COMPLETE ANSWER · MERCHANDISING

Did the Walmart Rollback work for the supplier? What to close before you call it a win

Why units up and ROAS above target do not settle a Walmart Rollback: the unit-versus-AUR test, the OTIF charge that bills later, and the store position to confirm.

By Matt Drake, practicing retail advisor ·

DIRECT ANSWER

A Walmart Rollback wrap-up, as of September 2026, clears only when unit growth in Scintilla Channel Performance outpaces the AUR decline with penny profit held. Read Ad Center ROAS as click-date credit, not proof; wait for the OTIF charge that bills quarterly in HighRadius; confirm Repl InStock and valid stores held. Close all four first.

Not yet. A Rollback wrap-up built from Ad Center and a sales pull answers the easy question, whether units moved, and leaves the three that decide the outcome open. The first is the test itself: a Walmart Rollback succeeds only when the unit growth percentage is larger than the average unit retail (AUR) decline it was bought with, while penny profit holds. The second is a charge that has not arrived: event volume is where OTIF stress concentrates, and OTIF charges bill on a quarterly invoicing schedule as an accounts-receivable invoice in HighRadius, about five weeks after the quarter ends, so the P&L of a Rollback that ended last month is not closed. The third is the store position the event left behind, read as Repl InStock against the 97% healthy line and the valid-store count against the traited count. And the number the deck leads with, ROAS, is credit assigned to a click date under Walmart Connect’s attribution rules; it is evidence of where sales were credited, not proof the Rollback caused them. Close all four before the deck goes out, and if the wrap-up is for a client, close them inside that client’s own context, because the tier, the freight terms, and the calendar change what each one means.

What is the actual test for a Rollback?

Walmart’s rules for a Rollback are specific, and the wrap-up has to be judged against them, not against “units are up.” Walmart’s public Every Day Low Prices page calls Rollbacks targeted price reductions on the items families buy most; the supplier-facing rules are tighter. A Rollback is a temporary retail price reduction of at least 10% off retail; it must keep overall penny profit constant throughout, and in practice the supplier funds most or all of the reduction through a temporary cost reduction or allowance; and the reduced price holds for at least 90 and no more than 180 days. The success bar follows from the mechanics: a successful Rollback offsets the AUR decline with a unit-growth percentage larger than the AUR percentage decline, while maintaining margin. If AUR fell 15% and units rose 12%, the Rollback destroyed value even though every chart in the deck points up.

The unit read comes out of Channel Performance, and it is available on both Scintilla tiers. In Report Builder, run Store Sales & Inventory (or Omni Sales, a Charter dataset, if the item also ships to home) with a multi-time selection: the Rollback period and an equal-length previous period, thirteen weeks against thirteen weeks for a quarter-long event. Filter to the Rollback items, pull POS Quantity This Year and Last Year, and compute two numbers after export: units change against the prior equal period and units change year over year. Both are the headline unit reads, and they can disagree when the category itself was moving. The Sales Type Code field, where report code 7 is Rollback, isolates the units that rang at the Rollback price; inventory and the other non-sales metrics report only on the regular-sales rows, so a Rollback row showing zero inventory is expected, not missing. Comparison weeks follow the Walmart calendar, not the calendar month; the comp week finder and the week converter do that mapping.

The causality and customer reads are Charter. Guided Insights lists the root causes of an anomaly in your supplier dollar share in order of potential sales impact; for a Rollback window, Pricing Change should sit first, which is the platform’s evidence that the price drove the lift. If Replenishment Health sits first, the event ran into supply, and the story is about in-stock, not price. Shopper Behavior’s Promotions report splits the sales between existing customers and customers new to the category, and Performance in Detail shows whether penetration or frequency carried the growth. On Basic, none of those modules exist to open, so the wrap-up can prove the unit test and nothing about who bought; which tier the client sits on decides the shape of the deck before the first pull. One universal caution: Channel Performance restates daily back two years, so an analysis delivered weeks after the pull should be re-pulled, not re-sent.

Why is the ROAS not the proof?

Because attribution assigns credit, and it assigns it under rules that do not line up with the Rollback’s own clock. Sponsored Search attribution is click-based, with 3, 14, and 30-day windows and 14 as the default (the basis Walmart Connect’s public new-customer goal page also states), and every attributed conversion is credited to the date of the click, not the date of the sale. A daily Ad Center view therefore never reconciles with the daily sales in Supplier One, and a week that ended yesterday is still filling; judge attributed sales on ranges that closed at least a full window ago. The default ROAS counts online sales only; in-store credit sits beside it as Omni ROAS and the in-store metrics.

