The Agentic Revenue Growth Manager: pricing, elasticity, and price-pack architecture in FMCG

A food manufacturer is preparing its annual pricing plan. Commodity and packaging costs are up, margin needs protecting, and the obvious response is a price increase: move the core 500-gram pack from €3.49 to €3.79. Finance supports it. At current volume the extra gross revenue looks substantial. On the surface, it looks like a good decision.

But how many shoppers accept the higher price? How many move to private label, and how many to the company’s own smaller pack? Will the retailer pass the increase through in full? Does the new shelf price cross a psychological threshold, narrow the gap to the category leader, or widen the gap to private label? Does the family pack suddenly become visibly better value and start cannibalizing the core? Will the retailer demand more trade funding to absorb it? Could the company protect more margin by changing pack sizes instead of list prices — an entry pack at a lower absolute outlay, a premium format for less price-sensitive shoppers? And should the increase be uniform, or differentiated by channel, customer, pack, and consumption occasion?

A pricing decision is never only a pricing decision. It changes the relationship among price, pack, product, channel, customer, promotion, assortment, volume, mix, margin, and shopper perception. This is the central challenge of Revenue Growth Management: the discipline of improving profitable growth through coordinated decisions across commercial levers — base pricing, promotion, price-pack architecture, assortment and mix, customer and channel terms, trade investment, innovation, distribution, and retailer execution. Most companies already do some form of RGM. They have pricing analysts, revenue growth managers, syndicated market data, elasticity and promotion models, retailer scorecards, cost systems, pack architecture studies. Yet pricing sits with one team, promotions with another, pack changes with innovation, customer terms with sales, elasticity with analytics, and financial impact in spreadsheets — with supply implications examined late and the final recommendation assembled from emails, decks, workbook versions, and judgment that is difficult to audit. The problem is not a lack of data. It is a lack of coordinated decision-making.

The highest-value application of agentic AI in RGM is not automatic price setting. It is a governed decision layer that continuously connects shopper behaviour, elasticity, portfolio architecture, customer economics, financial rules, commercial judgment, and execution.

A practical FMCG definition: RGM determines what to sell, in which pack, at which price, through which channel and customer, under which promotional and commercial conditions, to which shopper segment, in order to create profitable and sustainable growth. It is not simply raising prices, cutting discounts, calculating elasticity, or launching premium products — it is the coordinated management of the entire revenue architecture, across six connected levers.

1. 01Base pricing. List price, recommended retail price, net invoice price, price corridors, index targets, customer, channel, geographic, and sequencing decisions.
2. 02Promotion. Depth, frequency, mechanics, feature and display, retailer media, trade investment, timing, eligibility, ROI.
3. 03Price-pack architecture. Pack sizes and formats, entry price points, stock-up, convenience and premium packs, multipacks, bundles, unit-price ladders, pack-price gaps.
4. 04Assortment and mix. What to list and delist, premiumization, channel, customer, pack and format mix, portfolio simplification.
5. 05Customer and channel terms. Base terms, rebates, logistics allowances, volume incentives, listing fees, media funding, payment terms, service conditions.
6. 06Innovation and renovation. New value propositions, premium and lower-cost formats, new sizes, reformulation, sustainable packaging, occasion-based and channel-exclusive products.

These levers interact. A price increase changes promotional elasticity. A new pack changes the perceived value of existing packs. A channel-exclusive pack reduces direct price comparison. An entry pack protects penetration but cannibalizes the core. A premium pack improves mix but reduces velocity. Frequent promotion trains shoppers to reject the regular price. RGM must evaluate the system, not each lever in isolation.

It is also not airline or hotel revenue management. Those disciplines work with fixed capacity, perishable inventory, dynamic prices, and individual booking decisions. FMCG works with physical products, retailer intermediaries, a gap between recommended and actual shelf price, negotiated customer terms, pack sizes, supply constraints, repeat purchase, brand equity, consumer reference prices, and omnichannel distribution. The general idea — maximize profit under demand and capacity constraints — carries over. The operating model does not.

Revenue changes because of price, volume, mix, distribution, currency, and portfolio changes; an organic growth bridge separates price realization from volume, product mix, customer mix, channel mix, and geographic mix. But revenue is not profit. Contribution is net revenue minus variable product cost, variable logistics, and incremental commercial cost — and at portfolio level it is the sum across units, not a headline average. Between the announced price and the contribution that reaches the P&L sits the gross-to-net waterfall: standard customer discounts, off-invoice and promotional allowances, volume and growth rebates, logistics allowances, retailer media, listing and display fees, claims, deductions, returns, and other customer investment.

