits revenue and COGS differently

The refund liability is updated at the end of each reporting period [2] and the offset goes to revenue, in the period of the change [2]. The return asset is updated for changes in the liability and for circumstances that suggest impairment, and its offset goes to cost of sales. In row 3, lower expected returns release $2,880 to revenue because 40 fewer refunds are expected, but the asset falls by $945 because 35 fewer resalable units are expected back; those 35 units are now kept sales whose cost belongs in COGS. A lower return estimate therefore raises revenue and COGS together, and the gross-profit effect is the difference, $1,935 here.

For unsellable goods, the asset measurement already deducts potential decreases in value [2], and inventory that comes back below cost and is measured using FIFO or average cost is then carried at the lower of cost and net realizable value [20]. A change in a returns estimate is a change in accounting estimate, accounted for in the period of change and later periods if affected [21]. By contrast, ignoring information that was available when the reserve was set is an error, not an estimate change [22]; the distinction matters for how a late-found mistake is corrected.

Reserve sensitivity (illustrative)

Assume cost of goods at 40% of price and 85% of returned units resalable. A 2-point move in the return rate shifts revenue by 2% of gross sales and the return asset by 2% x 40% x 85% of gross sales, so the gross-profit effect is the difference:

  • At $250,000 monthly sales: revenue $5,000, asset $1,700, gross profit $3,300.
  • At $500,000 monthly sales: revenue $10,000, asset $3,400, gross profit $6,600.
  • At $1,000,000 monthly sales: revenue $20,000, asset $6,800, gross profit $13,200.

A higher rate lowers revenue and raises the asset (lowering COGS); a lower rate does the reverse.

Figure 02
From December sale to window expiry in the hypothetical jacket cohort
  1. DecemberSale

    Recognize $131,200 in revenue, a $28,800 refund liability, and a $9,180 return asset at sale.

  2. JanuaryActual returns

    Process 310 returns and reduce the refund liability by $22,320. The reserve absorbs the returns with no P&L effect.

  3. January 31Month-end true-up

    Reduce the refund liability to $3,600 and the return asset to $1,080. Revenue rises by $2,880 and cost of goods sold rises by $945.

  4. ExpiryWindow expiry

    Process 46 further returns and release the remaining refund liability and return asset, bringing both balances to zero.

NetSuite Design: Where the Reserve Lives

This section is configuration and workflow design. It describes what Oracle's help documentation states about returns postings and proposes a reserve design that sits beside them. It does not describe any NetSuite behavior as a defect; the operational returns flow is built to record returns as they occur, and the reserve is an estimate layered on top.

What the documented returns flow posts

Oracle's help pages describe the following sequence. A return authorization has no accounting impact until items are received back into inventory [5]. The return receipt posts an increase to an asset account [5]. A credit memo decreases the total amount of income from sales [5]. With Advanced Receiving enabled, a return must be credited before it can be refunded [23], and a credit memo that has been refunded cannot be deleted [24]. For items with Multi-Location Inventory, a return linked to a sale uses the exact cost from the original sale [25].

The consequence for design is straightforward. Each of these documented postings records a return when it happens, against revenue and COGS. None of the pages reviewed describes an automatically calculated returns reserve, refund liability or return asset. Advanced Revenue Management (ARM) is documented as creating non-posting revenue arrangements for sales and returns [26], with instructions to record returns using return authorizations and merge the positive and negative arrangements [26]; a return authorization created from a sales order generates an arrangement with a negative revenue element [27], while a credit memo created from a return authorization generates no additional arrangement [27]. The ARM overview page is written around fair value allocation across elements [28]. These pages do not address expected-return estimation, and this report makes no claim about ARM behavior beyond them; confirm current capabilities with Oracle or your implementation partner.

A balance-based reserve journal

Because the documented postings run through revenue and COGS as returns are processed, a clean design is to let them run unchanged and add a period-end reserve journal that moves the refund liability and return asset to their required balances. The journal posts the difference between the required balance and the book balance, so it self-corrects. Using the Table 3 example at January 31:

  • Refund liability: book balance 28,800 (nothing relieved yet), required 3,600. The journal is Dr Refund liability 25,200; Cr Revenue (returns reserve) 25,200. The January credit memos and refunds already debited revenue by 22,320, so net January revenue is +2,880, matching Table 3.
  • Return asset: book balance 9,180, required 1,080. The journal is Dr COGS 8,100; Cr Return asset 8,100. The item receipts already credited COGS 7,950 and the recovery cost debited 795, so net COGS is +945, again matching Table 3.

