A checkout that looked like a demand problem.
A client workflow review. There was traffic, a respected handmade product, and customers willing to talk, but no verified clean website sale in months. The obvious answer was more marketing. The useful answer came from tracing the whole path from product discovery to deposited payment, and separating real customer friction from misleading order and analytics data.
About this brief.
This is a real engagement, published without identifying the client. The business is a one-person manufacturer of handmade custom guitar cabinets, bass cabinets, and pedalboards. The findings, repairs, tests, campaign figures, and unresolved constraints are all real. The business and owner are not named, product line names have been generalized, and no customer information appears here. The work began as ongoing website and conversion support rather than a one-week review, and has been reorganized into the format an AI Workflow Review delivers.
- Client: a one-person custom music-gear manufacturer.
- Workflow: website visitor to completed order.
- Period covered: June to August 2026.
- Platform: WordPress, WooCommerce, WooPayments, Mailchimp, and UPS shipping calculation.
A Workflow Review defines one business question, maps the real workflow, finds the constraint, compares the realistic options, assigns an owner, and states what the evidence does and does not prove. That last part is why this brief is worth reading. It ends without the result anyone wanted.
The question we were asked.
The owner's concern was direct: the store had not produced a clean website sale in months. He was still creating some business through manually sent invoices, but activity was slow, and summer seasonality did not fully explain the silence from the store.
The natural response was to generate more demand. Send more email, attract more traffic, add something clever to the storefront. But low sales come from two very different conditions: not enough people want the product, or interested people cannot reliably reach and complete a purchase. Those look identical from the revenue line and are distinguished only by where in the path people stop.
Scope of the review. One workflow: the path from a visitor discovering the shop to a valid order, a successful payment, a confirmation, and a deposit. That included the homepage, navigation, product configuration, cart, shipping calculation, checkout, payment state, order records, analytics, and follow-up email.
How the work actually ran.
This is not a high-volume catalog business. The owner designs and builds custom gear, answers product questions, handles sales conversations, and manages fulfillment. A single product may require the buyer to choose cabinet dimensions, covering, grill cloth, hardware, and other options before the order is even ready to place.
That makes the owner's attention the scarce resource. Any proposed solution that adds another dashboard, another daily task, or another system needing care is working against the constraint rather than with it. A recommendation has to remove uncertainty or help a qualified buyer complete the next step.
- Discover: a player finds the shop through organic search, social activity, email, a referral, or prior knowledge of the brand.
- Navigate: the visitor uses the homepage or menu to reach the relevant cabinet or pedalboard category.
- Consider: they review photos, dimensions, configuration choices, price, and the maker's credibility.
- Configure: they select the required product variations and add-ons.
- Cart: the chosen product and configuration enter the cart correctly.
- Ship: the store calculates a UPS Ground estimate and updates the order total.
- Pay: a valid payment path is presented, whether card fields, Apple Pay, Google Pay, or PayPal.
- After: the buyer receives confirmation, the order is visible to the owner, payment reaches the connected account, and the post-purchase email sequence begins.
Where it broke.
The revenue line made this look like a demand problem. Tracing the path found multiple failures and blind spots much closer to the purchase.
Payment state
Payments had previously been left in test mode. One order looked paid and completed inside the store, but the order itself showed the payment gateway was in test mode. It created a normal-looking record without accepting real funds. That order could not be used as evidence that live checkout worked, and it could not be counted as revenue.
Cart and checkout configuration
Earlier work had identified stale cart behavior and competing payment options that could create dead ends. These were cleaned up before the later end-to-end audit.
Purchase controls
Some product purchase controls did not behave reliably. A pedalboard button had to be repaired so it worked when the page loaded. During later testing, configurable guitar cabinets sometimes required a second Add to Cart click after the selected variation initialized. That was not assumed to affect every product, and it remained a separate product-level issue rather than being folded into the main diagnosis.