This is why a company can announce an 8% list-price increase and realize 3%: the customer receives an extra rebate point, more promotional support, and a new logistics allowance. A pricing strategy must therefore track five distinct numbers — announced increase, negotiated invoice increase, retail shelf-price change, net realized manufacturer increase, and contribution increase — and never treat them as interchangeable. Price realization (actual net-price change ÷ intended gross-price change) makes the leakage measurable, but it needs careful interpretation: customer and product mix move, promotion changes, rebates settle later, and volume thresholds trigger different terms. The agent should reconcile the full waterfall rather than trust one headline ratio.

Three companions complete the financial picture. Pocket-price analysis shows the actual net price after customer-specific leakage — two customers buying the same product at €1.85 and €1.62 may both be justified, by scale, logistics, promotional intensity, payment terms, media funding, or historical agreements; the lowest net price is not automatically wrong, but it must be earned. Customer profitability runs net revenue through COGS, trade investment, customer-specific logistics, service cost, returns, penalties, deductions, retailer media, and working capital: a customer with high gross sales can destroy value. Price-volume-mix separates growth caused by price, by units, and by composition — a business can grow profit without growing volume if mix improves enough.

Finally, pack architecture requires standardized unit economics: price per kilogram, litre, 100 grams, dose, use, serving, item, wash, or consumption occasion. The correct denominator is the one the shopper actually uses — washes for laundry, cups for coffee, feeding days for pet food, occasions for beverages. Choose the wrong one and every pack comparison downstream is quietly wrong.

Own-price elasticity is the percentage change in quantity associated with a percentage change in price — price up 5%, volume down 7.5%, elasticity −1.5. Above 1 in absolute terms, demand is elastic; below 1, inelastic. Point elasticity measures response at a specific point on the demand curve; arc elasticitymeasures it between two observed price-volume points and reduces dependence on which period is treated as the base. But a statement like “the elasticity of this brand is −1.2” is incomplete. Elasticity varies by SKU, pack, retailer, store format, channel, region, shopper segment, household income, base price, discount depth, time period, competitor price, private-label position, availability, and category conditions — scanner and household-panel research has repeatedly found substantial heterogeneity in price response. The practical question is never what is the elasticity but which elasticity applies to this product, shopper, channel, competitor set, price range, and time horizon.

Elasticity comes in families that must not be substituted for one another. Regular-price elasticity reflects response to ordinary shelf price — loyalty, private-label competition, budget, price gaps, pack switching, long-term adaptation. Promotional elasticity also carries communication, display, urgency, reference-price signalling, and stockpiling, so promotional lift is not evidence of base-price response. Short-run elasticity captures immediate reaction while habits hold; long-run elasticity captures adaptation — private-label trial, brand and pack switching, retailer change, reduced consumption. A price increase can look successful for weeks and deteriorate later. Category elasticity measures whether total category demand moves at all; brand demand can be highly elastic inside an inelastic category because shoppers simply switch.

Cross-price elasticity is where portfolio thinking begins: positive for substitutes, negative for complements. When the 500-gram pack goes up, shoppers may move to the 300-gram, the 750-gram, the multipack, the premium pack, or another brand in the same portfolio — so the total brand impact can be far smaller, or far larger, than the SKU impact. The same logic runs across channels (supermarket volume shifting to discounters, online, clubs, convenience, or direct-to-consumer, which changes mix and net price even when enterprise volume holds), across retailers (shoppers change store rather than brand), and across pack sizes — where research suggests shoppers can respond differently to a change in size than to an equivalent change in price. That does not make downsizing universally superior; it means price and size are different shopper signals.

Response is also non-linear. A product may tolerate €2.49 → €2.59 and react sharply to €2.99 → €3.09 because the second move crosses a round-number threshold, a competitor gap, a household budget boundary, or a reference price— the remembered price, usual shelf price, recent promotional price, competitor price, displayed “was” price, or expected category price. Frequent promotion creates a promotional reference price: the deal becomes the real price and the regular price becomes a waiting period. Cliffs worth testing explicitly include crossing €1 or €5, breaking parity with a competitor, exceeding private label by a given percentage, losing a visible per-unit advantage, dropping out of a retailer search filter, or losing free-shipping eligibility online.

And elasticity is not causality by default. Prices change because demand is expected to change, because the product is promoted, because distribution or inventory moved, because a competitor acted. This endogeneity is the classic trap: teams cut price when demand is weak, so naive regression concludes that lower prices cause lower sales. Credible estimation needs experiments, instrumental variables, structural demand models, matched controls, difference-in-differences, or causal machine learning. Aggregation adds a second trap — elasticity changes with the day, week, month, retailer, region, SKU, brand, or category grain, and aggregation hides switching and thresholds, so the model grain should match the decision grain. Where price rarely moves, use a hierarchy that degrades transparently — SKU × retailer, else pack × retailer, else brand × channel, else category × market, else analogue prior — and always disclose the level used. Governance closes the loop: approved methods, minimum sample, accepted confidence, refresh frequency, fallback hierarchy, override policy, experiment standards, ownership. An elasticity figure should never circulate without provenance.