The two approaches differ only in which transactions carry the P&L effect; the balance-based version needs no change to RMA, credit memo or refund workflows. Mechanically, NetSuite's journal entry record supports this: a reversing journal entry is an exact opposite of the original [29], and it cannot be edited directly [29], which is useful evidence that a reversal was not altered. Reversing and rebooking the full required balance each month, or posting only the difference, is a design choice: reversal is simple to audit, while a difference-only journal keeps a continuous rollforward.

Accounts, subsidiaries and currency

  • Accounts: a refund liability account of a liability type, a return asset account of an asset type, and a contra-revenue or revenue account for reserve movements. Oracle's account-type page notes that asset and liability categories appear on the balance sheet [30]. Keep the return asset out of the main inventory account so the inventory subledger reconciliation stays clean.
  • Subsidiary scope: in a OneWorld account, a regular journal entry cannot be associated with more than one subsidiary [29], so the reserve is booked per subsidiary; entries between subsidiaries use the intercompany journal type.
  • Currency: the subsidiary must be selected before the currency can be accepted or chosen [29]. Oracle's revaluation page notes income statement accounts are not revalued [31]; decide deliberately whether the refund liability account is flagged for revaluation, and have your accountant confirm the treatment for the entity's functional currency. For ARM users, a return element and its sale element must have the same foreign currency exchange rate for the documented merge [27].

Shopify and Amazon inputs

Reconciling the reserve to channel data requires knowing what each export contains. Shopify defines a refund as one issued to a customer for an order [32]; in the transaction CSV, a refund row's amount is the refunded amount [32]; the orders export carries a refunded amount column [33]; the Finances report shows sales reversals as negative numbers on the date the return was processed [34]; and the payout reconciliation report groups refunds and related fee adjustments [35]. Two Shopify specifics matter for cohort data: a refund issued without a return does not record a return reason [36], and a refund cannot be reversed once initiated from the admin [36]. For Amazon, the Selling Partner API returns reports cover only seller-fulfilled returns [37], and settlement reports cannot be requested or scheduled [38]. Define the cohort key (order date or delivery date) before merging feeds.

Figure 03
NetSuite returns postings and the proposed reserve journal
Operational returns
  • A return authorization has no accounting impact until inventory is received.
  • The return receipt increases an asset account.
  • A credit memo reduces income from sales.
Period-end reserve journal
  • Adjust the refund liability and return asset to their required balances at period end.
  • Post the difference between the required and book balances.
  • Offset refund liability changes to revenue and return asset changes to cost of sales.

Reconciliation Tests and Audit Evidence

The reserve is only as credible as its tests. Three recurring tests, run monthly, cover most of the risk, and Oracle's Transaction Detail report (which lists all transactions in chronological order for a period [39]) and System Notes (which record a note when a field changes [40]) provide supporting evidence.

Test 1: reserve versus actual returns by cohort

For every closed cohort, compare the final refund rate to the rate that was reserved at the first, second and third month-end. Track the signed difference in dollars and points. In Table 2, for example, closed cohorts finished between 18.1% and 18.6%; a reserve set at 20% for any of them would have over-reserved by 1.4 to 1.9 points, a pattern that, if persistent, indicates a biased estimate rather than random noise. Investigate any cohort whose error exceeds a threshold finance sets in advance.

Test 2: return-window roll-off

When the window for a cohort closes, the remaining refund liability and return asset for that cohort should be zero after the expiry entries in Table 3, row 4. List any cohort past its window with a non-zero balance. Edge cases to check:

  • Late deliveries: windows measured from delivery, not order date.
  • Overrides: window overrides by collection or product [12].
  • Out-of-window requests: Amazon's seller help states a seller can accept or deny them [17]; flag these rather than reserving for them.

Test 3: liability and asset rollforward

Prepare a rollforward each month for both balances:

  • Refund liability: opening balance, plus new reserve at sale, less refunds processed, plus or minus true-up, less expiry release, equals closing balance.
  • Return asset: opening balance, plus new asset at sale, less goods restocked, plus or minus true-up, less expiry release, equals closing balance.

The closing balances should tie to the general ledger accounts, and refunds processed should tie to the Shopify transaction export refund rows [32] and to the credit memos and refunds posted in NetSuite. Breaks usually trace to exchanges mis-keyed as refunds, restocking fees not deducted [12], or refunds issued outside the returns workflow.

