Why is primary-plus-secondary harder than pure-primary or pure-e-commerce logistics?

Because there are two different decision owners (brand and distributor) with two different economic incentives, and the hand-off is contractual rather than operational. Pure-primary and pure-e-commerce have a single operator making the decisions end-to-end. FMCG requires consensus across the interface, and consensus is what kills real-time decision-making.

Can agentic AI work when part of the network is operated by partners we don't control?

Yes, and this is where the deployment pattern matters. The agent layer sits on top of the FMCG brand's TMS and reads partner data through standard integrations. Partners don't need to adopt new systems — they just expose the data the agent needs to reason over. The Heineken and Coca-Cola deployments operate this way.

Where does the on-trade / HoReCa complexity fit in?

On-trade is the hardest sub-problem within secondary because order cadence is high-variability, drop sizes are small, and SLA tolerance is very low. It is the single best pilot domain for agent-driven routing in FMCG because the status quo is so expensive to operate.

Does this apply to regional FMCG operators or only to global brands?

It applies at least as strongly to regional FMCG. Brands like Orbico, pladis, and Ripplr-pattern distributors face the same primary-secondary interface problem at smaller scale, with thinner margin to absorb inefficiency. The technology economics work down-market better in 2026 than they did in 2022.

Why FMCG primary-plus-secondary is the hardest problem in logistics — and how Heineken and Coca-Cola cracked it

By Soham Chokshi, CEO

Ask any FMCG logistics leader what their hardest operational problem is and you’ll get a clean answer: the handoff between primary distribution and secondary distribution. It is the single highest-cost, lowest-visibility, and most human-dependent interface in the FMCG supply chain. After a decade of tech investment, most operators are still solving it with Excel and a handshake. The ones who have broken out did something structurally different.

What most CXOs believe

The mainstream FMCG distribution view is that primary distribution (plant to distributor) and secondary distribution (distributor to retail) are two separate logistics problems with a contractual interface between them. Primary is the company’s cost and responsibility. Secondary is largely outsourced to distributors and partners, with the FMCG brand holding sales-out visibility but little operational control.

This model has run the industry for 30 years. It works tolerably for mature, slow-moving categories with high margin per unit. It breaks down in three places that matter now: (1) when distributor margins compress and the FMCG brand is forced to subsidize or insource, (2) when retailers demand SLA performance that requires real-time orchestration across the handoff, and (3) when on-trade / off-trade / e-commerce / modern-trade channels each demand different fulfillment patterns and the secondary layer is not built for that variability.

The default CXO response has been: add more visibility. Distributor management systems, secondary sales tracking, retailer scorecards. These help but they don’t solve the operating problem, which is that primary and secondary distribution are still two separate decision loops that don’t talk to each other in real time. The brands that have moved past this understand that the unlock is not visibility — it is a unified decision layer that treats primary + secondary as one flow.

What’s actually happening

Two large-scale FMCG deployments tell the story of what the new operating model looks like.

Heineken — 70 countries, unified settlement and dispute resolution. Heineken’s distribution network runs across company-owned breweries, bottlers, wholesalers, and HoReCa partners in 70+ countries. The financial flow across this network — freight invoicing, rate-card reconciliation, carrier and vendor disputes — used to take hundreds of humans across regional finance teams weeks to close. Vera (Shipsy’s autonomous dispute resolution agent) has autonomously resolved $25M+ in carrier and vendor disputes at Heineken — clearing what used to be a multi-quarter backlog of financial exceptions into a steady-state automated workflow. That is a direct margin unlock that flows straight to the P&L, not a dashboard improvement.

Coca-Cola bottler (Aujan / HCCB) — primary-plus-secondary unification. A Coca-Cola bottling operation deployed Shipsy’s TMS across primary distribution and into secondary routing, delivering $5M in brewery-distribution savings through a combination of load consolidation, route optimization, and visibility-driven exception reduction. The unlock was treating primary and secondary as a single optimization problem: if the primary load to a distributor is routed with knowledge of the distributor’s secondary dispatch schedule the next morning, you can consolidate deliveries, reduce touches, and hit SLA at lower cost. This is not achievable in a disconnected-decision-layer world.

The lesson from these two deployments, plus others across AB InBev-style operators, Orbico, pladis, and Ripplr-pattern distributors: the FMCG margin unlock sits at the primary-secondary interface specifically, and it requires both a unified record-and-decision layer and an agent-driven financial reconciliation capability. Either alone is insufficient.

What to do in the next 90 days

Map the handoff, not the flow. Most FMCG logistics diagrams show the full flow from plant to shelf. That is not the diagram you need. Draw the interface between primary and secondary: how does a primary load trigger a secondary plan? How is deviation communicated? How is settlement reconciled? For most FMCG brands this map reveals 5–10 manual handoffs, none of which are real-time, all of which are cost centers.

Prioritize autonomous settlement before autonomous dispatch. Counter-intuitive for most ops teams, but the Vera pattern is that autonomous dispute resolution lands faster and produces cleaner P&L impact than autonomous dispatch. Deploy the settlement agent first. The financial unlock funds the broader transformation, and the internal credibility of automated decision-making builds on a domain where the rules are deterministic and the outcomes are measurable in cash.

Deploy a unified TMS across primary and secondary — even if secondary is partner-operated. The partner operating model does not prevent unified visibility and unified routing intelligence. Shipsy deployments across Heineken, Coca-Cola bottlers, and similar operators run the same TMS layer into distributor systems so the primary planner knows the secondary schedule and the secondary partner has visibility into upstream delays. This is a 6–9 month implementation that produces 2–4 points of distribution cost savings by year two.

Run an agent-driven load consolidation pilot on your highest-frequency lanes. Where the same primary lane feeds the same distributor multiple times per week, consolidation-by-agent produces immediate cost savings without disrupting the operating model. This is a low-political-risk entry point for agent deployment and establishes the decision-layer credibility you need for bigger initiatives.

Rebuild your distributor scorecard around SLA outcomes, not activity metrics. Most FMCG distributor scorecards measure activity — orders processed, delivery attempts, returns. The metric that matters is outcome: did the retailer receive what they ordered, on time, in full? An agent-driven OTIF (on-time-in-full) measurement layer across the secondary network is how you get there.

If you are a regional GM or distribution head — carve out a pilot country. A single-country pilot with unified primary-secondary tech, autonomous settlement, and agent-driven load consolidation produces the internal reference case you need to scale globally. Heineken’s deployment followed exactly this pattern.

Why this matters now

FMCG margin is being squeezed from both ends — input costs rising, retailer power concentrating — and the traditional response (SKU rationalization, plant consolidation, promo cuts) is largely exhausted. Distribution is the last major cost bucket with meaningful unautomated decision layers. The 2–5 point operating-margin unlock available from agentic distribution automation is the largest uncontested margin pool in FMCG for the next 24 months. The brands that move in 2026–2027 will compound it. The ones who wait will not have the option by 2028, because their retail partners will demand the operating model.