Most manufacturers and distributors I meet have a master spreadsheet, three copies of it, and one person who knows which copy is right. Getting out starts with deciding what is true, then moving it.
Why is our product data such a mess?
Because it grew one spreadsheet at a time. Purchasing keeps one with costs, sales keeps one with prices, marketing keeps one with descriptions, and the website pulls from whichever was handy the day it was built. Nobody owns the whole thing, so every copy drifts.
What is a single source of truth for product data?
One system that every other system reads from. It can be your ERP, a product information system (a PIM), or a well-built database. Everyone else stops editing copies and starts reading from it. The spreadsheet can still exist as a view, but nobody types into it.
Who should own each field?
The person whose job breaks when the field is wrong. Purchasing owns cost and supplier. Sales or the owner owns price. Engineering or quality owns specs and weights. Marketing owns descriptions and photos. Write the list down, one owner per field, and give only that owner the right to change it.
Should we clean the data before we buy software?
Yes. Clean the categories first. If one spreadsheet says "countertops" and another says "surfaces, quartz," new software will import both and you will have paid to move the mess. Agree on the categories and the names with the people who sell and buy the products, then load it. AI tools are good at the cleanup itself: finding duplicates, matching the same part under two numbers, and filling missing specs from supplier PDFs for a person to check.
When do we need a knowledge graph instead of a database?
When the value is in how things connect. Parts that fit certain models, products that substitute for each other, kits built from components, customers tied to projects tied to suppliers. Rows and columns handle a list well and handle relationships badly. A graph stores the connections directly, so you can ask "what else fits this" or "who else is on this job" and get an answer.
I am the fractional CTO at Quartz Importers MN. I built an AI agent platform and a knowledge graph for the sales team that connects companies, people, and projects, and the platform took in the whole market the company sells into in about 30 days. The same approach works for products that relate to each other. The details are in the Quartz Importers case study.
What does it cost to get product data out of spreadsheets?
Two ways to pay for it:
- A fractional CTO at $3,500, $6,500, or $11,000 a month, three months minimum. I pick the system, set up the field owners, run the cleanup, and keep it running. See fractional CTO.
- A scoped build with a fixed quote, when you already know what you want and need it built. See AI agents and workflow automation.
Not sure which fits? Do I need a fractional CTO or a dev shop walks through it.
Answered by Jared Lukes, Minneapolis. Twenty-five years in brand, UX, and technology, currently holding fractional AI and technology seats at three companies. About Jared.