Ecommerce automation problems and fixes: the 12 sellers raise most
Most automation problems are not mysterious.
They come from the same few causes: bad data, loose limits and nobody checking. Find yours below, in the words sellers use, with the fix and the number that tells you it worked.
Quick answer
The most common ecommerce automation problems are oversells from slow or wrong stock sync, "where is my order" tickets nobody prevents, reorders placed too late, listings that drift from the truth, chatbots promising things your policy does not say and automations that break marketplace rules. Each one has a known cause, a fix and a measure.
This page is part of our free AI automation guide, which follows the path from first automations to governance.
The 12 ecommerce automation problems at a glance
| # | Problem, in the buyer's words | Usual cause | Relay Module or service |
|---|---|---|---|
| 1 | "We oversell when the same stock is listed on more than one channel." | Sync from the wrong number; slow sync; bad mapping | RELAY-03 Stock Flow |
| 2 | "Our inbox is full of 'where is my order?'" | Customers not told before they ask | RELAY-06 WISMO Breaker |
| 3 | "We keep running out of our best sellers." | Reorders happen when someone notices | RELAY-03 Stock Flow |
| 4 | "Our listings changed and we did not change them." | No daily monitoring; several data sources | RELAY-09 Catalog Ready |
| 5 | "Our chatbot told a customer something we do not offer." | Bot answers beyond written policy | RELAY-06 WISMO Breaker rules |
| 6 | "An automation got us a policy warning." | Rules not written against platform policy | RELAY-05 Account Health Watch |
| 7 | "Our repricer sold below cost." | No floor, or a floor on stale costs | RELAY-02 Margin Guard |
| 8 | "Our AI-written listings got flagged or suppressed." | Unchecked claims; wrong format | RELAY-09 Catalog Ready |
| 9 | "Something stopped working and nobody noticed for weeks." | No failure alerts, no owner | Module Care; Ops Desk |
| 10 | "We approve so many AI suggestions that we stopped reading them." | Approvals on low-risk actions | Governance (module 07) |
| 11 | "An app has more access to our account than we realized." | Broad permissions, shared logins | 14-Day Ops Foundation; security and access |
| 12 | "AI shopping assistants never mention our products." | Incomplete, inconsistent product data | RELAY-09 Catalog Ready |
1. Why do we oversell when the same stock is on several channels?
What it looks like: canceled orders because a unit was already sold elsewhere; bundles selling units the warehouse does not have; FBA or WFS stock not mapped to Shopify locations.
Cause: one physical unit appears on several channels and the count is synced from the wrong number (expected or received instead of put away), too slowly or with wrong SKU and bundle mapping.
Fix: sync only the warehouse put-away count; map every SKU, bundle and location; set sync frequency and a small buffer per channel on fast sellers; alert a person when two systems disagree. Walmart's cancellation standard is 2% or below (Walmart Marketplace Learn, updated Jul 28, 2026), so this is an account-health issue as well as a customer one.
Measure: cancellation rate by channel, weekly. Learn more: module 06; the Scale guide. Related: overselling across channels.
Here is how a small timing gap becomes a cancellation (illustrative · sample figures). A SKU sells 40 units a day across three channels, and stock syncs every 30 minutes. Two units are left, and all three channels show 2 until the next sync. In a busy half hour the channels sell 3 between them, so one order is canceled. Syncing from put-away, more often for fast sellers, and holding back a buffer of 2 units per channel removes most of these cases.
2. How do we stop "where is my order?" tickets?
What it looks like: agents copying tracking numbers into replies; response times slipping in busy weeks; customers asking twice.
Cause: customers ask because nobody told them.
Fix: send shipping and delay updates before customers ask; give the helpdesk live order and tracking data; let AI draft or send order-status replies from that data; send exceptions (lost parcels, refunds, upset customers) to a person.
Measure: order-status tickets per 100 orders and first response time. Learn more: Start guide, Automation 1. Related: where is my order tickets.
3. Why do reorders always happen too late?
What it looks like: top sellers out while slow SKUs sit in storage; supplier lead times that live in someone's head.
Cause: sales velocity, lead time and inbound status live in different places, so the reorder happens when someone notices.
Fix: set a reorder point per SKU (average daily units x lead time in days + safety stock), track inbound shipments to arrival and alert before a top seller drops below its cover target. Let the system draft the PO; a person approves it.
Measure: stockout days on top SKUs, per quarter. Learn more: Start guide, Automation 2; module 03. Related: stockouts on best sellers.
4. Why do our listings drift from the truth?
What it looks like: titles, images or attributes changed without your approval; different specifications on different channels; low listing quality scores on Walmart.
Cause: product data comes from several places (spreadsheets, channel edits, other sellers, AI rewrites) and nobody checks daily. On Amazon, titles still over the 75-character limit after Jul 27, 2026 "will be automatically rewritten by Amazon's AI systems" in every category except media (PPC Land, Jun 13, 2026, citing an Amazon Seller Forums announcement; trade source; confirm in Seller Central).