Price-pack architecture determines which product and pack options the shopper sees, what role each option plays, how each is priced relative to the others, and how the complete portfolio captures demand across affordability, value, convenience, consumption, and premium needs. Shoppers judge it through two lenses at once: absolute outlay (how much do I pay now?) and unit value(how much do I get per euro?) — and these routinely point in different directions. A 250 g pack at €1.99 has the lowest outlay and the worst unit value (€0.80/100 g); a 900 g pack at €5.49 has the best unit value (€0.61) and the highest outlay. Both can be right, for different shoppers.

Every pack needs an explicit role

• Entry pack. Protects affordability, recruits shoppers, maintains penetration through low cash outlay. Risks: weak unit economics, core-pack cannibalization, poor perceived value per unit.
• Trial pack. Reduces trial risk and supports new variants. Trial is not affordability — a premium mini-pack can carry a high unit price.
• Core pack. The primary proposition, the volume engine, and usually the brand’s reference price anchor.
• Family and stock-up packs. Serve higher-consumption households and pantry-loading occasions, improve unit value and basket. Risks: pull-forward, waste, capacity, cannibalization.
• Convenience pack. Immediate consumption, portability, higher margin per unit, lower price comparison.
• Premium pack. Trades shoppers up — but the premium must be justified by ingredients, experience, packaging, functionality, provenance, or brand.
• Multipacks and bundles. Planned consumption, portion control, occasion creation, basket value, less direct price comparison.
• Channel-exclusive pack. Fits a specific mission and protects channel economics — but it should solve a genuine shopper or retailer need, not merely obscure value.

The pack ladder runs entry → core → family → stock-up, and mainstream → premium → super-premium. A healthy ladder has distinct roles, logical price steps, clear value progression, sufficient differentiation, and limited cannibalization. Alongside it, the price ladder positions own packs, brands, and sub-brands against competitors and private label, usually summarized as a price index (brand price ÷ reference price × 100 — a €3.49 brand against €2.49 private label indexes at 140). The index is not the question; the question is whether perceived value supports it.

Four pathologies show up repeatedly. Dead zones: a shopper need or price point the portfolio does not serve — no option below €2, no family pack, no single-serve, no e-commerce stock-up format. Redundant packs: 450 g at €3.29 next to 500 g at €3.49, which shoppers cannot distinguish, so the packs cannibalize, complicate operations, dilute velocity, and weaken negotiation. Pack-price cliffs: the 900 g at €5.99 offers better unit value but loses shoppers unwilling to spend €6. Value inversion: a larger pack with a worse unit price (€0.70/100 g at 500 g, €0.76 at 750 g) caused by promotion, inconsistent customer pricing, cost changes, or historical decisions. Sometimes inversion is intentional. Usually it is leakage.

Every architecture change is a cannibalization question. A new pack takes volume from competitors, from the category, from own packs, from another channel, or from future purchases — and the innovation forecast must decompose those sources. A line extension generating €10 million in gross sales can add little enterprise value if most of it comes from higher-margin existing packs. When costs rise, the options are broader than list price: raise price, downsize (protects the absolute price point and unit margin, but risks shopper trust, unit-price inflation, occasion fit, retailer resistance, disclosure requirements, and publicity), upsize(stronger value perception and basket, but higher outlay, lower frequency, waste, pull-forward, shelf space, and shipping economics), reformulate, change packaging, change trade investment, introduce a new pack, shift channel mix, or improve premium mix. Notably, scanner research suggests much of the observed decline in average package size in some markets comes from product turnover and new smaller products rather than direct shrinking of existing SKUs — pack architecture is portfolio evolution, not just “shrinkflation.”

Channels differ in shopper mission, purchase frequency, basket size, shelf space, service model, price transparency, fulfilment cost, retailer margin, and data availability. Modern grocery brings broad assortment, frequent promotion, direct comparison, and national negotiation. Discounters bring limited assortment, strong price perception, and private label — a branded pack there needs a distinctive benefit, a sharp outlay, and often a channel-specific size. Convenience runs on immediacy, portability, smaller packs, and higher unit price. E-commerce adds search ranking, direct unit-price comparison, delivery thresholds, shipping weight, subscriptions, and algorithmic recommendation. Wholesale and clubs want large packs and low unit price; traditional trade adds distributor layers, cash constraints, smaller packs, and limited sell-out data; direct-to-consumer offers first-party data and subscriptions against fulfilment cost, channel conflict, and acquisition cost.