Two 2026 changes matter for any wrap-up that compares against last year. Since June 1, 2026 Sponsored Search reports under a unified, deduplicated last-touch model, and orders canceled within 72 hours of the order date come out of attributed sales while returns never do. A Nil Pick, an online order the store could not pick, is one of the cancellations that rule removes. So a Rollback that outran store execution shows up twice: as Nil Picks in Scintilla’s store-fulfillment metrics on Charter, and as attributed sales quietly removed from the ROAS. History was not restated, which means a comparison against a pre-June benchmark mixes two methodologies, and any automated rule keyed to ROAS was calibrated on undeducted orders.

Then there is what the ROAS was measuring. During a Rollback the item page carries the Rollback badge and the reduced price, and paying to advertise items that are already heavily discounted is one of the common trade-promotion mistakes, so an event-period ROAS is read next to the price state, not as a media result on its own. Walmart’s advice to advertisers is to treat ROAS as a guide against a per-item break-even, the reciprocal of margin, not as a goal, and the closed-loop measurement that proves cause is a separate product: the public Walmart Connect measurement page lists attribution and incrementality as separate lenses; Display incrementality carries a minimum of $150,000 in eligible spend while search incrementality runs through the account team. Spend also cannot create serving. An item that goes out of stock online or loses the Buy Box stops serving, the eligibility rule on the public Sponsored Search guide, while the other items in the campaign keep spending, so a campaign-level ROAS through an event that hit in-stock trouble is partly a report on which items were available to advertise.

Where is the OTIF charge, and when does it land?

Event volume is where OTIF stress concentrates, because feature and event quantity is bought as incremental, one-time volume on top of replenishment flow, and the same program measures it. The goals since February 1, 2024 are 90% on time for prepaid freight, 98% Collect Ready for collect freight, and 95% in full, applied to each merchandise alignment. The charge is 3% of the cost of goods on non-compliant cases when the metric falls below goal, and cases Walmart is accountable for are excluded. Compliance is measured at the DC on Walmart’s arrival data; a proof of delivery showing an on-time arrival does not override it. For system-generated orders the measure runs against the original MABD and original quantity unless an approved PO Edit Reason Code in NOVA adjusted them, and that code is the whole accountability question: a supplier-accountable cancellation drops its cases into Not In-Full, while a Walmart-accountable one moves them to Canceled POs and out of the denominator. The OTIF scorecard’s PO Details tab shows the PO Edit Flag and Reason Code columns for every line, and the Charges tab excludes Walmart-accountable cases while the Performance Summary includes them, so the two views do not tie by design.

The dollars per PO are small and the timing is the trap. A 1,000-case prepaid PO at $12 of cost per case with 150 cases late runs 85% on time, below the 90% goal, and the charge is 150 cases times $12 times 3%, about $54. At $50 million of annual cost of goods, each sustained point of non-compliance is roughly $15,000 a year, and the scorecard damage before a line review usually costs more than the fines. Charges are invoiced on the fiscal year’s OTIF invoicing schedule: MABD dates are grouped into quarterly ranges, data freezes monthly, and each group is invoiced once on a published date, generally about five weeks after the quarter ends. The FY27 schedule as published invoices on April 29, July 30, and October 29, 2026 and January 28, 2027, and those dates change without notice. The invoice arrives as an AR bill in HighRadius, netted against payment, with copies posting within about 72 hours, and it disputes there, not in APDP, as SPS Commerce’s public guide to disputing OTIF fines also describes. A projected charge that has not been invoiced cannot be disputed in the portal at all, and that conversation runs through the merchant team. So a Rollback that ended in early September has an OTIF exposure that bills in late October at the earliest, and the wrap-up’s P&L line is a projection until then. The scorecard itself sits in Supplier One under Performance, refreshed weekly, and its Charges Breakdown by accountability is the dispute list: charges sitting in Walmart-accountable buckets are the ones to challenge. In Scintilla, OTIF is Charter-only, through the Omni OTIF dataset and the Inventory tab; a Basic client reads it outside Scintilla. How the charge is routed and disputed is its own subject, and SQEP is a separate monthly bill that can land on the same PO; shipping against a canceled PO is itself a SQEP defect.

What did the event leave behind in the stores?

Two numbers, and the second is the one that walks into the next line review. Repl InStock, item-store-weeks in stock over item-store-weeks expected, is the shelf availability read; 97% is the healthy line used in line reviews and decline diagnosis, and 85% is the threshold below which an item is a modular-placement problem worth diagnosing. It is a harsher measure than plain Instock %, which counts a store as in stock with any on-hand quantity and divides by the valid store count, because it compares on-hand to the forecast window. Read it week by week through the event and for at least four weeks after: the under-forecast case (out of stock, then erratic catch-up orders) shows up there first, and the over-forecast case (overstock, then erratic ordering) shows up in weeks of supply and aging DC inventory. The Walmart merchant and replenishment manager set the final forecast lift for a Rollback; the supplier’s leverage is the quality of the data fed to them before the event, and the forecast itself is a projection of unconstrained demand, never a commitment to order.

