What are businesses struggling with right now—and what practical software, automation, and engineering answers could help?
This research report translates recent public complaints from merchants, app developers, Canadian small businesses, industrial teams, and boat owners into focused solution opportunities. It separates what people reported from the proposed response. Source links support the observed problems; tool ideas and opportunity scores are SolvProb analysis, not claims made by the original posters.
Quick answer: What problems are businesses searching for help with right now?
Across the latest discussions, recurring problems include incorrect inventory after refunds, ecommerce systems that disagree with each other, missing bulk actions, misleading analytics, failed platform migrations, manual evidence gathering, incomplete maintenance work orders, controls-software version risks, and poorly structured marine service intake.
The practical opportunity is rarely “add more AI.” It is usually: take data people already have, identify exceptions, explain the result, and produce a file they can safely use.
Jump to a problem area
- Ecommerce and retail operations
- Apps, analytics, and platform reliability
- Canadian small-business operations
- Electrical engineering, controls, and industrial maintenance
- Marine vessels, ownership, and service
- What these problems have in common
- Frequently asked questions
Ecommerce and retail operations
These are not abstract feature requests. They affect stock, pricing, fulfilment, advertising decisions, and customer trust.
1. Refunds can create phantom inventory
Search intent: Why does Shopify show inventory that I do not actually have?
Observed problem: A merchant explained that the default “Restock items” choice during a refund can increase inventory even when no physical item was returned. The nonexistent unit may then sell, creating another refund and customer failure.
Source: Read the original Shopify Community complaint
Practical answer: Reconcile refunds, physical returns, and inventory adjustments. A Phantom Stock Auditor could import those exports, flag refunds without receiving evidence, and produce a recount and correction queue.
2. Marketplace deposits hide channel fees
Search intent: How do I reconcile Shopify with Amazon, Walmart, or eBay payouts?
Observed problem: Shopify can record gross marketplace orders while the marketplace subtracts fees before sending the bank deposit. The merchant is left matching three different views: sales, settlement, and cash.
Source: Read the marketplace-fee discussion
Practical answer: Import order, settlement, and bank CSVs into one clearing-account reconciliation. Output matched orders, fee categories, missing deposits, and unexplained differences for bookkeeping review.
3. Per-unit rounding changes discounts on low-price items
Search intent: Why is Shopify discount math wrong for products sold by length?
Observed problem: A fabric merchant demonstrated a case where per-unit rounding produced a €0.80 discount instead of the €1.36 expected from the line subtotal. Even a proposed free app reportedly reproduced the wrong answer.
Source: See the discount-rounding evidence
Practical answer: A deterministic calculator should show unit price, quantity, raw discount, rounding stage, and final line total. This is safer than hiding the calculation inside another discount app.
4. B2B prices can change between Admin and POS
Search intent: Does Shopify POS support B2B catalog pricing?
Observed problem: A merchant discovered that POS would not directly pull prices from a B2B customer catalog. A suggested workaround starts the sale as a B2B draft order in Admin and then loads it into POS Pro.
Source: Read the B2B pricing discussion
Practical answer: Before adopting the workaround, reconcile catalog price, company location, tax treatment, loaded draft, and final POS line price. Do not create a second cost-plus pricing system unless it can be kept synchronized.
5. Storefront search can look relevant while returning catalog noise
Search intent: Why did Shopify search show hundreds of unrelated products?
Observed problem: Two merchants independently reported irrelevant search results. One said a search for “duck earrings” found the right items and then more than 900 unrelated products; another saw unrelated results for “deer.”
Source: Read the first search complaint
Practical answer: Audit real buyer queries against titles, tags, descriptions, variants, and semantic matches. Report top-20 precision and explain why each questionable product matched before changing synonyms or catalog content.
6. Hundreds of held orders still become manual work
Search intent: How do I bulk release Shopify orders on hold?
Observed problem: A long-running Shopify discussion says bulk order release remains unavailable. Merchants previously described more than 400 preorder orders and roughly four hours of repetitive release work.
Source: Read the bulk-release request
Practical answer: Use an export-first planner to identify eligible and ambiguous orders, reconcile totals, create safe batches, and generate a Flow recipe or reviewed mutation file without immediately modifying the store.
7. Draft-order updates can erase operational identifiers
Search intent: Why did my Shopify custom SKU disappear?
Observed problem: A developer reported custom-item SKUs disappearing from Shopify Admin. Editing another draft-order line allegedly reset custom SKUs to null, allowing paid orders to reach downstream processing without identifiers.
