A friend of ours runs a small wine shop in Jerusalem, and Daniel got curious about what's actually happening behind the counter when someone buys a bottle. Not the card network side — not interchange fees and settlement between banks — but the merchant's own systems. The till, the inventory, the ledger, the reorder trigger, the tax authority. The full idealized stack that almost nobody actually runs. Daniel wants us to trace that stack from the bottom up: the one-person wine shop, the local tech retailer with three branches, the regional chain, then IKEA. At each rung, what's really deployed, what's missing, what breaks or gets quietly faked. And he flagged something strange about the bottom rung: in Israel, since June, any B2B invoice over five thousand shekels needs an allocation number from the Tax Authority's own servers before it can legally be issued. That means the wine shop's distributor invoice has to be machine-cleared in real time — but the bottle sold over the counter doesn't. The supply side gets force-digitized before the sales side. So we're tracing a continuum where regulation is pushing from the bottom, and at every rung the stack gets simplified, collapsed, or papered over. Let's start with what the full stack even looks like — and then watch it get dismantled.
The idealized merchant stack has five layers, plus a sixth that wraps around all of them. Layer one, payments processor — the thing that talks to the card networks and gets the money from the customer's bank to yours. Layer two, the till — the point of sale, the screen the cashier taps, the thing that totals the items and prints the receipt. Layer three, inventory — a record that decrements when something sells, so you know you have eleven bottles of that Cabernet left instead of twelve. Layer four, the accounting ledger — double-entry, debits and credits, the thing that tells you whether you made money today. Layer five, reordering — a trigger that says inventory dropped below some threshold, time to call the distributor. And then the sixth layer, tax authority integration, which in Israel's case is now a real-time clearance system for B2B invoices. That's the theoretical maximum. Nobody runs all of it cleanly.
Nobody runs all of it cleanly, and the interesting question isn't who runs the full stack — it's what each rung actually runs, and what it pretends to run.
Right. The stack is a theoretical maximum, not a baseline. At every rung, parts get left out, collapsed into one vendor's bundle, done by hand at month-end, or quietly faked. So we're going to climb from the bottom to the top, and at each rung we'll ask three questions. What's real? What's missing? What's being papered over?
And the Israel regulation is going to be our anchor — the one force that pushes digitization from the bottom up, but asymmetrically. It hits the supply side before the sales side, and that asymmetry shapes everything.
Let's start at the bottom, where regulation is doing something strange.
The one-person wine shop. I've been in this shop. It's on a side street off Emek Refaim, run by a guy who knows every bottle on his shelf because he put them there himself. What's actually running?
A single-vendor POS bundle. Something like a Square terminal or a local equivalent — it handles the payments processor and the till in one package. You tap your card, the terminal talks to the card network, the receipt prints. That part is real. Inventory is where it gets fuzzy. The POS app probably has an inventory feature, but whether the shop owner keeps it updated is a different question. In practice, inventory is approximate. He knows he ordered two cases of the Galil Mountain red, he knows he's been selling it, but the number in the system might say fourteen when he actually has eleven. The decrement happens when he remembers to scan the right barcode, and sometimes the barcode on the bottle doesn't match what's in the system because the distributor changed the SKU.
So the till is real, payments are real, inventory is... aspirational.
Aspirational is the right word. The accounting ledger is a spreadsheet. At the end of the month, he sits down with his bank statement and his till reports and types numbers into Excel. He's not running double-entry accounting software. He's matching deposits to sales and hoping the totals are close. Reordering is done by eye — he walks the shelves, sees the gap where the Sauvignon Blanc used to be, and calls the distributor. Tax compliance is a monthly ritual: gather the invoices, type the totals into the tax authority's online portal, file the return. The stack is real but skeletal. Payments and till exist. Inventory is approximate. Ledger is manual. Reorder is instinct. Tax is a calendar event.
And then June happened.
