Why Beverage Giants Are Rethinking the 400B Yuan Restaurant Channel: From Shelf Space to User Relationships

Introduction: Money Flows from Shelves to Tables

In the first half of 2026, beverage companies' financial reports and investor communications show a sharp increase in focus on the restaurant channel. Dongpeng lists restaurant, special channels, and vending machines as key expansion scenarios; Master Kong raises direct-supply ratios in restaurants; Uni-President densifies terminal coverage; Xiang Piao Piao increases freezer placements. All push to bring products closer to consumers.

Bain's 2026 China Shopper Report shows Q1 beverage sales down 2.9% year-on-year, with ready-to-drink tea up 4.1%, bottled water up 2.9%, juice up 1.4%, carbonates up 0.7% — the overall market shrinks but sub-categories diverge. Nielsen IQ reports May full-channel beverage sales down 8.7%, offline down 10.1%. The shelf side is struggling.

Meanwhile, Hongcan Industry Research Institute estimates 2025 non-alcoholic meal beverage market at nearly 400 billion yuan (~$55B); Kantar Consumer Index shows restaurant channel penetration exceeding 45%, with average ticket price growing 5-8% annually for three years. The dining table is hot; money flows from shelves to tables.

Where the Money Goes: Three Familiar Plays

Investment in restaurant channels concentrates on three tactics: direct delivery to stores, restaurant-specific SKUs, and equipment (freezers/displays) for fixed placement. All share one premise: securing visibility.

1.1 Head Brands Simultaneously Increase Restaurant Spend

Direct supply improves speed and margin space; denser coverage raises repeat visibility; freezers and prime displays increase pick-up probability. Different actions, same target: the product and its location.

With extremely fragmented restaurant outlets (tens of thousands per city, low per-store volume), these actions are costly. Delivery, visits, merchandising, and payment terms must be negotiated store by store. Owner changes, transfers, relocations are common; active outlet lists require frequent reconciliation. Dayao's restaurant network — 31 provinces, 1,000+ distributors, 1M+ offline terminals — took years to build; high barrier, tangible returns.

1.2 Incremental Budget Shifts from Shelves to Tables

Modern retail (supermarkets) faces price wars and private-label pressure; same slotting fees yield thinning volume. Bain notes much of the 2.9% decline occurs in modern channels. Restaurant penetration at 45% means over half of restaurants don't treat beverages as routine business.

A restaurant has limited freezer space, category slots, and staff recommendation bandwidth. Early entrants fill capacity; latecomers must pay more or find new ways to displace incumbents. The same budget now buys entry tickets and better placement — but the counterparty is the store, not the end consumer.

This playbook inherits from deep distribution: maximize outlet coverage, maximize display prominence, incentivize recommenders. It solved "buyable" and "visible" in the shelf era. But as position prices rise yearly, the incremental volume no longer keeps pace with spend. The next question: the product is buyable and visible — what about the person?

Position Prices Rising, Not Controlled by Brands

2.1 A 2-Yuan Water Bottle: 0.7 Yuan to Channel

A brokerage teardown of a 2-yuan mineral water: ~0.7 yuan (35%) goes to channel (distribution, slotting, freezer placement); packaging ~25%; marketing ~12.5% (0.25 yuan). The largest cost slice is "where the goods sit," not production or brand. Distribution cost is nearly 3x marketing. Brands wanting to shift spend to product or consumer engagement find little room; budgets flow to visible placement.

Crucially, slotting/freezer fees are paid before sell-through. Brands pay upfront, then verify movement; if it stalls, the money is sunk. Stores get certainty upfront; brands bear the risk.

2.2 Channel Fees Are Rigidly Rising

Research shows strong KA systems' comprehensive fee rates approaching 25-30% of sales; prime end-cap fees rise ~10% annually. Pricing shifts from annual fixed negotiation to highest-bidder-wins. Raw materials, labor, packaging can be optimized; channel fees are rigid, upward, and pricing power sits with the retailer. Position is a time-limited lease; fail to bid, lose it.

The "traffic gets pricier, where does growth come from" question becomes concrete: position spend rises, but position returns don't grow. The gap appears elsewhere — compressed NPD budgets, cut consumer research, missed experiments.

Position Returns Thinning, Stores Pull Back First

3.1 Stores Cut Display by Revenue per Square Meter

Beverages are low-ticket, thin-margin, and freezers consume electricity. Stores rank SKUs by revenue per square meter; categories yielding higher margin per area get the space. Many stores actively reduce beverage SKUs and display area, reallocating to higher-margin items (e.g., dairy drinks). A slow-turning beverage ties up capital and electricity; the same space with a high-margin item may yield more profit than dozens of water cases. Stores don't care about brand fame; they care about monthly margin per square meter. When beverages fall in this calculus, pulling freezers is a natural, unaided decision.

