How a Compounding Supplier Restructured 7,000 Formulas and Automated a 5,400-Record Update
A compounding supplier moved 7,000 formulas into a structured Airtable library with automated PDFs, then added member analytics across Xano, Stripe, and Retool.
TL;DR: The U.S. division of a multinational pharmaceutical compounding supplier had about 7,000 formulas and 2,000 ingredients behind a members-only formula library, with ingredients crammed into single cells and every PDF built by hand. Connex restructured the data in Airtable, automated PDF generation and approvals, and pushed a library-wide change across 5,400 formula records in one automated run. Later work gave the team its first real view of which members log in, view, download, and cancel.
Compounding formula library automation: a pharmacist's mortar and ingredient bottles. Photo by David Trinks on Unsplash
The situation
The company is the mid-market U.S. arm of an international supplier of compounding ingredients and services for both human and veterinary pharmacies. Part of its offering is an educational platform: subscribing pharmacy professionals log in, browse a library of compounding formulas, and download a formatted PDF for each one.
The library ran on a no-code stack built over several years by different hands. A Webflow site (with Wized and a Xano backend) served the member portal. Stripe handled subscriptions. Retool sat on top as an internal reporting screen. Microsoft tools ran the rest of the business, including Dynamics 365 for customer order history. The formula data itself lived in Airtable, while other global teams kept their versions in Excel.
The problem: a formula library that couldn't be edited safely
On the first call, the team's description of the data was blunt. Each formula's ingredients sat in one text cell. There was no separate ingredient list, no clean quantity or unit-of-measure fields, and no way to say "show me every formula that uses this ingredient."
That caused three practical problems:
PDFs were hand-assembled. Each downloadable formula sheet needed the right disclaimer, notes, branding, and dosage-form details based on its category. Old PDFs lived in shared drives with no reliable link back to the record.
Bulk changes were all-or-nothing. When a disclaimer or wholesale detail changed, thousands of records had to be updated, re-reviewed, and republished to the site. Nobody wanted to do that one record at a time.
Nobody could see member behavior. A year into the project, the team asked questions the stack couldn't answer: how many people got stuck in sign-up, how many cancelled, and whether people were re-registering to restart a 30-day trial.
What they'd already tried
The portal had been built by an outside agency on Webflow and Wized. The team wasn't happy with the ongoing support and was shopping for someone who could own the data side long term. They were also clear about constraints: everything had to stay in a Microsoft-friendly environment, and they wanted to keep the site they already had rather than rebuild it.
What we built
The work ran in phases over roughly two years, each scoped in small blocks of hours.
1. A structured formula database
A Connex consultant split the ingredient cell into a proper formula-to-ingredient join table with quantity and unit-of-measure fields, cleaned the existing data, and linked legacy PDFs back to their records. Field-level permissions limited who could edit what.
2. Review, approval, and PDF generation
New or changed formulas now go through an intake form and an approval step before they reach the database. Approval triggers automations that generate the formatted PDF and sync the record to the website. A duplicate-a-formula form lets staff start a variant from an existing formula, and an "Is it a duplicate?" flag keeps the original reference intact so the automation doesn't overwrite it with a generic value.
3. A library-wide update in one run
When the team needed to push a wholesale change across the whole library, the consultant built a Make scenario that moves each record to review, creates the live PDF, and returns it to approved. Before running it, the consultant sized the cost: about 7 operations per record, or roughly 38,000 operations for the full set. That set the plan choice (a monthly 40,000-operation tier rather than an annual commitment) before anything ran. The automation then processed all 5,400 records in one pass. (Make replaced operations with credits as its billing unit in August 2025, converting 1:1, so the same sizing math still applies.)
4. Member analytics the stack could finally answer
The first finding was architectural. Retool was only a display layer; Xano held the real activity data. So the work went into Xano:
Login timestamps written to each user record.
Stripe webhooks feeding a subscriptions table that tracks new, cancelled, and paused subscriptions.
Formula view events logged as they happen, alongside downloads.
A 10,000-record cap on the reporting endpoints removed, so totals stopped silently truncating.
New formula downloads written to an Excel sheet by webhook for the Microsoft-side team.
On top of that, the Retool dashboard now shows monthly versus annual members, active versus inactive status, last login date, downloads and real-time views by account, most popular formulas, member counts by country (with an "unknown" bucket instead of dropped rows), and downloads by team with internal staff accounts filtered out so they don't skew the numbers. Historical Stripe billing details were backfilled into the same records.
5. Fixes when something broke
Late in the engagement, checkout sessions started failing whenever a product's old price had been archived in Stripe. The fix pulls the current price dynamically at checkout and attaches a client reference ID to each session.
The results
All figures below are real, taken from the project records. We have not put a dollar estimate on this one: the client never gave us a per-record editing time, and we won't invent one.
~7,000 formulas and ~2,000 ingredients moved from single-cell text into a structured, queryable library.
5,400 records updated, re-reviewed, and republished in one automated run instead of one at a time.
~38,000 operations sized before launch, so the automation cost was known up front.
10,000-record reporting cap removed, so download and popularity totals reflect the full history.
First member-level visibility: logins, views, downloads, subscriptions, and cancellations in one dashboard, answering the "who's stuck and who's cancelling" question the team raised.
The broader outcome is control. Formula changes go through an approval step, PDFs regenerate themselves, and the team decides what to ship to the site instead of waiting on someone else to do it.
Client voice
We don't have a quotable testimonial for this one. What we do have: the team rated individual work sessions "Happy" in our feedback records throughout 2025 and kept booking new phases for roughly two years, moving from data cleanup to approvals to analytics as each piece landed.
FAQ
Do we have to rebuild our site to get this?
No. This project kept the existing Webflow, Wized, Xano, Stripe, and Retool setup and worked inside it. The main job was finding where the data actually lived (here, Xano rather than Retool) and connecting to that.
How do you keep automation costs from surprising us?
Measure a single record first. Here, one record used about 7 operations, which made the full 5,400-record run predictable before it started.
Can we track what members search for?
Not reliably in this setup. Search terms come in with misspellings and variations, so the team tracked views, downloads, and logins instead, which give a cleaner signal of what members actually use.
What access does this kind of project need?
Admin or API access to each system involved (here: Airtable, Xano, Stripe, and the automation platform), plus someone on your side who can approve test runs before anything touches live records.
Is your data stuck in one cell too?
If your team is maintaining a large library, catalog, or document set by hand, and every bulk change feels risky, we can help you structure it and automate the busywork without throwing out the tools you already use. Book a call with Connex.
A 30-person cabinet and countertop company swapped a macro-heavy sales spreadsheet for an Airtable bid form, AI-built PandaDoc contracts, and a Pipedrive sync.
A 50-person health-tech company connected HubSpot and Zoom to a HIPAA-scoped Notion workspace in six weeks, with 12 months of customer history backfilled.
A Czech smart-metering operation used Smartsheet, Pipedream and Adobe document generation to auto-build 100–200 photo-rich apartment handover PDFs a month.