You Don't Have a Data Problem. You Have a Plumbing Problem.

I hear this in almost every first call with a business owner: “We need to get our data in order before we can do anything with AI.”

It sounds reasonable. Responsible, even. But nine times out of ten, it's the wrong diagnosis. You don't have bad data. You have data locked in systems that were never designed to share it.

Your ERP knows your inventory. Your eCommerce platform knows your orders. Your accounting software knows your margins. Your CRM knows your customers. Your email platform knows who opened what and when. Every system is doing its job. The problem is that none of them know what the others know.

That's not a data problem. That's a plumbing problem.

The Silo Tax

According to McKinsey's research on data-driven enterprises, companies that break down data silos see 20-30% improvements in operating margins. Not because they collected more data — because they connected what they already had.

Think about what happens in a typical mid-market company when a customer calls to ask about their order:

• The rep opens the CRM to find the customer record
• Switches to the ERP to look up the order status
• Opens the shipping portal to check tracking
• Maybe pulls up email to see if anyone sent the customer an update
• Possibly checks Slack to see if fulfillment flagged anything

Five systems. One question. And that rep just spent three minutes doing what a connected system does in three seconds.

“The average enterprise uses 130 SaaS applications. Only 29% of them are integrated with each other.” — Productiv, State of SaaS 2024

That 71% gap is the silo tax. You're paying for it in slower decisions, duplicate data entry, missed insights, and employees who spend half their day copying information from one screen to another.

Why “Clean the Data First” Is a Trap

The “we need to clean our data first” instinct comes from a good place. But it creates an infinite loop. Here's why:

Data gets dirty because of silos. When the same customer exists in three systems with slightly different spellings, that's not a data quality problem — it's a plumbing problem. When order totals don't match between your ERP and your marketplace, that's not bad data — it's two systems that never reconcile automatically.

Cleaning data without fixing the plumbing is like mopping the floor while the pipe is still leaking. You can do it. It'll look great for about 48 hours. Then you're back to mopping.

Harvard Business Review found that most companies spend 80% of their data budget on cleaning and preparation — and only 20% on actual analysis. That ratio is backwards, and it stays backwards as long as the underlying connections are broken.

What the Plumbing Actually Looks Like

When I walk into a new engagement, I don't start with a data audit. I start with a systems inventory. I map every tool, every database, every spreadsheet, every API — and most importantly, every gap between them.

Here's what I typically find in a $5M-$20M company:

• 8-15 core systems running the business (ERP, eCommerce, CRM, accounting, email, shipping, POS, marketplace integrations)
• 3-5 of them share data in some automated way
• The rest are connected by humans — manual exports, copy-paste, “Sarah checks that every morning”
• At least 2 critical spreadsheets that someone built years ago and nobody fully understands but the entire operation depends on

The fix isn't replacing all of those systems. The fix is connecting them. API integrations, webhooks, scheduled syncs, unified data layers. The vendor trap isn't having too many tools — it's having tools that don't talk.

The Three Levels of Integration

Not everything needs real-time sync. I think about integration in three tiers:

Level 1: Visibility. Can you see data from System A while working in System B? Even a read-only dashboard that pulls from multiple sources eliminates 80% of tab-switching. This is the fastest win and costs the least. A unified intelligence dashboard that connects your ERP, eCommerce, and CRM into a single view.

Level 2: Sync. When something changes in System A, System B updates automatically. Orders placed on Shopify appear in your ERP. Customer updates in HubSpot reflect in your email platform. This eliminates manual data entry and the errors that come with it.

Level 3: Orchestration. Systems don't just share data — they trigger actions in each other. A delivery confirmation from the shipping API triggers a follow-up email from marketing, updates the CRM, and flags the rep for a phone call. This is where AI agents live — sitting on top of connected systems, watching the data flow, and acting when patterns emerge.

The Real Blocker Isn't Technical

I've never walked into a company where the technology was the actual blocker. APIs exist for virtually every major business tool. The connectors are there. The documentation is there.

The real blocker is organizational. It's the fact that the marketing team chose their email platform without talking to the sales team. It's the CFO who bought the ERP based on a demo and never asked whether it had an API. It's the “IT guy” who left two years ago and took all the institutional knowledge about how the systems connect.

Gartner estimates that through 2026, organizations that promote data sharing will outperform their peers on most business value metrics. The emphasis is on sharing — not collecting more.

You don't need more data. You need fewer walls between the data you already have.

A Real Example

A precious metals company I work with had seven reps, a NetSuite ERP, a Shopify storefront, an eBay store, a Klaviyo email platform, a HubSpot CRM, and a VoIP phone system. All functioning independently. Here's what they didn't know:

• Which customers had been contacted by which rep this month
• Whether a customer who called in also had open orders on eBay
• What their actual blended margin was across all channels
• Which email campaigns were driving phone calls (not just clicks)

None of that required “better data.” NetSuite had the orders. Shopify had the products. Klaviyo had the engagement. HubSpot had the pipeline. The phone system logged every call. All the data existed — in seven different places that never compared notes.

We connected them. Built a unified data layer in Supabase. Synced NetSuite orders, Klaviyo engagement, and phone records into a single view. Added AI agents on top to monitor the whole picture. Now a rep can see a customer's complete history — orders, emails, calls, browsing behavior — in one screen. The full case study goes deeper.

The data wasn't dirty. The pipes were just disconnected.

Where to Start

If this sounds like your company, here's what I'd do Monday morning:

1. List every system your business touches. Every SaaS tool, every spreadsheet, every database. Include the ones you forgot about.
2. Draw the connections. Which systems share data automatically? Which ones require a human to move information between them?
3. Find the highest-pain manual bridge. Where does your team spend the most time copying data between systems? That's your first integration.
4. Check the APIs. Almost every modern tool has one. If yours doesn't, that's a different conversation — but that's a vendor problem, not a data problem.

The Stack Audit tool on this site will walk you through it interactively if you want a structured approach.

You're closer to AI-ready than you think. The data is there. The systems work. You just need someone to connect the pipes.