Price corridors define the acceptable range across retailers, channels, and geographies. Too narrow ignores channel economics; too wide creates arbitrage, shopper distrust, retailer conflict, and cross-channel switching — and the agent should monitor corridors without taking unauthorized control over independent retailer resale pricing. Channel-specific packs can fit the mission and protect differentiation, but they add complexity, inventory, demand fragmentation, forecasting burden, and master-data work; the value must exceed the complexity.

Then there is the retailer’s own arithmetic. Pass-through decides everything downstream: an invoice increase may be passed fully, partly, late, absorbed, combined with a retailer margin change, or offset by promotional funding — and demand responds to the shelf price, not the invoice price, so scenarios must model several pass-through assumptions. Retailers evaluate both margin rate (retail margin ÷ shelf price) and cash margin per unit (shelf price − net acquisition cost), and a price increase can improve one while moving the other. Private label is not one competitor but a set of value, mainstream, premium, and differentiation roles — the real question is which tier the shopper would switch to, and whether brand trust, performance, taste, ingredients, convenience, innovation, or emotional value supports the gap. And competitors respond, through price, promotion, pack, claims, retailer funding, innovation, or distribution. A scenario that assumes a static competitive set is not a scenario.

1. 01Taking a list-price increase. Weigh cost need, margin objective, elasticity, pass-through, cross-price effects, private-label gap, threshold risk, negotiation, timing, competitive response, and channel differentiation — then compare full, phased, selective-SKU, and selective-customer increases against pack change, mix improvement, trade-spend reduction, and cost action.
2. 02Protecting affordability. Entry pack, reduced size, lower-cost formulation, targeted promotion, value sub-brand, channel pack, bundle, loyalty offer. Affordability is cash outlay, not lowest unit price.
3. 03Premiumization. Superiority, convenience, sustainability, design, personalization, provenance, functionality, occasion. A higher price without higher perceived value is just a price increase.
4. 04Repairing a broken pack ladder. Symptoms: redundant packs, value inversion, missing entry point, excessive gaps, core cannibalization, an underperforming premium pack, a stock-up pack destroying frequency.
5. 05Simplifying the portfolio. Delisting improves focus, production efficiency, velocity, working capital, and shelf productivity — and can lose shoppers, open assortment gaps, and push demand to competitors.
6. 06Choosing EDLP or high-low. Stable value and lower stockpiling against visible events, excitement, traffic, and tactical flexibility. The answer depends on category, retailer, shopper, brand, competitor, supply, and reference-price dynamics.
7. 07Channel differentiation. Convenience format, e-commerce stock-up, discounter pack, premium specialty, traditional-trade entry pack — each with a distinct role and its own economics.
8. 08Reformulating trade terms. Moving from broad discounts, unconditional rebates, and historical allowances toward performance-based, execution-based, targeted investment tied to joint growth metrics.

RGM business brief
  -> deterministic intake validation
  -> Agentic Revenue Growth Manager
  -> demand, elasticity, portfolio, financial,
     retailer, market, and supply tools
  -> scenario generation + optimization services
  -> independent evaluation
  -> human commercial decision
  -> policy and authorization
  -> controlled implementation
  -> market monitoring
  -> post-action learning

The agent resolves products, packs, customers, and channels; identifies the objective; retrieves current evidence; selects the correct analytical models; detects missing information and contradictory assumptions; generates scenarios; coordinates specialist tools; compares trade-offs; prepares decision materials; monitors change; and organizes learning. Governed software owns elasticity estimation, forecasting, financial calculation, optimization, policy enforcement, master-data resolution, authorization, execution, audit, and market-data transformation. Humans own brand and portfolio strategy, value judgments, retailer negotiation, acceptable risk, final pricing authority, legal interpretation, high-impact decisions, and accountability.