Disclosure and review

The standard requires disclosure of the methods, inputs and assumptions used to measure returns and refunds [2] and to assess whether variable consideration is constrained [2]. Preserve, for each close, the cohort file, the approved estimate, the posted journal and the rollforward; an auditor will ask for each.

Figure 04
Monthly reserve reconciliation tests
  1. 01Backtest closed cohorts

    Compare each closed cohort’s final refund rate with the rates reserved at its first, second and third month-end. Track the signed variance in dollars and points.

  2. 02Check return-window roll-off

    Check that the refund liability and return asset reach zero after expiry. List expired cohorts with a balance.

  3. 03Reconcile both rollforwards

    Tie the closing refund liability and return asset to the general ledger. Match refunds to the Shopify export and the credit memos and refunds posted in NetSuite.

The journal posts the difference between the required balance and the book balance, so it self-corrects.

Data Analysis and Evidence

Published return-rate data and why it cannot set your reserve

Two independent surveys give very different headline rates, and the difference is methodological. The NRF and Happy Returns 2025 report projects $849.9 billion in total retail returns for 2025 [41], with retailers estimating that 15.8% of total annual sales [18] and 19.3% of online sales [41] will be returned. The retailer sample was 358 ecommerce professionals at US merchants with revenue of $500 million or more, alongside 2,006 consumers [18], so it describes large merchants and retailer estimates, not audited data and not small DTC brands. Appriss Retail and Deloitte put 2024 returns at 13.21% of sales [4], with an in-store return rate of 8.72% and an online rate of 24.52% [4]. Appriss Retail's 2026 benchmark reported $706 billion in 2025 merchandise returns [42]. These figures use different bases and years, so they should not be blended. The Census Bureau estimated e-commerce at 17.1% of total US retail sales in the second quarter of 2026 [43], a figure it has since flagged as superseded by revised estimates [43].

What public filers disclose

Filings show how large reserves are relative to the business and how the policies read. As of the dates stated in each filing: Warby Parker's allowance for returns was $3.6 million and $2.6 million at December 31, 2025 and 2024 [14], adjusted in the period the returns occur [14]; Revolve's returns reserve was $77.0 million at December 31, 2025 [15], with an offsetting asset in prepaid expenses and other current assets [15]; and FIGS reported a returns reserve of $4.2 million and $3.9 million at the same dates [44], recording the expected inventory right of recovery as a reduction of cost of goods sold [44]. Nike's sales returns reserve was $1,099 million and $1,277 million at May 31, 2026 and 2025 [13], with the related estimated inventory cost presented as a separate asset [13]. FIGS and Nike both state that revisions to the estimate flow into revenue in the period determined [44] [13], and Revolve notes that actual returns may vary from its period-end estimates [15].

These amounts are not comparable with each other (different fiscal years, mix and channel) and are not a benchmark for any other company. They show the structural features in this report in practice: a reserve established at sale from history, a separately presented recovery asset, and re-estimates through revenue.

Gaps in the evidence

No public benchmark exists for the amount of history sufficient for a new SKU family, for DTC reserve accuracy by category, or for how window length changes aggregate return rates. Claims here are limited to what the cited sources state, and all example figures are illustrative.

Implications and Future Directions

What changes for the close process

The reserve moves the returns conversation from what refunds posted to what refunds will post. Three practical changes follow:

  • A cohort file with an owner: the month-end close needs a named preparer, reviewer and review date.
  • Estimate movements in revenue and COGS: margin analysis must separate them from underlying trading.
  • A policy calendar shared with finance: new windows, free-return promotions and holiday extensions can shift the curve, and the NRF survey found 72% of merchants had begun charging for at least one return option [18].

When to bring in an implementation partner

A single-channel brand in one currency with one subsidiary can often run the reserve as a monthly journal and a spreadsheet. Complexity tips the balance when:

  • Return windows differ by channel or product: Shopify overrides, Amazon's category windows and wholesale terms need separate curves and a consistent cohort key.
  • Marketplace refunds arrive on settlement timing: Amazon documents that settlement reports cannot be requested or scheduled [38], so refund timing is partly outside the brand's control.
  • Multiple currencies or subsidiaries are in scope: regular journals are single-subsidiary [29], and currency, revaluation and intercompany settings need deliberate design.
  • Volume makes manual journals fragile: scripted or integration-driven journals, saved searches and an approval workflow are worth designing once the monthly file runs to thousands of refunds.