Fix: one catalog source for product data; push from it to every channel; monitor top listings daily for changes and suppressions; rewrite titles to each channel's rules yourself, before the platform does.
Measure: listings changed or suppressed, and days to fix. Learn more: module 08.
5. What do we do when our chatbot says something wrong?
What it looks like: a bot promises a refund, a delivery date or a discount your policy does not include.
Cause: the AI answers from general knowledge or guesses instead of your written policy, and nothing stops it at the edge of that policy.
Fix: load only your written policies and live order data; block answers on refunds, exceptions and anything unclear and route them to a person; review a sample of conversations weekly; correct the source, not just the reply. Treat its promises as yours: in Feb 2024 a Canadian tribunal held Air Canada responsible for misleading bereavement-fare refund information its chatbot gave a customer (Moffatt v. Air Canada, 2024 BCCRT 149, Feb 14, 2024).
Measure: policy errors found in the weekly sample and escalation rate. Learn more: module 05.
6. Can an automation cause a marketplace policy violation?
Yes. Automation that breaks policy carries the same risk as a person breaking it, and it repeats the mistake faster.
Common causes: listing claims you cannot document; pricing rules that trip a marketplace's price checks (Walmart automatically unpublishes offers that break its Pricing Rule); review requests outside platform programs; messages that break marketplace messaging rules; tools that log in with shared credentials or pretend to be a browser.
Fix: write each rule against the platform's policy; use official connections where the tool identifies itself; least-privilege, named access; a person approves every marketplace submission; daily account health checks with same-day action.
Measure: policy warnings per quarter and days from warning to action. Learn more: module 07; the Beyond $1M guide.
Where automation stops
Let AI do the busywork. Three actions always wait for a person:
- 01Price changes outside the limits you set.
A pricing mistake can sell your stock at a loss in minutes.
- 02Refunds above a set amount.
Small ones can be automatic. Big ones get a second look.
- 03Anything touching account health.
Appeals and policy replies are written and checked by people.
We add three more to every build: account appeals and reinstatement, policy decisions (what your store promises customers) and any situation nobody has written a rule for. Everything else: automate it, log it and review it every week.
Five guardrails sit under every workflow: limits on every action, approval for anything risky, only the access each tool needs, a log of every step and one person who owns the outcome. What your AI says, you said.
7. Why did our repricer sell below cost?
Cause: no price floor, a floor set on old costs or competing rules that chase each other down.
Fix: a floor and a ceiling per SKU built from current landed cost, fees and shipping; automatic repricing only inside that band; a person approves anything outside it; costs refreshed when fees or tariffs change.
Measure: orders sold below floor (target zero) and SKUs below margin floor. Learn more: module 07.
8. Why did our AI-written listings get flagged?
Cause: the AI added claims you cannot support ("waterproof", "non-toxic"), used restricted terms or ignored a channel's format.
Fix: draft from your own product data; check every claim against documents; fit each channel's title and attribute rules; keep a person approving every listing change. Amazon says sellers remain responsible for the accuracy of AI-generated listings (sell.amazon.com, Nov 3, 2025).
Measure: listings flagged or suppressed per month. Learn more: module 05.
9. Why did an automation fail without anyone noticing?
Cause: a platform update, an expired connection or a changed field broke the workflow, and there was no failure alert and no owner.
Fix: a register of every automation with a named owner; alerts on any failed run; a weekly log review; a re-test after every platform change you see on the Seller rule tracker.
Measure: failed runs per month and days to detect. Learn more: the Beyond $1M guide.
10. What if we approve AI suggestions without reading them?
Cause: too many approvals on low-risk actions, so people click through the risky ones too.
Fix: remove approvals from reversible, low-risk actions (tags, drafts nobody receives) and review those by weekly sample; keep approvals for money, prices, appeals and promises; rotate reviewers; track how often people change the AI's draft.
Measure: override rate and errors found in the weekly sample. Learn more: module 07.
11. What if an app has too much access to our account?
Cause: apps and contractors added over time with broad permissions or shared logins.
Fix: list every user and app with access; remove what is not used; give named users the least access the job needs; never share the owner password (Amazon's service provider authorization for outside providers and User Permissions for employees, a limited Walmart role with Admin kept by you, Shopify collaborator access); no tool gets bank, payout or legal-entity permissions.
Measure: users and apps with access, reviewed quarterly. Learn more: our security and access page.
12. Why don't AI shopping assistants mention our products?
Cause: assistants answer from structured data. Missing attributes, inconsistent specifications, out-of-stock items and thin reviews make a product hard to recommend. Adobe named individual product pages as "one area of challenge," scoring 66% on its AI Content Visibility Checker (vendor estimate, Adobe, Apr 16, 2026).