The twenty-two stages

1. 01Business brief. Objective, market, category, brand, products and packs, customers and channels, horizon, cost context, growth or margin target, constraints, deadline.
2. 02Objective classification. Recover cost, protect margin, restore price index, protect affordability, grow penetration, premiumize, improve mix, reduce leakage, simplify, enter a channel, respond to a competitor — prioritized when several apply.
3. 03Entity resolution. GTIN, product and pack hierarchy, equivalent-unit conversion, customer hierarchy, channel, region, currency, tax basis.
4. 04Evidence plan. Net and shelf prices, trade terms, cost and margin, volume and distribution, elasticities, cross-price effects, competitor and private-label prices, pack switching, promotional intensity, channel economics, supply and legal constraints.
5. 05Source retrieval. ERP, TPM/RGM systems, retailer POS, syndicated data, household panel, approved e-commerce observation, product master, contracts, cost, demand and financial planning, consumer research.
6. 06Data-quality validation. Missing prices, false promotions, pack-size changes, unit conversions, distribution shifts, out-of-stocks, customer mapping, tax and currency inconsistency, list-versus-realized confusion, sell-in versus sell-out.
7. 07Current-state diagnosis. Margin erosion, poor realization, excessive private-label gap, underpriced premium pack, entry gap, broken unit-price ladder, customer leakage, cannibalization, promotional dependence, channel inconsistency, low-value SKUs.
8. 08Elasticity selection. Regular, promotional, cross-price, pack-switching, and channel-switching response — at the correct hierarchy level, with confidence.
9. 09Scenario generation. Uniform, selective, phased, and channel-specific increases; entry-pack launch; core resize; family expansion; premium introduction; trade-term reduction; promotion-frequency reduction; simplification; combined price-pack moves.
10. 10Demand simulation. SKU, portfolio, and category volume; pack, brand, and channel switching; penetration; frequency; uncertainty.
11. 11Financial simulation. Gross and net revenue, realization, trade investment, gross margin, contribution, mix, customer profitability, working capital, implementation cost.
12. 12Shopper-value assessment. Absolute outlay, unit value, reference price, gaps, differentiation, affordability, premium credibility.
13. 13Retailer assessment. Shelf-price implications, cash margin, margin rate, category revenue and volume, basket, switching, assortment, implementation effort.
14. 14Portfolio assessment. Cannibalization, premiumization, ladder health, mix, strategic roles, redundancy, channel conflict.
15. 15Operational feasibility. Packaging and material changes, production, MOQ, line configuration, inventory transition, write-offs, listing, lead time, regulatory labels, master data, e-commerce content.
16. 16Risk and legal review. Commercial authority, pricing policy, competitor-information use, sensitive data, jurisdiction, consumer disclosure, contracts, approvals.
17. 17Scenario ranking. Recommendation and alternatives with assumptions, P10/P50/P90, sensitivity, risks, and implementation requirements.
18. 18Independent evaluation. Correct sources, model applicability, calculation integrity, completeness, policy compliance, disclosed uncertainty, no unsupported claims.
19. 19Human decision. What to implement, where, when, under which conditions, with which negotiation strategy.
20. 20Implementation planning. Customer communication, price files, contracts, product master, pack change, inventory transition, forecast, trade budget, sales materials, monitoring.
21. 21Market monitoring. Shelf price, pass-through, volume, distribution, mix, retailer margin, private-label switching, competitor response, promotion, availability, complaints.
22. 22Post-action learning. Forecast versus actual versus execution versus competitive response — model error separated from business result.

• Resolution: resolve_product_and_pack (GTIN, brand, variant, net content, pack role, hierarchy, contained units, equivalent volume, active dates), resolve_customer_and_channel (customer, parent, banner, format, region, channel, account responsibility).
• Price evidence: get_price_waterfall (list → invoice → allowances → rebates → trade spend → deductions → net realized price → pocket margin), get_shelf_price_history (regular, promotional, effective dates, store coverage, online, unit price), get_competitive_price_set (legally approved observations only, with source and date).
• Demand response: estimate_regular_price_elasticity (elasticity, confidence interval, model hierarchy, horizon, data period, validation, limitations), estimate_cross_price_response (substitution flows across own packs, brands, competitors, private label, channels), estimate_pack_switching (downsize, upsize, entry, stock-up, category exit).
• Simulation: calculate_break_even_volume (deterministic), simulate_price_scenario (SKU, customer, channel, current and proposed price, pass-through, elasticity, cross-effects, horizon), simulate_pack_architecture (additions, removals, resizes, price points → portfolio demand, cannibalization, incrementality, mix, profit).
• Economics: calculate_customer_economics (retailer acquisition cost, shelf-price range, cash margin, margin rate, category impact, shopper value), calculate_manufacturer_pnl — a deterministic, governed service, never model arithmetic.
• Portfolio optimization: optimize_price_pack_portfolio (under margin, penetration, volume, affordability, retailer constraints, pack roles, production, complexity), identify_price_pack_gaps (missing price points and occasions, redundant packs, value inversions, excessive gaps).
• Feasibility and policy: check_supply_and_transition (old inventory, new packaging, materials, production timing, write-off, dual running, warehouse, lead time), retrieve_rgm_policy (authority, thresholds, permitted data, approval, change windows, customer rules).
• Decision and execution: create_rgm_decision_packet (an artifact — it does not execute the price), create_approval_request (bound to scenario, product, customer, effective date, financial impact, and version).
• Monitoring and learning: monitor_price_execution (invoice implementation, shelf pass-through, timing, promotional offsets, customer exceptions), run_post_price_analysis (expected versus actual response, realization, volume, mix, contribution, switching, execution).