June 2026. The Israel Tax Authority flips the switch on real-time clearance for B2B invoices over five thousand shekels. This is a clearance system, not a reporting system. Under the old model, you issued an invoice, recorded it in your books, and reported it at the end of the month or quarter. The tax authority found out about it after the fact. Under the new model, before you can legally issue a B2B invoice over the threshold, you send the invoice data to the Tax Authority's servers, and they issue an allocation number. The allocation number goes on the invoice. No allocation number, the invoice isn't valid — the buyer can't claim the input tax credit, and the seller is in violation.
So the distributor who sells wine to the shop — that invoice has to be machine-cleared in real time before it exists legally.
The distributor's system sends the invoice to the tax authority, gets back an allocation number, and only then prints or emails the invoice to the wine shop. The wine shop's supply side is now force-digitized. Every case of wine that comes through the door has a machine-verified paper trail before it arrives. But the bottle sold over the counter? That's a consumer sale. The clearance mandate doesn't touch it. The shop owner swipes your card, the till goes beep, and the tax authority finds out about it at the end of the month when he files his return.
So you've got this asymmetry. The supply side is verified in real time by government servers. The sales side is a black box with a till on top.
And that asymmetry creates a weird incentive. The shop owner is now running a hybrid system. On the supply side, he's part of a machine-verified clearance network whether he likes it or not — his distributor forced him into it. On the sales side, he's still on a basic POS and a spreadsheet. The two halves don't talk to each other. The inventory that the distributor's invoice says he received — that's been verified by the tax authority. But the inventory in his POS, the thing that decrements when he sells a bottle? That's still approximate, still manual, still disconnected from the verified supply data.
What breaks?
Inventory drifts. The distributor's invoice says he received twelve bottles of something. The tax authority knows he received twelve. His POS says he has eight because he sold four. But he actually has seven because one broke and he forgot to write it off. Month-end comes, he does his reconciliation, and the numbers don't match. He ends up adjusting the spreadsheet to match what he thinks is true, not what any system says is true. The reorder trigger — if it exists at all — is still him looking at the shelf. The gap between the verified supply side and the unverified sales side is where the fudging lives.
The shop owner thinks he's compliant because his B2B invoices are cleared. But the sales side is unverified guesswork.
And to be clear, the tax authority knows this. They know consumer sales are a black box. The clearance mandate is a first step — it closes the biggest loophole, which was fake B2B invoices used to claim fraudulent input tax credits. The consumer side is harder to fake in aggregate because the till totals have to roughly match the inventory depletion.
Let's move up one rung. The local tech retailer with three branches. Daniel mentioned this as the sweet spot — the scale where something close to the real stack might genuinely run.
This is where it gets interesting. A three-branch retailer selling laptops and peripherals — they're big enough to need real systems but small enough that they're not running SAP. What they typically have: a proper POS system that actually decrements inventory. Not the bundled feature in a payment terminal — a real point-of-sale application, maybe something like Lightspeed or a local equivalent, running on a proper terminal at each counter. Inventory is tracked per branch. When someone buys a laptop at the downtown store, the system knows that store now has one fewer laptop. The accounting ledger is a proper package — QuickBooks or something comparable, with double-entry, chart of accounts, the works. Reordering is semi-automated: the inventory system flags items below a threshold, and the manager reviews the list and places orders with distributors.
So they have layers one through five. Payments, till, inventory, ledger, reorder. What's missing?
Integration. The POS talks to the inventory system, but the inventory system doesn't automatically talk to the accounting package. Someone exports a report from the POS at the end of the day and imports it into QuickBooks. Or worse, they re-key the totals. The reorder trigger is semi-automated but it doesn't know about returns or damaged goods until someone manually updates the inventory. And the tax authority integration is still batch — they file returns monthly or bi-monthly, same as the wine shop. The clearance mandate doesn't apply to them directly because most of their B2B invoices are under the threshold, or their suppliers handle the clearance on their side.
So the gap here isn't missing systems — it's missing integration between systems that already exist.
Every layer is present. But they're islands. The POS island, the accounting island, the reorder island. They're connected by manual processes — exports, imports, re-keying, someone walking around with a clipboard. The stack exists, but it's stitched together with human effort. And that stitching is where things break.
What's the most common breakage?