Source: Read the SKU-preservation report
Practical answer: Compare intended draft-order lines with current and paid-order exports. Flag missing or changed SKUs and prepare a restoration file plus evidence for support.
8. Partial pickup status can misrepresent the order
Search intent: Why did a customer think every item was ready for pickup?
Observed problem: A baby-gear retailer said an email correctly listed two items ready for pickup, while the customer account page implied that all five were ready. Some customers travel three hours each way.
Source: Read the partial-pickup complaint
Practical answer: Generate a line-by-line “ready now” versus “still preparing” summary from order and fulfilment exports, plus corrected notification copy and a staff exception queue.
9. Refund activity is not the same as a physical return
Search intent: Why does Shopify report more returns than I physically received?
Observed problem: One merchant reported a platform return rate near 25% while physical returns were closer to 3%. Another example showed a negative return amount despite no parcels coming back.
Source: Read the returns-metric discussion
Practical answer: Separate cancellations, edits, exchanges, refunds, and received return parcels. Produce both financial-refund and physical-return rates with transparent definitions.
10. Product availability signals can contradict each other
Search intent: Why does AI say my in-stock product is unavailable?
Observed problem: A merchant said AI assistants repeatedly described an available product as out of stock. Inspection found correct availability in one product endpoint but duplicate structured data, invalid JSON, and several machine-readable surfaces. The exact source used by an AI assistant was not proven.
Source: Read the AI-availability discussion
Practical answer: Compare visible purchase state, Product and Offer schema, product JSON, canonical URLs, markets, currency, and agent-facing catalog output. Report contradictions without pretending to know which source a model used.
Apps, analytics, and platform reliability
Many expensive failures happen at transitions: migration, refresh, deployment, uninstall, or a race between two valid operations.
11. Expiring-token sessions can vanish between background requests
Search intent: Why did my Shopify offline session disappear?
Observed problem: An application logged a Prisma offline session with a refresh token on September 18, then found no row on September 19 despite no explicit delete in that code path. Opening the app recreated the session. The exact cause was not established.
Source: Read the session-disappearance report
Practical answer: Combine session snapshots, database audit events, refresh calls, and background-request timing into one state timeline. Look for concurrency windows and missing observability before blaming a specific library.
12. Theme settings and rendered CSS can diverge
Search intent: Why does Shopify Theme Editor show a change but the storefront does not?
Observed problem: A developer reproduced a deployment where Theme Editor showed a new value while a CSS asset still served the previous one. A second CLI push corrected one path, while settings-only API writes reportedly remained stale.
Source: See the reproduction steps
Practical answer: After deployment, compare expected settings with fetched assets and version parameters. Treat Admin confirmation as one signal, not proof that customers received the new output.
13. Device cart and server-side bundle state can disagree
Search intent: Why did Shopify POS change my bundle quantities or total?
Observed problem: A developer documented a Cart Transform case where a merged bundle disappeared server-side and loose quantities increased while POS continued showing the expected cart. One example compared 3,074.90 on the device with a quoted 2,262.40.
Source: Read the detailed POS bundle report
Practical answer: Reconstruct line identity across device snapshots, transform inputs and outputs, and completed orders. Add regression fixtures for full component sets, duplicate bundles, and repeated additions.
14. Eligible checkouts may bypass the post-purchase impression
Search intent: Why are 15–20% of post-purchase offers never displayed?
Observed problem: Two developers reported post-purchase eligibility returning true while checkout reached the thank-you page before the extension rendered. One measured approximately 15–20% missing impressions.
Source: Read the corroborated discussion
Practical answer: Reconcile eligible checkouts with actual render, offer, and order events. Segment missing impressions by checkout path, device, and latency without overstating causation.
15. Billing migrations risk losing entitlement truth
Search intent: How do I migrate away from Mantle without breaking app billing?
Observed problem: Developers said Mantle had handled plan catalogs, entitlements, usage visibility, subscription state, and migration tooling. Its shutdown left gaps around discounts, usage billing, and merchant migration, with a September 30 billing horizon discussed.
Source: Read the Mantle shutdown discussion
Practical answer: Create a per-merchant continuity ledger that compares plans, usage, entitlements, and live subscription records. Every merchant needs an explainable expected state before migration.
16. False purchase events can contaminate advertising decisions
Search intent: Why does my Meta Purchase event fire on the homepage?
Observed problem: A merchant repeatedly produced a Meta Purchase event by loading the homepage. Network and configuration investigation later identified an active Event Setup Tool rule based on a storefront-domain condition.