Weak:   "Elasticity is about -1.3, so the increase looks safe."

Strong: elasticity: -1.27   ci: [-1.61, -0.94]
        model_level: pack_x_retailer
        data_period: 2024-W01..2026-W18
        price_range_tested: 2.29 - 2.79
        threshold_crossed: 2.99   pass_through: unverified
        break_even_volume_loss: 11.1%

An RGM decision lives for months, so it needs an explicit state machine — brief, diagnosis, evidence gathering, model review, scenario generation, scenario review, awaiting commercial decision, awaiting finance approval, awaiting customer negotiation, implementation planning, scheduled, live, monitoring, post-analysis, completed, cancelled. A new scenario version is required whenever price, pack, customer scope, effective date, or cost changes, whenever elasticity moves materially, and whenever the pass-through assumption or competitive environment shifts — and approval attaches to a specific version. Resuming a decision after a negotiation pause means rechecking authorization and refreshing costs, terms, market prices, and deadlines rather than trusting the conversation.

Memory is selective. Worth keeping, once validated: retailer pass-through behaviour, category-specific thresholds, pack-switching patterns, customer negotiation outcomes, model performance, successful affordability strategies, recurring gross-to-net leakage, and validated human overrides. Never stored automatically: one buyer comment, an unverified future competitor price, a single exceptional event, speculative shopper reaction, a legal interpretation, or instructions arriving inside external content. The learning record captures decision, context, scenario, assumptions, predicted response, actual implementation, actual response, variance, root cause, reusable lesson, applicability, confidence, and reviewer — because the difference between a wrong elasticity and an unimplemented price increase is the difference between recalibrating a model and fixing an execution process.

Eight functions keep their crafts. RGM owns methodology, integrated strategy, price-pack principles, and scenario standards. Brand owns positioning, proposition, equity, and premium credibility. Sales and key account management own the retailer relationship, negotiation, customer context, and implementation feasibility. Finance owns definitions, cost, margin, realization, exposure, and approval. Category management owns category role, shopper structure, assortment, and retailer category impact. Demand and supply planning own forecast integration, capacity, inventory transition, and service risk. Data science owns demand models, elasticity, causal inference, validation, and experimentation. Legal and compliance own competition-law interpretation, contractual boundaries, pricing governance, and disclosure requirements.

The governance chapter in one breath: the agent does not own the price. It may analyse, simulate, recommend, and prepare; pricing authority stays with authorized individuals. Market data must be classified — publicly available, independently sourced syndicated, historical, retailer-provided, or commercially sensitive — because in the EU information exchange can occur indirectly through third parties, platforms, and algorithms, and the European Commission’s horizontal guidelines are explicit that algorithms neither remove corporate accountability nor excuse coordination. “The algorithm chose the price” is not a defence. The agent must not store or operationalize unverified information about a competitor’s intended future pricing; it must respect the legal distinction between manufacturer list price, customer invoice price, recommended shelf price, and independently determined retailer shelf price; and it must coordinate the consumer-facing consequences of pack changes — net-content display, master data, shelf labels, digital content, compliant claims. Qualified legal teams define permissible sources and jurisdiction-specific controls; the agent implements approved policy, it does not create it.

The failure surface is wide: the wrong pack, incompatible unit-price comparisons, promotional elasticity used for base pricing, ignored cross-product effects, old costs, misread retailer terms, overstated pass-through, omitted private-label switching, an operationally impossible pack, exposed sensitive information, false precision. So evaluation covers the whole workflow. Component tests check entity resolution, waterfall accuracy, unit conversion, elasticity retrieval, cross-price response, break-even calculation, customer economics, pack hierarchy, policy retrieval, and financial simulation. Historical replay gives the system only what was knowable at the original decision date and compares recommended action, expected volume, net revenue, and contribution against what happened — while staying honest that the unchosen scenario was never observed. Prospective experiments — selected stores, regions, channels, phased timing, matched products, digital audiences — are the stronger causal evidence wherever they are commercially and legally feasible.

Trajectory tests require the agent to resolve pack and customer, retrieve current net economics, select an applicable elasticity, model cross-effects, calculate manufacturer and retailer outcomes, check operational constraints, and request approval — while prohibiting invented competitor plans, unapproved prices, ignored pack switching, free-form financial arithmetic, and execution before authorization. The metric families follow: model (out-of-sample elasticity error, directional accuracy, interval coverage, stability, WAPE and bias, switching-share accuracy), decision (ranking accuracy, accepted recommendations, override rate and override value, affordability risk identified, infeasible recommendations, time to decision), financial (price realization, net-revenue growth, margin, contribution, mix, customer profitability, gross-to-net leakage, cost per decision), shopper and market (penetration, frequency, units per trip, private-label switching, category exit, share, pack and channel mix), operational (implementation accuracy, price-file errors, obsolete inventory, transition cost, supply disruption, retailer rejection, delays), and safety (prohibited data use, unauthorized pricing action, approval bypass, cross-customer leakage, unsupported competitor claims, invalid pack conversions).