Timing. The POS knows inventory in real time. The accounting package knows inventory as of the last import, which was last night. The reorder trigger runs off the accounting package's numbers, not the POS's numbers. So if the downtown store sells out of a popular laptop model at eleven in the morning, the reorder trigger won't fire until the next morning's import — and then the order goes out, and the distributor takes two days to deliver. The store is out of stock for three days because the systems aren't talking in real time.
And the manager papering over that gap by keeping a mental list of things to reorder urgently.
Or a Post-it note on the monitor. Which works fine at three branches. It doesn't work at twenty.
Which brings us to the regional chain.
Twenty stores, maybe a couple of hundred employees, serious revenue. At this scale, you're running an ERP. Something like NetSuite or SAP Business One — an integrated system that handles inventory, accounting, and reordering in one package. The POS is still a separate system, but it syncs with the ERP nightly. Every layer exists, and most of them talk to each other. The payments processor is integrated with the POS, the POS syncs inventory to the ERP, the ERP handles accounting and triggers reorders, and the tax filing is generated from the ERP's numbers.
So what's the catch?
The catch is latency and scope. The nightly sync means inventory is always a day behind. A store could sell out of an item at nine PM, and the reorder trigger in the ERP won't fire until the sync runs at six AM. The order goes out the next morning, and the item arrives two days later. That's a three-day gap between selling the last unit and getting a new one — and during those three days, the website still shows it as in stock because the website pulls from the ERP, not the POS. Customer walks in, asks for the thing they saw online, and the store doesn't have it.
The classic "the system says we have three" problem.
The system says three because the sync hasn't run yet. The reorder trigger also has blind spots. It's watching inventory levels, but it doesn't account for returns sitting in the back room waiting to be processed, or shrinkage — theft, damage, items that walked out the door without being scanned. The accounting ledger is accurate, but it's accurate as of the last sync, not accurate right now. And the tax authority integration is still batch. The clearance mandate doesn't hit this chain directly because their B2B invoices are mostly inbound from suppliers who handle clearance on their side, and their outbound B2B invoices — if they do any wholesale — are below the threshold or processed through the ERP's tax module as batch filings.
So the gap shifts again. At the bottom rung, the gap was missing systems. At the three-branch retailer, it was missing integration. At the regional chain, it's latency. The systems exist and they're integrated, but they're not real-time.
And that latency is expensive. A three-day stockout on a popular item at twenty stores — that's real money. But closing the latency gap means moving from nightly batch syncs to real-time integration, and that's a different order of complexity. You're talking about message queues, event-driven architecture, making sure the POS can talk to the ERP in real time without bringing down the whole system if one endpoint fails. Most regional chains decide the latency is cheaper than the engineering.
Which brings us to IKEA.
IKEA. Every layer exists and talks to every other layer. POS decrements inventory in real time. The inventory system triggers reorders automatically when stock drops below threshold. The reorder flows into the supply chain management system, which calculates optimal order quantities based on lead times and demand forecasts. The accounting ledger updates continuously. Tax filings are generated automatically. The stack is complete, integrated, and mostly automated.
Mostly.
Mostly. Even at IKEA's scale, there are gaps. The tax authority integration is still batch for most transactions. IKEA files enormous tax returns, but they're filed periodically, not in real time. The real-time stack is really a series of near-real-time integrations with human checkpoints. When a reorder trigger fires, it doesn't just send a purchase order to the supplier automatically — it goes into a queue for a human buyer to review. The buyer checks the numbers, maybe adjusts the quantity, and approves the order. That human checkpoint exists because the cost of an automated ordering error at IKEA's scale is enormous. A system that accidentally orders a million Billy bookcases because of a data glitch — that's a real risk, and the human checkpoint is the insurance policy.
So even the top rung isn't fully automated. It's just that the human intervention has moved from data entry to oversight.
Right. The difference from the bottom rung isn't the presence of systems — it's the latency between them, and where the human sits in the loop. At the wine shop, the human is doing everything — counting bottles, typing spreadsheets, calling the distributor. At IKEA, the human is watching dashboards and approving exceptions. The stack didn't get more complete so much as the human moved from operator to supervisor.