Source: Read the event investigation
Practical answer: Import a sanitized HAR file and order timeline. Identify event emitters, purchase events without order IDs, value, or currency, browser and server duplicates, and the likely contamination period.
17. Embedded mobile printing can fail silently
Search intent: Why does print do nothing inside Shopify mobile or POS?
Observed problem: window.print reportedly produced no dialog, PDF screen, console error, or visible failure inside Shopify Android. Another developer reported similar behaviour across POS iOS versions.
Source: Read the mobile-print report
Practical answer: Maintain a compatibility matrix by environment, version, document type, and API path. Every test should record dialog appearance, promise resolution, print styling, and fallback download or share behaviour.
18. Delayed cart writes can reverse authorization state
Search intent: How do I prevent stale writes from overwriting newer Shopify cart data?
Observed problem: An app developer described an older delayed authorization-manifest write landing after a newer revision. Cart metafields did not provide compare-and-swap, ETag, or fencing primitives to enforce monotonicity.
Source: Read the stale-write discussion
Practical answer: Model the operation timeline, identify unknown timeout states, and test serialized write, readback, and repair strategies under synthetic latency. Backend truth alone does not prevent a late platform write.
19. Legacy customer pricing lacks a safe bulk migration path
Search intent: How do I move thousands of merchants to new app pricing?
Observed problem: An application with approximately 1,500 installed stores needed to preserve tenure discounts. Shopify confirmed that individualized App Pricing discounts did not have a bulk workflow.
Source: Read the bulk-pricing migration discussion
Practical answer: Import subscriptions, classify approved cohorts, calculate target pricing, and produce per-merchant tasks with before and after reconciliation. Do not hide exceptions inside one mass action.
20. Visible app-owned definitions may still be unwritable
Search intent: Why can I read a Shopify metaobject definition but not write it?
Observed problem: A TOML-defined app-owned metaobject appeared in GraphQL but an upsert said no definition existed. The setup worked the previous day and failed after a development-server restart despite the reported app prefix and write scopes.
Source: Read the metaobject identity report
Practical answer: Compare TOML, scopes, authenticated app identity, definition ownership, environment, and mutation errors in one preflight. Visibility and write ownership are separate questions.
Canadian small-business operations
The strongest Canadian signals involve administrative work that is too important to ignore but too small or specialized for existing software.
21. BC caregiver payroll and health-authority reporting are still manual
Search intent: Is there easier payroll reporting for family-funded care in BC?
Observed problem: A sleep-deprived BC family caregiver manages payroll for three to four full-time employees and up to three casual workers, pays by cheque, handles remittances, and manually prepares monthly reports containing regular hours, sick hours, and vacation pay.
Source: Read the original BC complaint
Practical answer: Do not replace payroll first. Convert existing payroll and shift exports into the required monthly report, highlight missing hours, and preserve a human-review trail. Validate the format with a BC bookkeeper or program administrator.
22. Cross-border brokerage holds need better evidence packets
Search intent: Why are Canada-to-US shipments held even with HS codes?
Observed problem: A BC seller reported shipments being held despite exact values and HS codes; another seller was said to have 19 flagged packages.
Source: Read the BC shipping discussion
Practical answer: Turn shipment exports, invoices, origin information, and broker requests into one evidence packet per parcel. The tool should organize facts for a customs broker, not make unsupported clearance determinations.
23. Growing firms cannot see who owns their technology risks
Search intent: When does a BC small business need managed IT?
Observed problem: A BC business with roughly 10–50 employees described growing work around Microsoft 365, backups, access, phishing, device setup, and staff departures. Full-time IT felt excessive, while relying on whoever was “good with computers” was no longer sustainable.
Source: Read the BC managed-IT discussion
Practical answer: Create a neutral ownership matrix for users, devices, access, backups, restore tests, vendors, and incidents. Use it to compare MSP scope and response promises without claiming to certify cybersecurity.
24. Small businesses cannot document everything—but key knowledge sits with one person
Search intent: How do I decide which business processes to document?
Observed problem: A Canadian owner said more processes were known by only one person, creating absence and onboarding problems. Writing a complete operations manual felt impossible.
Source: Read the Canadian process discussion
Practical answer: Score processes by frequency, customer and revenue impact, single-person dependency, recovery time, and change rate. Document the top ten first, assign owners, and set review dates.
25. Receipt digitization fails on the messy documents specifications ignore
Search intent: What scanner works for wrinkled and torn receipts?