Navigation and merchandising
The homepage and navigation exposed discontinued, empty, or broken destinations. Rack-case links returned 404 pages, speaker destinations were empty or malformed, a discontinued pedalboard line was still linked, some cabinet links pointed to empty categories, and the main shopping calls to action did not lead to the intended storefront.
Measurement
The analytics dashboard reported zero purchase events even after a controlled live transaction succeeded. This is the finding that undermines every other number. Without it, the store could not distinguish no sales from untracked sales, and no amount of additional traffic would have resolved the ambiguity.
Fraudulent noise
The order list contained a large cluster of failed attempts with suspicious identities, addresses, and network patterns. Several checked addresses appeared fraudulent. These were not treated as genuine customers who tried and failed to buy. They created operational noise and made the order history harder to read honestly.
The cost of getting the diagnosis wrong. Sending more traffic into this system would not have answered the sales question. It would have mixed real visitors, broken routes, inconsistent product controls, untracked conversions, controlled tests, and likely automated fraud into a single misleading number, and it would have cost money to do it.
Options we compared.
Rejected for now. The site already drew 535 visitors in the reviewed 28-day period, with organic search supplying the largest share. Buying traffic before fixing navigation, product controls, and measurement would amplify the uncertainty rather than resolve it.
Rejected for now. An assistant could eventually help buyers choose between custom configurations. It could not fix test-mode payments, dead links, variation behavior, fraud noise, or missing purchase events. It would have put a polished layer over unresolved commerce plumbing.
Rejected. The controlled live transaction proved the existing platform could complete a real charge. The platform was not the constraint. Its configuration, navigation, product behavior, and measurement were.
Chosen. Repair the transactional path, remove discontinued and broken routes, make the real storefront obvious, keep a direct line to the owner for configuration questions, and verify every step across devices before spending more on demand.
The second option is the one worth dwelling on. It was the most fashionable answer available, it would have been straightforward to build, and it would have been the wrong thing to do. A review that cannot reach that conclusion is not worth commissioning.
Recommendation: change.
Repair and verify the sales path before investing in more traffic or adding new software.
- Restore and verify live payments. Remove test-mode ambiguity and confirm the connected payment account with a controlled live transaction.
- Clean the cart and checkout. Remove conflicting payment paths and stale cart behavior, then verify that shipping and totals update correctly.
- Repair purchase controls. Make sure products can be configured and added to the cart without hidden initialization steps.
- Rebuild navigation around current products. Remove discontinued lines from customer-facing navigation and point the header and storefront links only to relevant, populated categories.
- Make the shop the real storefront. The primary homepage and footer shopping links should lead to an actual product-shopping experience.
- Keep a human contact path everywhere. Show the public phone and email so a qualified buyer can resolve a configuration question instead of abandoning the purchase.
- Connect post-purchase email. Capture completed customers and begin the follow-up sequence automatically.
- Add fraud protection carefully. Reduce automated card-testing noise without automatically rejecting legitimate high-value custom orders.
- Repair purchase tracking. Align orders, payments, deposits, and analytics so a completed order appears consistently in the reporting.
How we will know it worked
Every customer-facing header, submenu, homepage, shop, and footer link reaches a valid and relevant destination.
Tested products enter the cart with the selected configuration intact.
UPS Ground returns an estimate and the checkout total updates.
A controlled live payment succeeds and appears in both the gateway and the connected account.
Valid purchases appear as purchase events and reconcile with the order records.
Failed automated attempts decline without a rise in legitimate high-value declines.
A real customer, not the owner and not a test order, completes a website purchase.
Where a person stays in the loop
- The owner decides which products and categories are genuinely still offered.
- The owner personally handles high-value configuration questions and reviews ambiguous declines.
- No plugin subscription or paid service is purchased without his approval.
- Campaign copy and post-purchase messages stay subject to human review.
- I verify technical behavior. The owner confirms whether the resulting order and payment match the business reality.