3.2 Why Products Stocked in Restaurants Don't Sell

Channel visits reveal widespread underperformance: inventory turnover stretched, net margins near zero; distributor reorder rates drop due to service gaps. Qualitative but clear: goods enter, but don't move as expected.

Wrong SKU: family-pack formats placed in single-table small eateries.

Price misalignment: product priced 2x the house pour, rarely ordered.

No in-store management: freezer stock hidden behind other boxes; consumers see competitor's brand.

Each error is small, but replicated across tens of thousands of stores; manual correction costs exceed returns. One stocking's cost is shared by brand and distributor; poor sell-through sinks money in the channel — no volume, no consumer insight. Brands know how many cases shipped and sold; they don't know who drank them, why, or if they'll return.

Position is the only accounted operating object: clear counterparty, expiry, settlement slip. The consumer side has no counterparty, no settlement date, no quarterly review column. The algorithm measures position output; omni-channel fan sales aims to measure how many times the same person returns — two rulers on different dimensions, but only the first enters budgets.

What Happens at a Meal Never Reaches the Brand

4.1 Multiple Signals at One Table

A meal generates simultaneous signals: cuisine type dictates beverage function (spice-relief, mild, carbonation); party size dictates pack format; occasion (family, business, solo, gathering) dictates price band and ritual need; revisit frequency defines loyalty. Hongcan 2026 survey: 46% prioritize spice-relief; over 30% say suitable meal-pairing beverages are scarce. Demand is clear; supply lags. The demand side speaks; the brand side hears silence.

4.2 Real-Time Readable vs. Cross-Visit Knowledge

Sales reps can read table-type signals on the spot (hot pot → spice-relief; elderly/kids → mild; business banquet → premium) with 70-80% accuracy — a skill honed by experience. But this relies on individual judgment, varies by person/region, lives in heads, not databases. It answers "what's this table's type," not "who at this table is a repeat customer." The former supports a single recommendation; the latter requires identity recognition — a different capability.

Experience works because it reads table types, not identities. Same table, different rep → same type judgment. Reusable across tables, but cannot build a relationship: types repeat, identities don't.

No trace links a diner's current visit to their last: staff rotate, owners change, receipts don't link. To answer "how often they return," you need a recognizable identity stringing each appearance together. Bottle caps stay with waitstaff; bottles cleared instantly. QR code scanning in restaurants faces three-party friction: who scans, who benefits? In supermarkets, consumers self-scan at checkout; at tables, any unmotivated party breaks the chain.

Brands count cases shipped and bottles sold; they cannot count how many people bought a second time.

4.3 Hot-Pot Table vs. Family Banquet: Different Needs

Four table types show stark differences:

Hot pot/BBQ: spice-relief & carbonation first; ice-cold matters more than flavor.

Family banquet: low stimulation, shareable; large bottles & gift packs beat single-serve.

Fast food/work meal: small bottles, quick consumption, value; one-hand packaging.

Banquet/light social: non-alcoholic ritual; a normal-looking drink beats a medicinal herbal tea.

These map to spec, formulation, price-band forks. Same spice-relief sparkling drink: hot pot → large ice-cold shared bottle; fast food → small grab-and-go. Same juice: family → large premium pack; delivery → crush-proof, easy-open. Dayao's "one city one flavor" in restaurants does exactly this: adjust flavor to local diet structure, not one national recipe.

Using one product, one promo across all four tables treats four distinct needs as one. Scene resonance means product, spec, price, and rights change with the scene. Stocking solves "is it there"; scene resonance solves "does it fit this table."

4.4 Same Bottle, Different Records at Both Ends

Brand systems see a bottle as an outbound record, inventory number, receivable; only after scan does a timestamp/location open-event appear. No scan, no record — not unwillingness, no entry point.

Brand reviews only show goods flow: in, sold, stuck. Conclusions only about goods. Any people discussion ends at "where's the data," but people data requires identity first. Invisibility → no discussion → no road built. This asymmetry self-reinforces cycle after cycle.

Position renewal has a clear bill; relationship payoff waits for next visit, but most review cycles don't last that long. A yearly-renewed position vs. a multi-year compounding relationship on the same budget sheet — position wins because it delivers a verifiable slip this month.

Treating the table as another display position to buy means spending the same money again for "goods are here." "Will this person return" cannot be bought — it requires recognition at the table, then a next time.

Who Gets the Benefit, How Fast, Decides Placement

5.1 Rewards Must Match Relationship Depth

Three parties at table: owner, waitstaff, diner. Benefits today mostly flow to owner only — volume/opening-based settlement, B2B accounting. Waitstaff who recommends gets unclear, delayed payout. The actual decision-maker has the least certainty.

