Ask a regional sales manager at an Indonesian FMCG company how sell-out looked yesterday, and the honest answer is usually "give me until Thursday." The data isn't missing. It sits in a dozen places that don't talk to each other, and someone still has to stitch the recap together by hand before anyone can read it.
That gap between "the data exists" and "someone can answer the question" is the real bottleneck in FMCG distribution. Sell-out, coverage, days of supply, promo payback: every one of those numbers is already being recorded somewhere. What's missing is a place where they line up on the same morning, so the questions that actually run a commercial team stop taking until Thursday to answer.
The questions an FMCG team asks every day
Commercial and supply chain teams don't ask exotic questions. They ask the same ones, repeatedly:
- Which distributors are behind on this month's target, and by how much?
- Which SKUs are stocked out or close to it, in which territories?
- What did general trade sell yesterday versus modern trade?
- How many days of supply are left on the top movers, distributor by distributor?
- Which must-stock SKUs are missing from outlets that should be carrying them?
- Is the current promo actually paying back the trade spend behind it?
- What's coming back as returns or sitting close to expiry, and where?
- Which outlets haven't been visited this cycle, and where is margin slipping?
None of these require sophisticated modeling. What they require is current numbers pulled across systems that were never built to talk to each other: a distributor's own recap sheet, a retailer's POS feed, and the company's ERP. Today, answering any one of them well usually means a planner opening three tools and a spreadsheet, then waiting for someone else's data to land.
Why does distributor data arrive late and inconsistent?
Distributors are independent businesses. Most run whatever system they've always run, sometimes a proper distributor management system, sometimes a spreadsheet, sometimes a notebook that gets typed up at week's end. Head office doesn't own that system, so it can't simply query it. It waits for a report.
That report is where consistency breaks down. One distributor's "sell-out" might mean shipments to sub-distributors; another's might mean confirmed retail sales. Units get recorded in cases by one team and pieces by another. By the time a national recap is assembled, someone has quietly normalized a dozen small definitional differences, and nobody upstream saw it happen.
The result is data that is both late and untrustworthy in ways that stay hidden until a number gets challenged in a meeting. Asking distributors to try harder won't fix it, because the cause is structural: the data was never built to be queried, only to be reported.
General trade and modern trade run on different clocks
General trade (traditional warungs and small outlets) and modern trade (chain retailers) don't just differ in format. They differ in how fast truth reaches head office.
| General trade | Modern trade | |
|---|---|---|
| Primary data source | Field sales reps, distributor recaps | Retailer POS / EDI feeds |
| Typical latency | Days, often manual entry | Often near-real-time, but siloed per retailer |
| Main visibility risk | Under-reporting, inconsistent SKU coding | Feed format changes, delayed reconciliation |
| What "sold" means | Frequently means shipped to outlet, not sold through | Usually genuine sell-through at the till |
Most tooling picks a side. Field force automation tools are built for GT and barely touch retailer feeds. Retail analytics platforms are built for MT and have no concept of a distributor recap. A commercial team that sells through both channels ends up holding two half-pictures and reconciling them in a spreadsheet, which is exactly the kind of work that goes stale the moment it's finished.
The numbers hiding behind sell-out
Sell-out gets the attention because it's what the target is written in. But an FMCG commercial team lives or dies on a few numbers that sit underneath it, and each one breaks the same way: spread across systems that never reconcile.
Days of supply is the clearest. Knowing a distributor sold 400 cases last week means little without knowing how many days of cover are left behind that number. The first figure lives in the recap, the second in the distributor's own stock file, and nobody joins them until a fast mover has already run dry. Must-stock compliance has the same shape. Head office sets a list of SKUs every outlet in a channel should carry, but whether a given warung actually has them on the shelf is a field-visit fact that rarely meets the target sheet.
Trade spend is where the money hides. A promo is funded against expected incremental volume, yet judging whether it paid back means putting the spend, the baseline, and the actual sell-through in one view. By the time that reconciliation is done by hand, the budget for the next cycle is already committed. Returns and near-expiry stock close the loop: product coming back, or sitting a few weeks from its date in a distributor warehouse, is real margin quietly leaking, and it usually surfaces as a write-off long after anyone could have moved it.