Observed problem: An Ontario buyer needed to scan folded, crumpled, and torn receipts. A bookkeeper said long narrow gas receipts jam more often, papers should be sorted by width, and some documents need separation, repair, or phone capture. Another user raised firmware and computer-compatibility lifespan.
Source: Read the Ontario scanner discussion
Practical answer: Test the buyer’s actual sample batch before recommending hardware. Classify feeder-ready, pre-separate, repair, photograph, and manual-review documents, then calculate the exception rate.
26. Asset-backed Canadian projects fall between lender categories
Search intent: How do I finance a $2–3 million hospitality project in Canada?
Observed problem: A Canadian consultant planning a small resort or retreat in Ontario or BC described a worst-case cost near CAD $2.8 million: insufficient equity for conventional lending, too little history for institutional capital, and a project too small for many private-equity investors.
Source: Read the Canadian financing discussion
Practical answer: Build a sources-and-uses waterfall, funding-gap scenarios, sensitivities, and a missing-evidence checklist for review by a commercial mortgage broker, accountant, and lawyer. This should be planning support, not automated financial advice.
Electrical engineering, controls, and industrial maintenance
Engineering users repeatedly ask for better evidence, configuration preservation, and commissioning records—not black-box recommendations.
27. A 60× PID/PIDE unit trap can survive until commissioning
Search intent: How do I convert Rockwell PID gains to PIDE correctly?
Observed problem: A controls practitioner documented a 60× unit-conversion trap when transferring gains between Rockwell Logix PIDE and classic PID instructions. Proportional gain can remain unchanged, making the incorrect conversion look plausible.
Source: Read the PLC discussion
Practical answer: A verified calculator should show the source and destination equations, units, converted values, assumptions, and printable commissioning sheet. It should never hide the formula.
28. Normal operation does not prove safe-motion behaviour under faults
Search intent: How should safe motion be tested during commissioning?
Observed problem: A practitioner highlighted the difficulty of proving safe limited speed, safe stop, and position monitoring when an encoder fails, communication drops, or an axis is already moving near a limit.
Source: Read the safe-motion discussion
Practical answer: Create an engineer-reviewed fault-scenario matrix with initial state, injected fault, expected safe state, measured result, and evidence reference. It is an evidence tool, not automated safety certification.
29. Incomplete work-order preparation wastes technician hours
Search intent: Why did maintenance arrive without the correct parts?
Observed problem: A maintenance technician said a centrifuge work order contained the wrong parts, costing an hour, and estimated that poor preparation wastes hundreds of hours annually.
Source: Read the maintenance-planning complaint
Practical answer: Compare work orders, BOMs, inventory, reservations, and required tools before dispatch. Output missing or ambiguous parts, verified kit contents, and a job-readiness status.
30. Teams may not know whether the project file is authoritative
Search intent: How do I verify an offline PLC project matches the running controller?
Observed problem: A maintenance team needed to troubleshoot active alarms but could not verify whether its offline project matched the PLC. The alternative under consideration was wiping and downloading the presumed latest file.
Source: Read the PLC provenance problem
Practical answer: Create a read-only provenance pack: controller identity, firmware, upload and download timestamps, program hashes, hardware inventory, and comparison evidence. Do not recommend a destructive download when project authority is unknown.
31. Engineering-software upgrades can alter legacy project content
Search intent: Can newer PLC software delete unsupported devices from an old project?
Observed problem: An Omron user opened an older project in newer Sysmac Studio and received a warning that an unsupported network device was deleted. Running multiple versions or virtual machines introduced additional IT friction.
Source: Read the Omron compatibility discussion
Practical answer: Before opening the project, capture software, firmware, device-profile, and file-hash inventory. Compare object counts before and after, and preserve an immutable original plus a version-isolation plan.
32. Long-idle drives need manufacturer-specific restart evidence
Search intent: What should be checked before restarting a VFD after long storage?
Observed problem: A 60 hp, 480 V drive reportedly exploded during restart after a year-long shutdown. The restart schedule omitted a supplier-stated capacitor-reforming requirement, while practitioners disagreed about the exact cause.
Source: Read the industrial discussion
Practical answer: Build an asset register that pairs storage duration, model, manual revision, and manufacturer restart instructions with an engineer-review checklist. Do not infer the failure cause or generate unsupervised energization instructions.
Marine vessels, ownership, and service
The most useful marine tools prepare owners and technicians with better facts while staying clear of unsupported remote diagnosis.
33. “Free boat” decisions ignore title, transport, and first-year cost
Search intent: How do I know whether accepting a free boat is a bad idea?