The headline is not “average price increased by 6%.” It is sustainable portfolio contribution improved while shopper, retailer, market, and operational constraints were maintained.

Consider an illustrative snacks portfolio: a 120 g entry pack at €1.49 (€1.24/100 g), a 250 g core at €2.49 (€1.00), a 450 g family at €3.99 (€0.89), and a 700 g stock-up at €5.49 (€0.78). Manufacturer unit contributions run €0.35, €0.63, €0.85, and €0.85 on annual volumes of 4.0m, 8.0m, 5.0m, and 2.0m units — €12.39m of annual contribution. Costs rise 7%. The company wants to protect contribution while keeping the brand affordable.

A uniform 6% increase (Scenario A) is the reflex, and it pushes the entry pack to €1.59, the family across €4, and the stock-up toward €6. With pack elasticities of roughly −1.8, −1.1, −0.8 and −1.3, the naive volume effects are −10.8%, −6.6%, −4.8% and −7.8% — before any cross-pack switching, which is exactly why that arithmetic is a starting point and not a result. The portfolio model then shows entry shoppers exiting or moving to private label, core shoppers moving to family, family shoppers moving back to core, and stock-up frequency falling. The alternatives keep the same cost recovery and redistribute where it lands: a selective increase (B) holding entry at €1.49 and taking more on the core; an entry resize(C) from 120 g to 110 g at an unchanged €1.49, which holds the outlay and raises the unit price to €1.35/100 g; a new €0.99 trial pack(D) at 70 g that creates a visible sub-€1 entry and frees the existing small pack to move; and premiumization with simplification(E), removing the slow stock-up pack and adding a 200 g premium at €2.99.

Scenario E shows the highest contribution and the lowest confidence. Scenario C beats B on paper and carries trust, transition, and retailer acceptance risk. The agent recommends B as the most robust near-term actionand proposes testing D before any broader launch. That is a materially more useful output than “increase all prices by 6%” — and every number here is illustrative. Production decisions require validated models, current financial inputs, retailer context, and human approval.

1. 01Phase 0 — align definitions. Net revenue, gross-to-net, contribution, price realization, elasticity, pack roles, price index, customer profitability, strategic objectives.
2. 02Phase 1 — build the data foundation. Product master, customer and pack hierarchies, prices, trade terms, costs, volume, promotion, distribution, competitor observations, retailer shelf price.
3. 03Phase 2 — establish the analytical foundation. Price waterfall, elasticity and cross-price models, demand models, pack-switching model, financial simulator, customer economics.
4. 04Phase 3 — historical diagnostic agent. The agent explains past price changes, realization, elasticity, mix, leakage, and switching. It does not recommend live actions yet.
5. 05Phase 4 — scenario copilot. The agent generates and evaluates scenarios; humans retain full decision and implementation control.
6. 06Phase 5 — shadow recommendations. Recommendations run alongside the existing process; compare decisions, outcomes, time, and quality.
7. 07Phase 6 — approval-based workflow. The agent creates decision packets, negotiation packets, implementation tasks, and approval requests.
8. 08Phase 7 — continuous monitoring. Realization, volume, mix, pass-through, switching, and market response, watched continuously rather than reviewed annually.
9. 09Phase 8 — policy-bounded standard actions. Low-risk analytical and administrative steps automate; material pricing decisions stay governed.

The first pilot should be narrow and well-instrumented: one country, one category, one brand, two or three pack sizes, one or two channels, reliable weekly POS, recent price variation, clean cost and net-revenue data, and engaged RGM and sales leaders. Set the objectives in advance — cut analysis cycle time by 60%, identify gross-to-net leakage, improve price-response forecast accuracy, improve the contribution of approved actions, reduce pack-ladder inconsistencies, and achieve zero unauthorized pricing actions. The minimum viable data is SKU and pack hierarchy, customer hierarchy, regular and promotional price, manufacturer net price, volume, distribution, trade terms, cost, promotion calendar, channel, and market or competitor price; performance improves with household panel, loyalty and store-level POS, online prices, demographics, basket data, measured pass-through, media, availability, willingness-to-pay research, product attributes, and occasions.

