Industry: Media & Publishing | Size: Small business | Tools: Zapier + HubSpot + Google Sheets | Problem: Workflow Automation
When a small magazine publisher decided to stop entering ad sales data by hand, they did what a lot of small teams do: built a Zap with Zapier assistance. The logic was sound. New or updated rows in a Google Sheet—tracking ad placements across three magazine issues—should flow into HubSpot as contacts, companies, and deals. Account managers should be assigned automatically. It should just work.
Photo by Growtika on Unsplash — Google Sheets to HubSpot Zapier workflow automation
It didn't. The workflow crashed on every run. The publisher contacted Zapier support. Multiple attempts. Still failing.
What the workflow was trying to do
The Zap had real ambition: pull ad sales rows from a Google Sheet, clean up contact names and emails, create or update contacts and companies in HubSpot, associate them, create deals with revenue amounts from three issues, and assign each deal to the right account manager by name. For a first automation built with AI, it was architecturally solid. The problems were in the details.
Seven things that were silently wrong
A Connex consultant reviewed the workflow live on a 35-minute call. Seven issues surfaced:
Wrong field references. The Zap was using raw unformatted names from Step 1 instead of the cleaned output from the formatter steps. Every contact name entered HubSpot dirty.
Email on the wrong record type. The workflow was trying to write an email address into the company record. HubSpot doesn't store emails at the company level—they belong to contacts.
A redundant filter. One step checked whether a company existed before creating it—but the preceding step already handled that. The filter added friction without adding logic.
A broken deal-to-contact association. Step 14 was linking deals to the wrong step reference—pointing at Step 1 (the trigger) instead of Step 8 (the contact creation). Every deal was created without a contact attached.
Incomplete JavaScript for owner assignment. The code that mapped sales rep names to HubSpot owner IDs only handled one rep. Every other account manager was left unassigned.
No dynamic owner lookup. There was no mechanism to translate a rep's display name to their HubSpot-specific owner ID at runtime. The JavaScript patch was always going to be incomplete.
A contact duplication tradeoff that wasn't visible. Zapier's create-or-find pattern creates a new contact when no email match is found—it doesn't add a second email to an existing record. The consultant explained the tradeoff clearly, simplified the logic to allow duplicates that could be merged in HubSpot later, and documented it so the client understood what to expect in production rather than discovering it by surprise.
What was rebuilt during the call
The consultant fixed all seven issues live. The JavaScript owner-assignment code was replaced with a Zapier Formatter lookup table that converts each rep's display name to their email address, followed by a HubSpot "Get Owner by Email" step that retrieves the owner ID dynamically. Field references were corrected throughout. The redundant filter was deleted. The email field was removed from the company step. The deal association was pointed at the right contact ID.
The Zap was published before the call ended and ran without errors on the first test replay.
The outcome
The publisher now has a working pipeline from their ad sales spreadsheet to HubSpot. Three issues' worth of ad revenue data flows in as properly associated deals with correct owner assignments. The work that had blocked pipeline visibility—and had already consumed Zapier support exchanges and multiple self-repair attempts—was resolved in under an hour.
✅
After the call, the client's note: 'Solved on the call.' The Zap ran clean on the first replay after the session.
Why AI-built automations fail in specific, predictable ways
This is a common pattern. AI tools generate automation logic that is architecturally correct—the step sequence makes sense. Where they fall short is in operational specifics: which field reference is valid after a formatter step, how HubSpot separates contact emails from company records, how to translate a human-readable name to a platform-specific owner ID at runtime. These aren't creative problems. They're knowledge problems.
The gap between a well-structured Zap and a working one is often not the architecture—it's the seven small things that only surface when real data runs through.
If you have a Zapier workflow that looks right but keeps failing on real data, a single review session is usually enough to find what's off. Most of the issues above took minutes to fix once someone with Zapier and HubSpot experience was looking at the right thing.
Get help fixing your broken Zapier workflow
Connex consultants review Zapier workflows live—finding field reference errors, broken associations, and logic gaps that look invisible until real data runs through. Most workflow reviews resolve in a single session. Book a free consultation to get a second set of eyes on your workflow.
How a regional mortgage lender automated offline conversion tracking across Google, Bing, and Meta using Zapier and the Velocify API — no weekly manual exports.
How a mid-sized countertop manufacturer replaced their inbound status-inquiry phone queue with an AI voice agent connected to Moraware. Read the case study.