The second number is the valid-store count against the traited count. Per-store velocity and in-stock are computed on the valid base, so an event that caused a run of stores to stop replenishing the item can leave $/store/week flat while distribution quietly shrinks; the mechanics of that erosion are the same ones behind “Walmart stopped ordering.” A step change in the traited count is a modular decision and belongs to the buyer; a falling valid count under a flat traited count is stores exiting one at a time and belongs to the replenishment manager. Then the aftermath: post-Rollback hangover, where volume drops below the pre-event baseline as customers work through pantry stock, and cannibalization of the items that were not on Rollback. Incremental volume is what sits above a pre-event baseline once the recovery period is counted, not the lift in the event weeks alone.

How is the funding reconciled?

A Rollback is typically supplier-funded through a temporary cost reduction or an allowance, and since May 19, 2025 the Event Proposal workflow in Supplier One has been the standard mechanism for a supplier-funded co-op event. The proposal carries the funding type (a flat dollar amount per unit sold, a percentage of retail, or a percentage of cost), and the fund amount cannot exceed the item’s cost; approval by both merchant and supplier assigns the Co-op ID, and the amount is calculated on the proposal’s basis, sales or receipts, then reconciled after the event to actual volume. Only co-ops entered through that workflow are visible there, and approved co-ops land as deductions against later payments. Where the funding rides an allowance code instead, the Promotional Allowance is code 51, and applying it to the wrong PO scope is one of the common trade-spend errors that turns into a deduction dispute later. The wrap-up’s funding line is therefore also open until the reconciliation posts, and the honest number for the client is the funded amount on the proposal’s own basis, adjusted for whatever the post-event reconciliation changed.

What does a closed wrap-up look like?

Here is one, hypothetical but concrete. A four-count pack retailing at $6.48 rolled back to $5.48 for thirteen weeks, June 13 through September 11, 2026, funded at $1.00 per unit through an Event Proposal. AUR fell 15.4%; units against the previous thirteen weeks rose 34%, and 29% year over year. That passes the test: unit growth larger than the AUR decline with penny profit held. The deck’s 2.4x ROAS restates as attributed online sales over spend on a 14-day click window under the June model, with in-store credit reported separately; the campaign ran entirely after June 1, so no benchmark mixing, but two of the eight items were out of stock online in weeks three through five and did not serve, so the ROAS describes the six that did. Repl InStock fell from 97.8% to 90.6% in those weeks and recovered to 96.4% by week eight; Nil Picks doubled in the same weeks while the pre-substitution rate held (an offered substitution does not count as a Nil Pick, so the two are read together), which is the in-stock miss showing up as canceled online orders. Traited stores held at 3,400 and valid stores dipped from 3,180 to 3,140 before returning to 3,176, so no distribution was lost. On OTIF, three event POs delivered 420 cases late against a $9.10 cost and pulled the period’s on-time rate to 89%, under the 90% goal, so those cases are chargeable: a projected charge of about $115, plus the dip itself on the scorecard; a fourth PO was canceled by the replenishment manager with a Walmart-accountable reason code, so its 260 cases are out of the denominator. The last MABD weeks of the event sit in a group that does not invoice until late October at the earliest, about seven weeks after the deck is due. The closed wrap-up says: the Rollback passed the unit test and held margin; the ROAS is assigned credit under the June model, not incrementality; in-stock broke for three weeks and recovered without distribution loss; the OTIF charge is projected and the on-time dip is on the scorecard ahead of the line review; the funding line reconciles on actual units. That is a deck a merchant can argue with and a client can trust.

Who checks the wrap-up before the client sees it?

The four checks above cross five systems that keep different clocks: Scintilla and its tier, Ad Center’s attribution model, the OTIF scorecard, HighRadius, and Supplier One’s Event Proposal. A wrap-up drafted in one system’s terms quietly misreads the others, and the misread usually surfaces when the invoice arrives or the buyer opens the scorecard. That is the failure Retail Reason is built to catch: paste in the wrap-up plan, with the client’s tier, freight terms, event dates, and the numbers you intend to lead with, and it checks the test, the attribution read, the charge timing, and the store position, with the date each fact was last verified and a confidence class on the answer. For a firm running several clients, each check runs inside the selected client’s workspace. It answers questions and checks plans; it does not open Ad Center, HighRadius, or Scintilla, and it does not retrieve anyone’s data. See pricing, or contact Matt.

What are the limits of this answer?

The Rollback rules and success test are as Walmart teaches them to suppliers as of May 2025; the OTIF goals have held since February 1, 2024; the FY27 invoicing dates change without notice; the attribution rules apply to Sponsored Search reporting from June 1, 2026 forward; Report Builder and Supplier One scorecard labels are as of July 2026. The worked example is hypothetical. If the current portal differs from this article, use the current portal.

Retail Reason is an independent service, not affiliated with, sponsored by, or certified by Walmart Inc., Walmart Connect, or Walmart Data Ventures; it holds no credentials and files nothing. See the full service limitations.

Sources checked

← All complete answersChoose a plan →