The Portal That Kept Getting Asked For

A sales rep who wanted a product on the site emailed the web team, or called them. That works for a narrow catalog. Filmtools carries grip, electrical, camera support, lighting, audio and a constant churn of expendables, so requests arrived in two inboxes and a phone queue with no shared record, and some of them were never seen again.

Nobody could answer the question both teams kept asking, which was where a given request had got to. Sales chased, web repeated themselves, and the same item got submitted twice. Product updates slowed, which on an e-commerce site is lost revenue rather than lost tidiness.

Both teams were in Magento and NetSuite every day. Neither has a reputation for elegance, and that was beside the point. I built the portal out of their conventions on purpose, because a tool that looks like the two systems you already have open is a tool nobody has to be trained on.

The research pointed at one feature above the rest. People wanted to know where a request stood, ahead of anything more sophisticated, so status went to the front of the interface as a refreshable indicator both teams could read at a glance.

Submitting a request was meant to take under a minute, so the test sessions were ten. That fit inside a sales rep’s day without scheduling, and short sessions turned out to be enough to find what was wrong. We piloted with a group from sales, fixed what they found, then opened it to the rest of the team.

Two things came out of alpha that were not in the original scope: bulk uploads, and real-time notifications. Both went in. The rest of the changes were smaller, simplifying form fields and sharpening the status indicators until a submission held to its minute.

The portal went out and then sat in testing for two or three weeks while we kept watching people use it. That is the part of the timeline that mattered, because other departments could see it working, and purchasing came over with a problem of their own.

They were maintaining vendor inventory in NetSuite by uploading spreadsheets, hundreds of SKUs at a time, and the counts refreshed once a day in the morning. Everything after that was a guess. The staff cross-checked the ERP against the website by hand, kept correcting the same discrepancies, and phoned vendors to ask what was true.

A stale count is not an internal inconvenience. It reaches a customer as an item that was in stock at checkout and is not in stock afterwards.

I started from the request table the portal already had rather than designing a second thing. Purchasing had watched sales learn the interface, so reusing it meant the training cost was close to zero and the argument for adoption was already made.

The adjustments were about control. Accurate list counts and the pull and sync actions moved into the open instead of sitting behind a menu, and testing added per-item syncing so staff could correct one SKU without triggering a run across the whole catalogue.

Purchasing was the second department, not the last. The portal went on to absorb shipping through ShipStation, accounting through QuickBooks, order fraud prevention through Signifyd, and the direct B2B sales flow. By the time I left Filmtools it was the tool those teams ran the day on, which is a long way from a form for getting a product onto a website.

I designed all of them. The first two are the ones written up here because they are the ones I kept the artifacts for, and the pattern behind the rest is the one visible in the second: a team watches another team’s problem get solved in a tool they can see, and comes back with their own.

The portal moved request traffic off email and phone, and gave both teams a status they could check themselves. Onboarding a new sales rep got faster too, which is the dividend of having built the thing out of interfaces they already knew. The landing dashboard we had listed as a future step shipped, so requests got an at-a-glance view as well as a searchable one.

The vendor integration ended manual uploads for that vendor and put its inventory on an hourly sync. Purchasing got two to three hours a week back, and the customer-facing result is the one I would lead with: items that had been sitting as out of stock became buyable again, checkout went up 12%, and cancellations fell 8%.

I would not have found the inventory problem by interviewing my way to it. Purchasing brought it to me because they had watched a tool work for somebody else, and so did the three departments after them. Shipping the small version and leaving it where people can see it did more discovery than any research plan I could have written.

Two things I would change. We ran the first phase on a color-coded Google Doc with no project manager and no tracking tool, which held together and cost me time I could have spent on the design. And I found out too late that the vendor’s API needed access I did not have, which stalled testing on the second phase for days that were already scarce in a two-week window.