None of these are new metrics to invent. They're already recorded, just never in the same place at the same time. It's the same connective gap that runs through multi-store retail operations and the wider supply chain: the number exists, the join doesn't.
What does a connected commercial agent actually do?
This is where a department-scoped AI agent earns its place, not by predicting demand or replacing a distributor management system, but by sitting across the systems that already exist and answering the questions above the moment they're asked.
With live, governed access to distributor data, POS feeds, and ERP, an agent scoped to commercial or supply chain can:
- Answer "which distributors are behind target" against this morning's numbers, not last week's recap.
- Flag a stock-out risk as soon as inventory crosses a threshold, instead of after a lost sales figure shows up in a monthly review.
- Show days of supply and promo payback side by side, so a shrinking margin gets caught while there's still a cycle left to change it.
- Reconcile GT and MT sell-through into one view without anyone manually merging spreadsheets.
- Route the answer to whoever asked, a territory manager, a regional head, or the finance team checking receivables, with only the scope of data that person is entitled to see.
None of this requires the agent to be smarter than the people using it. It requires the agent to be connected to the right systems and to understand what the numbers mean: canonical definitions of sell-out, sell-through, and coverage that don't shift depending on who compiled the report.
A day, illustrated
Picture a national sales manager on a Monday morning, purely as an illustration of how the pieces fit together, not a claim about any real deployment. Instead of waiting for the weekly distributor recap, she opens WhatsApp and asks which territories are trending below target this month. The answer comes back with distributor names, the gap versus target, and which SKUs are driving the shortfall, pulled from this morning's distributor and POS data rather than last month's file.
A field rep in a different city asks the same kind of question from a warung, gets a stock-out alert for a fast-moving SKU, and flags it before the outlet loses a full day of sales. Neither of them opened a dashboard. Both got an answer scoped to what they're allowed to see.
Why does this belong in an agent, not another dashboard?
Dashboards are still useful for the deep monthly review. But the daily rhythm of FMCG distribution, a stock question at 8am, a coverage question at noon, a target-gap question before a Friday call, is exactly the shape of work department-scoped agents structured like an org chart are built to absorb. A commercial agent that reports into the same hierarchy as the commercial team, with the same access boundaries, turns "wait until Thursday" into "ask right now."
A field rep covering forty outlets on a motorbike is never going to open a BI tool between stops. They already run the day through chat, so the interface that gets adopted is the one already on the phone in their pocket, and a stock-out answer that lands there beats a dashboard nobody in the field ever logs into.
Where Nalar fits
Nalar is an AI intelligence layer built for Indonesian enterprises, and FMCG is one of six industry archetypes it ships with. It connects distributor, POS, and ERP data under live, governed access, models distribution the way a commercial team already reasons about it (by distributor, territory, channel, and SKU), and answers through department-scoped agents in chat, on a dashboard, and over WhatsApp.
If you'd rather not take that on faith, the BARI readiness diagnostic will look at your distributor and POS data and say plainly whether it's in shape to be connected this way. Either way, the interactive demo runs the whole thing on a realistic mock FMCG enterprise: the fastest way to see what "ask right now" actually looks like before talking to anyone.
Frequently asked questions
- Why is FMCG distributor data always late?
- Most distributors don't run the same systems as head office and report on their own cadence — often a manual recap sent by email or WhatsApp at the end of the week. By the time it's compiled, aggregated, and reviewed, the numbers describe last week's business, not this week's.
- Can AI fix stock-outs directly?
- Not by itself — stock-outs are a supply chain and replenishment problem. What AI can do is surface a stock-out or a near-stock-out the moment the data shows it, instead of it showing up as a lost sales figure weeks later.
- Does this replace our distributor management system or ERP?
- No. An intelligence layer connects to the systems you already run (DMS, ERP, POS feeds) and answers questions across them. It's not a replacement for any one system; it's what lets someone ask a question that spans all three.
- How does this apply differently to general trade versus modern trade?
- GT visibility depends on field reporting and distributor recaps, so the agent needs to work with irregular, sometimes manually entered data. MT visibility comes from retailer POS feeds, which are more structured but arrive through a different pipe entirely. A useful agent has to reconcile both, not just pick one.