What was deferred, and when to revisit.
Three ideas were set aside. Each could be useful later, and none addressed the immediate constraint.
AI-assisted product copy. The owner's spoken knowledge about cabinet construction, tone, materials, and buyer fit is valuable and time-consuming to turn into consistent product pages. AI could draft descriptions from his notes or from interviews, with him approving every claim before it is published. This is a genuine fit: the work repeats, the input already exists, and the person reviewing it can tell immediately whether the output is right.
A guided product assistant. A storefront assistant could eventually help a player choose a cabinet size, configuration, or pedalboard. It should be reconsidered only once product data is consistent and the underlying purchase controls are dependable.
Visual swatches. Covering-material swatches were placed behind the higher-priority sales-path cleanup. The licensing position on the existing swatch plugin could not be confirmed, and no replacement was to be purchased without approval.
Trigger to revisit. After thirty days of stable checkout, trustworthy purchase tracking, and enough real customer behavior to identify the next bottleneck. Not before, because until measurement is trustworthy there is no way to tell whether any of these helped.
What happened.
Transactional verification
The owner completed a controlled live debit-card transaction of $17.27. The gateway recorded the charge and sent the expected order emails. Shipping was intentionally not fulfilled, because this was a test product and a test transaction rather than a customer sale. A separate clean audit confirmed that the cart opened, UPS Ground calculated shipping, checkout totals updated, and the card, Apple Pay, Google Pay, and PayPal interfaces all loaded.
Navigation and storefront cleanup
- Top navigation updated to point at relevant products.
- Discontinued and 404 destinations removed from customer-facing navigation.
- The shop made the true storefront, with footer links reviewed and confirmed.
- Public phone and email contact options confirmed across the site.
- Navigation tested across platforms, and tested products entered the cart successfully.
Demand test
One focused campaign was sent to a known audience.
Open rate.
Click rate.
Clicks that went to the product page.
That is meaningful product interest. It is not, by itself, proof that the store was producing completed customer sales.
Traffic context
- 535 visitors across the reviewed 28-day window, up 3.9% on the preceding period, and 681 sessions in the store-performance view.
- 54.6% engagement rate, with organic search supplying about 54.9% of visitors and direct traffic about 32.3%.
- 3.1K impressions and 249 clicks reported in Search Console, with 297 unique visitors from search.
- Returning visitors down 27.8%.
- Zero purchase events reported by analytics, despite the controlled live transaction having succeeded.
The honest result
The work proved that live payment, shipping calculation, navigation, contact, and Add to Cart could all function, and it removed obvious customer-facing dead ends. It did not prove that independent customer sales had resumed. As of the last verified review there was no evidence of an independent website sale in the period examined, and this brief does not claim the work produced revenue.
What changed is what can now be trusted. The next constraint is trustworthy measurement and product-page conversion, not raw traffic. That is a smaller and more answerable question than the one we started with, which is the outcome a review is actually for.
Next step and owner.
- Next step: watch for the first verified independent website sale and reconcile it across the order records, the payment gateway, the deposit account, and analytics.
- Technical owner: me, for purchase tracking, the remaining configurable-product behavior, and verification of technical changes.
- Business owner: the client, for product availability, configuration rules, customer follow-up, and final approval of any paid plugin or service.
- Security follow-up: add and tune fraud protection so automated attempts fall without blocking legitimate custom orders.
- Conversion follow-up: improve the highest-traffic product pages and the configurable-cabinet purchase controls, measured by completed carts and purchases rather than pageviews.
- Revisit after 30 days of stable data: AI-assisted product copy, a guided product-selection assistant, visual swatches, and additional demand generation.
- Not doing yet: replatforming, paid traffic, or an AI storefront layer, without clean measurement and stable checkout completion.
The review ends with a named owner and an evidence threshold because "the website works" is not the business outcome. A real customer completing a tracked purchase is.