Fix: fill every attribute on every channel; write content that answers real questions; keep stock and price live from one source; earn real reviews; use Shopify's Agentic Storefronts page and your Merchant Center feed to find gaps.
Measure: attribute completion by channel and traffic from AI sources, where your analytics can see it. Learn more: module 08; What's new in 2026.
Which problem should you fix first?
The one that puts an account or a customer at risk, then the one that costs the most hours. In practice that order is usually:
- 01Anything touching account health:
oversells that become cancellations, policy warnings and access you cannot account for (problems 1, 6 and 11).
- 02Anything that promises customers something untrue:
chatbot errors and listings with claims you cannot prove (problems 5 and 8).
- 03Anything that loses money quietly:
repricing below cost and late reorders (problems 7 and 3).
- 04Anything that wastes hours:
order-status tickets, unread approvals and silent failures (problems 2, 10 and 9).
Write down the measure for each problem before you start, using the line under each one above, and check it again four weeks after the fix with the same method. If the number has not moved, the cause was different from the one you fixed.
A note on AI claims: the FTC has warned businesses not to exaggerate what their AI can do (FTC business guidance, "Keep your AI claims in check," Feb 27, 2023). We describe each tool from its own documentation, label vendor figures as vendor estimates and never say that AI will produce sales for you. AI here means software that drafts, sorts, checks and alerts. People still own the decisions.
Talk it through with a real person
Book a free 30-minute session with an ops lead. Bring one workflow you want off your plate, and we will tell you plainly whether to automate it, how and where it should stop. A real person replies within one business hour, 9am to 6pm US Central, Monday to Friday. No pitch deck.
A written report within 48 hours of working read-only access. You keep it whether or not you hire us.
Free
Ask a real expert. Free.
A real person replies within one business hour, 9am to 6pm US Central, Monday to Friday.
Request received.
That did not send.
You can reach an ops lead directly, and nothing you typed is lost.
What is the most common ecommerce automation problem?
Overselling from stock that is synced from the wrong number or too slowly. It turns into cancellations, and cancellations are a scored account-health metric on marketplaces. The fix is to sync from the warehouse put-away count, with a buffer on fast sellers and an alert when two systems disagree.
Can AI get my seller account suspended?
It can contribute, the same way a person breaking policy would. Write rules against each platform's policy, use official connections, give tools the least access they need and have a person approve every marketplace submission. The usual triggers are listing claims you cannot document, repricing that breaks a price rule, review requests outside platform programs and shared logins; problems 6, 7, 8 and 11 above cover each.
How do I reduce "where is my order" tickets?
Tell customers before they ask: shipping and delay updates, a tracking page and AI-drafted replies from live tracking data. Route lost parcels and refunds to a person. Measure order-status tickets per 100 orders before and after, with the same method.
What should I do when a chatbot gives a wrong answer?
Correct the customer quickly and honestly, then fix the source: restrict the bot to written policy, block the topic it got wrong and add it to your weekly review sample. If the answer promised something, decide quickly whether to honor it; customers will hold you to it.
How do I know an automation has stopped working?
Set an alert on failed runs, give every automation a named owner and review the logs weekly. After any platform change, re-test the workflows it touches. A sudden drop in an automation's run count is often the first sign. Write the expected weekly run count in your register, so a drop is easy to spot.
Should we turn an automation off while we fix it?
If it touches customers, prices, stock counts or a marketplace, yes: pause it, handle the work by hand for a few days and fix the rule before switching it back on. If it only tags orders or sends internal alerts, you can usually leave it running and correct the rule. Every automation should have a written off switch and a named person allowed to use it.
Is it safe to let software reprice on marketplaces?
Only inside a floor and ceiling built from current landed cost, fees and shipping, with a person approving anything outside that band. Refresh the floor whenever a fee or tariff changes, because a floor built on last quarter's costs is how repricers sell below cost. On Walmart, also keep prices inside its Pricing Rule, since offers that break it are unpublished automatically (Walmart Marketplace Learn, updated Dec 12, 2025).
Sources9 sources
- 01help.shopify.comchecked Oct 20, 2026Primary
- 02help.shopify.comchecked Oct 20, 2026Primary
- 03business.adobe.comchecked Oct 20, 2026Primary
- 04marketplacelearn.walmart.comchecked Oct 20, 2026Primary
- 05sell.amazon.comchecked Oct 20, 2026Primary
- 06marketplacelearn.walmart.comchecked Oct 20, 2026Primary
- 07ppc.landchecked Oct 20, 2026Primary
- 08decisions.civilresolutionbc.cachecked Oct 20, 2026Primary
- 09ftc.govchecked Oct 20, 2026Primary
Free and ungated. No sign-up wall. Ecom Relays is not a course or coaching program. We do not sell stores and we never promise income.
Next step
Ask a real expert. Free.
A real person replies within one business hour, 9am to 6pm US Central, Monday to Friday.