And that connects back to something Daniel was getting at. The stack gets simplified not because of cost, but because of necessity. A one-person shop doesn't need real-time inventory. IKEA does. The regulation — Israel's B2B clearance — is the only force that pushes digitization from the bottom, and it's asymmetric. It hits the supply side before the sales side.
That's the core insight of this whole continuum. The full stack is a myth. Nobody runs it completely, and nobody needs to. What you run depends on what breaks if you don't run it. The wine shop can survive with approximate inventory because the cost of being wrong is a disappointed customer, not a supply chain collapse. IKEA can't survive with approximate inventory because the cost of being wrong is millions in lost sales and a warehouse full of the wrong things. The stack isn't a checklist — it's a set of answers to the question "what happens if this layer fails?"
And regulation changes the answer to that question. The Israel Tax Authority decided that fake B2B invoices were a big enough problem that the cost of not having real-time clearance was higher than the cost of forcing every business over the threshold to integrate with their servers. So they forced the supply side to digitize. But they haven't forced the sales side — yet.
Which creates this weird hybrid. The wine shop's supply side is machine-verified. The tax authority knows exactly what the shop bought, from whom, and when. But the sales side is still a black box. The shop owner is running two systems that don't talk to each other — one that's forced on him by the government, and one that he chose because it was cheap and easy. The mandate doesn't digitize the whole business. It digitizes the part the government cares about.
The shop owner ends up with the worst of both worlds. The compliance burden of the digital system, plus the manual overhead of the analog one. He has to deal with allocation numbers and machine-cleared invoices on the supply side, and then go back to counting bottles and typing spreadsheets on the sales side.
There's a knock-on effect here worth pulling out. The clearance mandate creates a paper trail on the supply side that the sales side can't match. The tax authority knows the shop bought a hundred bottles of wine this month. The shop reports selling eighty-five. The tax authority has no way to verify that number — but they can see the gap. Fifteen bottles unaccounted for. Some of that is breakage, some is the owner's personal consumption, some is shrinkage. But the gap exists, and it's visible, and it creates a tension that didn't exist before the mandate. Before, the tax authority didn't know how many bottles the shop bought either — the whole thing was self-reported. Now half of it is verified and half isn't, and the verified half makes the unverified half look suspicious by comparison.
The asymmetry doesn't just create a hybrid system. It creates a spotlight on the un-digitized half.
That spotlight is probably intentional. The tax authority knows they can't force every corner shop to run a real-time POS with tax integration — the political and technical cost is too high. But they can force the distributors, who are fewer and more sophisticated. And once the supply side is digitized, the sales side's opacity becomes a problem for the shop owner, not the tax authority. The shop owner now has an incentive to digitize his sales side voluntarily, just to make the numbers match and avoid an audit. The mandate is a lever, not a blanket.
That's a much smarter regulatory strategy than trying to mandate everything at once. Make the supply side transparent, and let the transparency pressure do the rest.
It's the regulatory equivalent of lighting one end of the room and waiting for the cockroaches to scatter toward the other end — except the other end is a digital POS system that the shop owner now suddenly wants to buy.
I want to sit with the faking for a minute. Daniel mentioned it in the prompt — parts of the stack get quietly faked. We talked about inventory drift and approximate numbers, but I think there's a deeper kind of faking that happens at the bottom rung, and I suspect Hilbert might have something to say about this.
Hilbert, you've been quiet — and I suspect you have a story about this.
Hilbert: I ran a wine import business in Tel Aviv. Two people. Late nineties. Before any of this clearance business.
What did you actually run?
Hilbert: A paper ledger and a calculator. The ledger had columns for what we bought, what we sold, what we paid, what we were owed. I'd sit down at the end of the month and add things up.
How long did month-end take you?
Hilbert: Depends on how honest I was being. If I was being honest, maybe four hours. If I was being honest, the numbers never matched. The distributor's delivery note would say fourteen cases, my ledger would say thirteen, and I'd have no idea which was right. So I'd spend another three hours on the phone with the distributor, arguing about whether an invoice was paid.
