A distributor sitting on
a billion data points—
and pulling reports by hand.
Three hours on the floor. Twelve opportunities. One supplier-shaped elephant in the room. This is the operational ground-truth — and where the first fifty thousand feet of lift is hiding.
The headline in one sentence.
The client runs a high-volume, high-automation physical operation on a last-decade software stack — the warehouse is tighter than the intelligence layer, and the intelligence layer is where the next margin lives.
The physical plant is genuinely impressive — a million-dollar palletizer making perfect layers, voice-picking in the keg room, conveyor orchestration that would embarrass most warehouses twice its size. This is not a business that needs robots.
The leverage is one layer up. The operations team still spends an hour a week pulling reports out of the route-accounting ERP. Returns, mispicks, and route exceptions live in a group chat. Expiration dates from supplier bills-of-lading get keyed by hand. A salesperson walking into a major retail account cannot, in under five minutes, surface what the distributor has on that store's shelf.
The warehouse is a finished product. The intelligence layer is the next build.
The headline: a prompt-native command center sitting on top of the existing ERP data, with AI-assisted exception detection at the palletizer, mobile field intelligence for sales reps, and a smart-pallet play the distributor can lead rather than inherit.
Nothing proposed here requires ripping out the ERP, buying a $500,000 robot, or fighting the dominant supplier on its home turf. Every quick win sits next to the legacy stack, reads from it through APIs, and gives humans back cognitive time.
And the supplier constraint? It's real — but it only governs 80% of the sales call frequency, not the operations underneath it. That's the seam.
The stack as it stands.
Every system the client pays for, what it actually does, and where it's working versus where it's leaking. Systems described functionally; vendor names redacted.
Operations & Logistics
Admin, Comms & Exceptions
"I don't want to go into the system and run a report. I should be able to say — show me what this account has in distribution, what's our most profitable item, right there, while I'm standing in the store."
— Client Owner · recorded on siteTwelve plays, ranked.
Scored by speed-to-value and operational impact. Phase tagging indicates the sequence Approach would propose, not the ranking of importance.
Where the first month of work goes.
Five opportunities that share two traits: they sit on data the client already owns, and none of them require permission from the dominant supplier.
Stop running reports. Start asking questions.
Today, the loop is: someone wants to know how a specific SKU sold last week → someone logs into the ERP → filters → clicks → drills → ten minutes later, a spreadsheet. The client is paying for sophisticated data and getting last-century ergonomics on top of it.
The command center reads the ERP (and payroll, and fleet) through their APIs and lets anyone with permission ask — by territory, supplier, SKU, rep, day. "Show me top 10 accounts whose volume dropped quarter-over-quarter." "Which routes are under 600 cases tomorrow?" "Give me year-over-year performance by region." The answer appears in under five seconds.
This is identical in architecture to what Approach is already piloting for a PE-backed multi-state manufacturer running on three separate systems of record. The pattern holds; the data plumbing differs.
A camera at the right node, not a robot at the wrong one.
The ops lead walked through the conveyor flow: a picker can grab the wrong SKU from an adjacent bin, the palletizer will still wrap a perfect layer, and the first anyone knows about the mispick is when the account rejects it the next day. ~20 mispicks a day × ~$6 each = ~$44K in raw cost annually. The real cost is the customer call, the driver's wasted stop, and the rep's defensive posture.
The fix is a vision node — mounted cameras at the wrap point reading each case as it passes, cross-referencing against the ERP order manifest. Mismatch → screen goes red → operator corrects before shrink wrap. No new robots. No palletizer rip-and-replace. Just a smart eye at the moment it matters.
Approach's partner network includes vision AI vendors already deployed at national retail scale. This is a well-understood problem in other verticals — just not yet in this one.
The salesperson walks in. The phone already knows.
The client owner's own example said it best: walk into a retail account, want to know what the distributor has on that shelf, and it takes five minutes of running multiple ERP reports to piece together an answer. A rep in that same store, standing in front of a customer, doesn't have five minutes.