A NetSuite implementation partner is one of several kinds of provider that can scope such work; a brand's auditors and its own accounting team remain the owners of the policy.

Looking ahead

Return policies keep changing, which makes a reserve re-estimated on actual cohorts more robust than one fixed in an annual policy memo. Census Bureau data put e-commerce at 17.1% of US retail sales in the second quarter of 2026, an estimate the Bureau has since flagged as no longer the most up-to-date [43], so treat it as indicative of the channel's scale rather than a current figure. IFRS reporters should confirm with their auditors how the "probable" versus "highly probable" constraint wording in Table 1 applies to new product lines.

Frequently Asked Questions (FAQs)

What is the difference between a refund liability and a return asset?

The refund liability is the consideration the seller expects to give back; the return asset is the right to recover the products, measured at former carrying amount less expected recovery costs [2]. They are presented separately, not netted.

How do returns work as variable consideration?

The seller estimates expected returns with the expected value or most likely amount method [2], applies the constraint, and updates the estimate every reporting period [2].

How does the true-up affect revenue and COGS?

Changes in the refund liability go to revenue and changes in the return asset go to cost of sales, so the two lines move by different amounts, as Table 3 shows.

Does NetSuite revenue recognition for returns calculate the reserve?

The documentation reviewed describes return authorizations, credit memos and ARM revenue arrangements for recording returns [26] but does not describe an automatic reserve; a period-end journal design is the practical route.

Conclusion

Recording returns only when refunds post is simple, but it leaves period-end revenue overstated by the refunds that are already likely. ASC 606 and IFRS 15 require a different structure: revenue for goods expected to stay, a refund liability for goods expected back, and a separate return asset. The estimate belongs in finance's control environment, with a cohort file, a documented method and an owner.

The worked example shows the mechanics a controller can copy: a reserve at sale, a balance-based month-end journal, actual returns that relieve the liability, and a true-up that moves revenue and cost of goods sold by different amounts. On NetSuite, the documented returns flow records returns as they occur, and a reserve journal built beside it brings the balance sheet and P&L into line with the standard. The reconciliation tests turn the estimate into something an auditor can follow.

The numbers here are illustrative, and the published return-rate surveys differ by method and population, so a brand's own history is the right basis. Confirm policy decisions and disclosures with the brand's auditor or a qualified accountant before adopting them.

External Sources (44)

About

Houseblend

Make NetSuite work better for your finance and operations teams with Houseblend. We help design, implement, integrate and improve ERP systems, with practical support for the people who use them every day.

Houseblend is a NetSuite consulting firm serving finance and operations teams. We help organizations implement ERP systems, connect business applications, improve existing configurations and maintain the systems that support everyday work. Our audience includes finance leaders, controllers, operations managers, NetSuite administrators and implementation teams.

Implementation and architecture

Houseblend provides NetSuite implementation, architecture and data migration services. We help teams evaluate how business processes, reporting requirements and existing data should fit together in an ERP environment. Training supports the people responsible for adopting and operating the resulting system.

Integrations, customization and AI

Our services include NetSuite integrations and customization, as well as AI integrations and AI transformation work. These engagements connect ERP data and workflows with the broader application landscape. The right design depends on the organization's systems, controls and operating needs.

Improve and support an existing system

Houseblend offers NetSuite health checks, optimization, managed support and project rescue services. We also provide expertise for analytics and specialist workflows, including NetSuite Analytics Warehouse, warehouse management and field service management. Published educational material helps teams investigate options and prepare informed questions for their implementation or support work.

Work with Houseblend

Explore NetSuite implementation, integrations, managed support and AI integrations. Contact Houseblend to discuss your current system and priorities.

Article examples explain concepts rather than promising a particular license, product capability, delivery schedule or outcome. Engagement scope is confirmed with the Houseblend team.

Disclaimer

This document is provided for informational purposes only. No representations or warranties are made regarding the accuracy, completeness, or reliability of its contents. Any use of this information is at your own risk. Houseblend shall not be liable for any damages arising from the use of this document. This content was generated with assistance from artificial intelligence tools, which may contain errors or inaccuracies. Readers should verify critical information independently. All product names, trademarks, and registered trademarks mentioned are property of their respective owners and are used for identification purposes only. Use of these names does not imply endorsement. This document does not constitute professional or legal advice. For specific guidance related to your needs, please consult qualified professionals.