Waitstaff handle dozens of tables; recommendation takes seconds. Drivers: convenience, immediate perk, no complaint risk. An unknown product requiring extra explanation with month-end pooled payout won't get repeated pushes. The person holding recommendation power is often the vaguest link in incentive design.

Current volume-based rewards treat first-timers and weekly regulars equally. Omni-channel fan sales proposes tiered rewards matching relationship depth — regulars get different treatment; stores graded by sustained relationships, not single-order volume.

5.2 Scan Code: Cheapest Identity Confirmation at Table

Open-bottle scan is the lowest-cost identity capture at table, and currently the only action linking "this bottle opened by whom" back to brand. Not the only way, but shortest path, least change.

Benefits must be instant, visible: a few cents on the spot beats quarterly rebate of a few yuan. Benefits must go to the right person: rewarding only owners, not waitstaff, struggles to get codes into diners' hands.

Consumers may decline scan/authorization/opt-out — especially at a meal meant for relaxation. Low-frequency occasions unsuited for forced deep engagement. Treating every passerby as an asset to manage misunderstands relationship assets: the prerequisite is willingness to leave a thread, not unilateral recording.

Scan yields "who opened this bottle"; knowing "who this person is" remains a gap. Brands must then make post-meal interaction desirable — will they accept brand contact after the meal? Scan is entry, not destination. Many scans, zero relationships.

No One at the Table Owns the Repeat Customer

6.1 Scorecards Dictate What Reps See

Rep KPI: shipments & collection — daily focus on store shelves, stock, arrears. Distributor KPI: turnover & payment terms — focus on flow speed. Waitstaff incentive: bottles opened — push one, count one. Three scorecards, three segments; none points to "will this table's people return next month."

These metrics belong to an era of coverage-rate, outlet-count, collection-core — when the race was who could ship wider, faster, and hold more terminals. That worked, building China's vast FMCG reach. But environment changed; metrics didn't. Coverage is saturated; extra coverage yields sub-proportional volume. Scorecard subjects remain the same.

Distributors bear service/feedback cost, rewarded only on volume/scans; expecting them to understand users, discover scenes is unrealistic. For actions to persist, KPIs must shift with the operating object; otherwise, all user efforts get wiped by a monthly shipment sheet.

6.2 Who Signs for the Table's Repeat Customer?

If user identification and repurchase don't enter product definition and budget cycles, it's just another promotion: event ends, spend stops, back to square one. For continuity, someone in the org must sign — not for an event, but write it into product spec, fixed budget line, long-term KPI.

Signing means when "will this table's people return" collides with "this month's shipments," someone has authority and incentive to choose the former. These conflict on resources and time; conflict is constant. No column on the scorecard → structural choice of the latter — not attitude.

Check a company's budget: is there a line for users? Check product meetings: does anyone bring user data to propose changes? If user voice only appears in event post-mortems, not product definition, impact is limited — reports thicken, decision basis doesn't change.

Channel operations end at "goods arrived, sold"; user operations start at "this person recognized, and willing to return." That is the fundamental divide. Shifting operating object from position to person ultimately requires a specific person accountable for a specific result.

Terminals too fragmented for manual counting; channel fees rising rigidly, forcing budget reallocation; digital tools drop identity-confirmation cost to cents. Conditions are ready; waiting for someone to put "will this table's people return next month" under their own name.

Conclusion: The Next Time They Sit Here, Is It the Same Person?

Stores have clear data: cases in, bottles out, cash collected. They cannot say how much revenue comes from the same batch of people. 400 billion describes scale — how many meals, how many beverage-consuming person-times. It doesn't tell how many are repeat persons. The only real renewal certificate for a meal-pairing business is whether the same table orders it again. A contract expires and can be re-signed; a customer leaves silently — one evening they switch drinks, no one gets notified.

Omni-channel fan sales makes user-relationship assets the starting point: what enterprises accumulate shouldn't just be shelf positions, but those recognized faces who will come back.

Positions, once bought, belong to the payer; relationships don't. An end-cap bought excludes others until re-bid; a customer visiting ten times still belongs to no one — next time which store, which bottle, only they decide. Positions can be bought outright; user-relationship assets must be won repeatedly.

Relationship management has a prerequisite held by the other party: willingness to be recognized. Willing to leave the thread, free to retract it. A list ≠ an asset — list is unilateral; asset requires mutual recognition. That's why "user" comes first in "user-relationship asset": it's the user's first, then possibly the enterprise's.

A meal-pairing business's foundation isn't how many stores it covers. Store count stays on reports; whether customers return lives only in the next choice. This item never hits the balance sheet, yet determines if other line items can continue. Positions renegotiated yearly; relationships compound per visit. One resets at expiry; the other accumulates.