If you weren't being honest?
Hilbert: Twenty minutes. You just make the numbers match.
What does "make the numbers match" look like in practice?
Hilbert: You write down what you think you sold, not what you actually sold. The distributor says he sent twelve bottles of something. Your ledger says you have eleven. You know you sold some but you don't remember how many. So you write twelve in the inventory column and move on. The ledger looks right. The numbers add up. Nobody's going to check.
You were faking the inventory count.
Hilbert: I was making it consistent. There's a difference. The tax authority doesn't care if you have eleven bottles or twelve. They care if your purchases match your sales plus your closing inventory. If the numbers are internally consistent, they leave you alone. My job at month-end wasn't to tell the truth. It was to produce a set of numbers that didn't trigger a phone call.
The paper ledger was real, but it was also a fiction.
Hilbert: It was a fiction that was consistent with the invoices. That's all that mattered. The clearance mandate would have changed things. The distributor's invoice would have been machine-cleared before I ever saw it. The tax authority would know exactly what I bought. And then my month-end fiction — the sales side, the inventory count — would have to match a number that was already in the government's system. I couldn't just make it consistent with my own ledger anymore. I'd have to make it consistent with their servers.
Would that have saved you time or cost you time?
Hilbert: Both. It would have saved me the three hours on the phone with the distributor, because the invoice would be verified before it arrived. But it would have cost me the twenty-minute month-end, because I couldn't fudge the sales side anymore. The supply side would be locked, and I'd have to actually account for every bottle.
How many bottles went missing in an average month?
Hilbert: A few. Breakage. A bottle I opened for a customer to taste. A bottle I took home. Nothing criminal. But nothing I wanted to explain to a tax auditor either.
You'd write twelve when you had eleven, and the missing bottle disappeared into the gap between the delivery note and the ledger.
Hilbert: The gap was where I lived. The clearance mandate closes that gap on the supply side. But the sales side — that's still your word against the till. Unless they mandate real-time reporting for consumer sales too. And they will, eventually. Once they've got the supply side locked down, the sales side is next. It's just a matter of time.
The gap was where I lived. That's the whole episode in five words.
It really is. The stack isn't a set of systems — it's a set of gaps. And at every rung, someone is living in those gaps, making the numbers match, papering over the difference between what the system says and what's actually true.
Where does this leave us? The continuum we've traced — from the wine shop to IKEA — it's not really a story about technology. It's a story about what different scales of business can afford to be wrong about. The wine shop can be wrong about inventory and it costs them a disappointed customer. IKEA can't be wrong about inventory or it costs them millions. The stack at each rung is exactly as complete as the cost of being wrong demands.
Regulation changes the cost of being wrong. The Israel clearance mandate made it expensive to be wrong about B2B invoices — not in terms of business cost, but in terms of legal consequences. So the supply side digitized overnight. The sales side hasn't, because the cost of being wrong about consumer sales hasn't changed — yet.
The open question is whether the sales side gets force-digitized too. The asymmetry we've been talking about — verified supply, unverified sales — that's an unstable equilibrium. The tax authority can see the gap now. They know what the shop bought and what it claims to have sold. The gap is visible, and visible gaps tend to get closed.
If they close it — if the clearance mandate extends to B2C transactions — the wine shop's hybrid system collapses. The shop owner who was running a basic POS and a spreadsheet is now running a fully integrated stack, not because he chose to, but because the government demanded it. The full stack stops being a theoretical maximum and becomes a regulatory minimum.
The next time you buy a bottle of wine, the supply side was probably machine-verified before that bottle ever reached the shelf. The till? Maybe not. But the gap between them is getting smaller, and it's not the market closing it — it's the tax authority.
Thanks to our producer Hilbert Flumingtop, who apparently spent the late nineties living in the gap between his delivery notes and his ledger.
This has been My Weird Prompts. If you want to send us a question — about systems, stacks, gaps, or anything else — email the show at show at my weird prompts dot com.
We'll be back soon.