The fix: a phone-first, location-aware intelligence layer. Open the app in the parking lot, the account lights up — three-year history, current distribution, gap analysis against what the store should be carrying from the distributor's portfolio, last rejection or return, recommended next move. Tap, walk in, sell.
This is the CRM pattern Approach has been developing — "built by salespeople, for salespeople" — adapted to the distribution context. The feature set maps directly onto the client's DNA once the ERP pipe is open.
The feds require paper. They don't require manual keying.
The IT director surfaced this one directly. The dominant supplier sends trucks with paper bills-of-lading — which, yes, federal transport rules require for in-transit inspection. But the second that paper hits the receiving dock, it can be photographed, parsed, and cross-referenced against what the ERP expects. Today, expiration dates are keyed by hand from that paper into the system — and the supplier regularly gets them wrong. When the distributor catches the error, fine. When it doesn't, expired product ships.
Phone camera + OCR + a thin verification layer against ERP receiving expectations. The operator never leaves their workflow; the system catches the discrepancies instead of the customer catching them later.
Returns live in a group chat. That's the tell.
When a driver pulls up to an account that refuses a pallet, the loop today is: driver opens the messaging app, pings the rep, rep may or may not respond in time, driver cuts the invoice, the refused cases come back to the warehouse, and leadership finds out the next day, if at all.
The VP of Operations was candid: "we're 24 hours in the rear on figuring out problems." That's not a technology failure — it's a visibility architecture failure. The group chat is fine as a driver-to-rep tool. It's the absence of a layer above it that's costing the business.
The fix: a lightweight intake that promotes each exception — return, refusal, damage, mispick — into a structured record the moment it's logged, routed to the right rep or manager, and rolled up for leadership in the command center. The driver's workflow doesn't change. Everyone above them gets hours back and decisions made in real time instead of retrospectively.
Where the handcuffs stop.
The largest supplier dictates call frequency, return obligations, and data-sharing terms. It does not dictate how the distributor sees its own business.
Where the supplier draws the line
- Call frequency by account tier (A/B/C) on the dominant supplier's brands is contractually mandated
- Out-of-date products are the distributor's cost, even when ordered by the account
- Photo intelligence the distributor collects today is housed on supplier infrastructure — not the distributor's own
- Exclusive territory assignment means no lateral expansion in-market without supplier sign-off
- 80% of volume, so any fight here is expensive and probably losing
Where Approach goes to work
- 20% of volume is unconstrained — full flexibility on scorecard, routing, rep allocation
- The wine and spirits division has no call-frequency mandate at all
- All internal operations — palletizer, receiving, inventory, exception handling — are entirely the distributor's domain
- The distributor's own data layer — rebuilt, owned, prompt-queryable — is the long game
- Smart-pallet initiative: the distributor patents and licenses upstream, flipping the dependency
"We want our own data. Because the supplier doesn't share it well enough with us."
— VP of OperationsThree phases. Twelve months.
Outcome-based. Approach is paid against measurable acceleration, not against hours billed.
Phase 01 · Stand Up
- Command Center MVP Prompt-native reporting layer live against ERP read-only
- BOL Digitization Camera capture + date cross-check at receiving
- Exception Layer Group-chat exceptions promoted into structured intake
- Data Audit Formal inventory of every system, API surface, and data flow
Phase 02 · Accelerate
- Palletizer Vision Pilot One line, mispick detection, ROI measured at 30 days in
- Field Sales App MVP In-store intelligence on phone for a pilot rep cohort
- Dynamic Load Balancer 13-pallet truck optimization moved out of the dashboard
- Proactive Delivery Comms ETA + pre-delivery verification for priority accounts
Phase 03 · Compound
- Drone / Vision Inventory Cycle counting automation across warehouse zones
- Photo Shelf Intelligence Client-owned cooler data pipeline
- Network-as-a-Service Vendor consolidation engagement
- Smart Pallet Initiative RFID co-op with distributor peers · patent posture upstream
The warehouse is already world-class.
Now the data layer gets to catch up.
Nothing in this dossier requires the client to buy a half-million-dollar robot, replace their ERP, or take on the dominant supplier. Every first-phase play sits beside the legacy stack, reads what's already there, and gives cognitive time back to the people running the business.