Observed problem: A first-time owner was offered an unidentified boat for free and knew little beyond the fact that the engine was not seized.
Source: Read the boat-identification discussion
Practical answer: Turn photos, dimensions, title status, engine details, storage, and transport needs into an acceptance checklist and first-year cost range. The output should recommend a survey, not declare seaworthiness.
34. Owners lack a practical scope for stepping up in vessel size
Search intent: What training do I need before moving to a much larger boat?
Observed problem: A boat owner moving from a 22-foot centre console toward a substantially larger twin-engine vessel asked what training and preparation were appropriate.
Source: Read the larger-boat training discussion
Practical answer: Use vessel dimensions, propulsion, operating waters, and owner experience to create questions and a supervised training scope covering close-quarters handling, systems, emergency procedures, and passage planning.
35. Bilge problems arrive without fit, access, and wiring evidence
Search intent: How do I prepare information for a bilge-pump service call?
Observed problem: A boat owner described a sticking Sahara2 1200 float, unsuccessful cleaning, a replacement that did not fit, wiring beyond their comfort level, and repeated difficulty reseating a gasket in a cramped bilge.
Source: Read the bilge-pump discussion
Practical answer: A mobile intake should collect model and serial, compartment photos, replacement dimensions, fuse and switch observations, and safe test history. Generate a technician-ready work pack, not remote electrical instructions.
36. Marine electrical symptoms need structured fault evidence
Search intent: Why does my outboard trim motor keep running?
Observed problem: An owner said an outboard trim motor ran continuously after the battery was connected even after both trim switches were disconnected.
Source: Read the outboard-trim discussion
Practical answer: Record wiring configuration, switch-disconnection state, observed behaviour, fuse and relay identifiers, and safe voltage readings for a marine technician. Stop short of remote repair instructions when unintended motor operation creates injury risk.
What these problems have in common
Systems disagree
Admin, storefront, POS, exports, and accounting records each show a plausible but different version of reality.
Bulk work lacks safe review
Businesses can often automate a change, but cannot preview its exact effect, approve exceptions, or roll it back cleanly.
Evidence is scattered
The useful facts already exist in logs, CSVs, PDFs, photos, and manuals. The missing layer turns them into a decision-ready packet.
Transitions create failures
Refreshes, migrations, upgrades, uninstalls, refunds, and restarts are where otherwise healthy systems lose state.
Small teams need narrow tools
They do not need another platform. They need one precise utility that saves an hour, catches an exception, or prepares a professional handoff.
Human approval still matters
Financial, safety, customs, electrical, and marine decisions should keep a qualified person visibly in the loop.
Turn a repetitive problem into a working tool
SolvProb builds focused websites, browser utilities, and custom applications for work that costs time, creates mistakes, or falls between existing platforms.
Frequently Asked Questions
What kinds of problems make the best small web apps?
Look for a structured input, repeated decision, and valuable output. Examples include CSV reconciliation, evidence packs, compatibility checks, calculators, exception lists, and corrected export files.
Should a new automation connect directly to live business accounts?
Usually not at first. A local, read-only import tool is easier to validate, protects customer credentials, and lets the user review the proposed result before anything changes.
Can AI solve these problems automatically?
AI is useful for classification, explanation, and drafting. Deterministic rules should handle money, identifiers, engineering calculations, and record reconciliation wherever possible. High-risk decisions still need qualified review.
How should a business validate one of these ideas?
Collect 10–20 real examples, define the current time or failure cost, run the proposed output beside the existing workflow, and measure omissions, false positives, and time saved.
Why include links to public complaints?
Source links let readers inspect the original context, distinguish reported facts from interpretation, and continue the discussion. They also make the article more useful to search engines and answer engines seeking traceable evidence.
Do outbound source links create backlinks to SolvProb?
No. They are outbound citations. They improve credibility and topical context. Backlinks occur when other sites link to this article because it is useful, original, and easy to reference.
Research method and limitations
This article synthesizes public discussions posted or active between September 9 and September 19, 2026. Sources include Shopify Community, Shopify Developer Community, and public Reddit communities. Complaints are paraphrased to avoid unnecessary personal information. A forum report is evidence that someone experienced or described a problem; it is not proof that every user is affected or that the proposed cause is correct.
Opportunity scores and commercial estimates are SolvProb analysis. Engineering, cybersecurity, customs, financial, and marine concepts are framed as evidence and workflow tools, not substitutes for qualified professional judgment.