Every so often, a new generation of data infrastructure tools shows up promising to make everything faster — faster ingestion, faster modeling, faster time to a working dashboard. Those promises are easy to make in a slide deck. So instead of taking anyone’s word for it, we decided to test two of the newest ones the way we’d want a distributor to test us: by actually building something.
The rules were simple. Pick a vertical with a genuinely messy, multi-system operation. Pick a business problem specific enough that a generic dashboard couldn’t fake its way through it. Then build the whole thing — from raw data to a working, chat-enabled dashboard — and see how fast it could actually go, without cutting corners on accuracy or governance along the way.
The vertical we picked: food distribution. The variable we picked: shrink.
Shrink Means Something Different in Food
Shrink in a general distribution business is mostly damage and theft. Shrink in food distribution is a different animal. It’s expiry sitting on a shelf too long. It’s a temperature excursion on a truck that idled at the dock. It’s a customer reject on a pallet that missed a delivery window by six hours. It’s a short-date pull that has to move today or it’s a write-off tomorrow. Four or five distinct failure modes, each with its own root cause — and in most operations, each living in its own spreadsheet that nobody has time to reconcile with the others.
That’s the kind of problem that makes a good proof of concept: real, painful, and impossible to solve with a single native ERP report.
A margin dashboard that ignores shrink isn’t wrong exactly — it’s just answering last quarter’s question. In food distribution, the real margin story is written in the product that never made it to a paid invoice.
Meet PurePath Food Distribution
So the team built PurePath — a fictional, multi-location food distributor carrying the full operational stack a real one would run: an ERP, a CRM, accounts receivable, and the day-to-day chaos that comes with perishable inventory moving through a supply chain. Then they built the pipeline behind it: raw data connected and cleaned, a governed data warehouse structured around the questions a food distributor actually asks, and a visualization and chat layer sitting on top — so anyone on the team can ask a question in plain English and get an answer instead of a spreadsheet.
No named tools, no product pitch here. Just the four-layer architecture, built end to end, under a clock.
13 Days
From a blank slate to a production-ready dashboard — chat agent included
PurePath Proof of Concept Build Log
12.4M
Synthetic, AI-generated data rows built to power the model
PurePath Proof of Concept
132
Business measures defined across the dataset
PurePath Proof of Concept
27
Fact and dimension tables in the warehouse model
PurePath Proof of Concept
Underneath those numbers sat 49 relationships connecting the data together — and the part the team got genuinely excited about, 14 hidden signals the model surfaced without being explicitly asked to look for them. Patterns like a supplier whose temperature excursions cluster on one specific route. A customer whose reject rate spikes every time a particular SKU ships. The kind of thing a sharp analyst could eventually find by hand, given enough coffee and enough hours. The kind of thing a properly connected model finds by Tuesday.
What The Dashboard Actually Shows
The build wasn’t a single chart. It’s a full operational picture, with every module pulling from the same connected data instead of five different exports:
| Dashboard Module | What It Answers |
|---|---|
| Margin erosion by SKU, customer & territory | Which relationships are quietly losing money once freight, handling and shrink are counted |
| Inventory health | Days on hand, slow-moving stock, and which SKUs are closest to becoming a write-off |
| Supplier performance | Fill rate, return rate, and which vendors are the real source of temperature and short-date issues |
| Sales & territory performance | Revenue and margin by rep, account activity, and who hasn’t ordered in 60+ days |
| Cash flow & AR visibility | Aging by customer, collections risk, and working capital tied up in receivables |
Every one of those views is a starting point, not an endpoint. Click into any number and the dashboard drills straight through to the transactions sitting behind it. And for the questions that don’t fit neatly into a chart, there’s a chat interface layered on top of all of it — type “which customers had the highest return rate on refrigerated SKUs last quarter” and get an answer back, not a support ticket.
One honest note: every row of data behind this dashboard is synthetic. No real invoices, no real customers, no real shrink. That was intentional. The point of the challenge was to prove the entire pipeline — raw data to a trustworthy, drill-down-ready answer — could be built fast and built right, before a single real record ever touched it.
See It In Action
Rather than describe the dashboard any further, here’s the walkthrough — a full tour of PurePath’s inventory, supplier, sales, and cash flow views, plus the chat agent answering questions live.
None of this replaces the real work of connecting an actual distributor’s actual systems — that part still takes real discovery, real data, and real judgment about which questions matter most to your business. But the challenge answered the question we set out to ask. The newest generation of data infrastructure tools can move genuinely fast, without cutting corners on governance or accuracy to get there.
For a food distributor sitting on years of ERP data and no real way to see all of it at once, this is a meaningfully shorter runway than it used to be.