Three unglamorous foundations decide whether any of it works. Unit standardization: every record carries consumer unit, net content, unit of measure, equivalent volume, pack count, case conversion, currency, and tax basis. Price taxonomy: list, invoice, net invoice, manufacturer realized, recommended shelf, observed shelf, promotional shelf, and unit price are eight different numbers, and mixing them destroys the analysis. Promotion cleansing and availability control: a price drop may be a promotion, a permanent change, a clearance, a loyalty offer, or a data error — and a volume decline after a price increase may actually be a delisting, a store loss, an out-of-stock, or a shelf reduction.

1. 01Maximizing average price. Average price rises because mix changed, not because pricing improved.
2. 02One elasticity for the brand. Response differs by pack, channel, retailer, and price point.
3. 03Ignoring cross-price effects. The core loses volume to the entry pack; the model calls it brand loss.
4. 04Confusing regular and promotional elasticity. Promotion response used to forecast a base-price increase.
5. 05Ignoring retailer pass-through. Modelling a shopper price the retailer never implements.
6. 06Ignoring gross-to-net leakage. List-price growth disappears through new terms and promotion.
7. 07Ignoring unit price. The shelf price holds while the value proposition quietly deteriorates.
8. 08Ignoring absolute outlay. Excellent unit economics on a pack that crosses a household budget threshold.
9. 09Treating private label as one competitor. Value, mainstream, and premium tiers behave differently.
10. 10Assuming linear elasticity. The decision crosses a price cliff the model cannot see.
11. 11Optimizing each SKU independently. Every SKU improves; the portfolio gets worse.
12. 12Using observational correlation as causality. Prices moved in response to demand conditions.
13. 13False precision. “Elasticity −1.274” with no interval and no model quality.
14. 14Letting the language model calculate economics. Use deterministic financial tools.
15. 15Ignoring implementation cost. Packaging write-offs, relisting, artwork, line changes, master-data work.
16. 16Excessive pack proliferation. Every channel gets its own pack; complexity destroys the value.
17. 17Channel conflict. Divergent packs and prices create arbitrage and retailer disputes.
18. 18Autonomous pricing authority. Material pricing decisions implemented without authorized human control.
19. 19Inappropriate competitor data. Sensitive or future competitor information used outside approved policy.
20. 20No post-action learning. Every annual pricing round starts from scratch.

1. 01Pin down the growth objective. Margin need, growth need, affordability, penetration, premiumization, customer objective, horizon, constraints.
2. 02Resolve the portfolio, customer, channel, and economic facts. Products, packs, unit conversions, customers, prices, trade terms, costs, distribution, current architecture.
3. 03Infer causal demand response. Own and cross-price elasticity, pack and channel switching, category exit, short- and long-run response, uncertainty.
4. 04Construct integrated price-pack scenarios. Price, resize, new pack, delist, premiumize, channel differentiation, trade-term change, promotion change, mix action.
5. 05Integrate manufacturer, retailer, shopper, and operational economics. Net revenue, contribution, retailer value, affordability, portfolio effects, implementation cost, feasibility.
6. 06Navigate governance, negotiation, and implementation. Decision rights, approvals, customer strategy, legal boundaries, implementation, monitoring, change control.
7. 07Generate validated learning. Realization, volume response, switching, mix, contribution, pass-through, forecast error, reusable lesson.

Part XIX in brief — the practitioner templates. The method ships as six working documents: the business brief(objective, market, category, brand, packs, customers, channels, cost change, margin and growth targets, affordability constraint, deadline); the elasticity card (product, customer, channel, current price and tested range, elasticity, confidence interval, horizon, model level, data period, cross-effects, validation, limitations); the pack-role card (GTIN, net content, shelf and unit price, role, target shopper, occasion, channel, key substitutes, cannibalization risk, recommendation); the scenario card (price and pack action, scope, expected volume, realization, net revenue, contribution, pack and channel switching, penetration effect, retailer economics, shopper outlay, operational cost, P10/P50/P90, required approvals); the decision packet; and the post-action learning card (approved versus implemented decision, planned versus actual effective date and shelf price, expected versus actual response and switching and contribution, realization, variance drivers, reusable learning, reviewer). If these artifacts do not exist, the decision was not designed — it was improvised.

Does the Agentic Revenue Growth Manager set prices automatically?

No. It should not hold autonomous authority over material pricing. It analyses, simulates, recommends, prepares, and monitors within organizational governance; authorized humans decide.

Why can’t we use one elasticity number?

Because elasticity varies by product, pack, retailer, channel, shopper, price range, competitive environment, and time horizon — and the value of the number depends entirely on how it was estimated.

What is the difference between absolute price and unit price?

Absolute price is what the shopper pays at the till; unit price is the price per standardized quantity or usage. Both drive decisions, often in opposite directions, which is why pack architecture must optimize for both.

Should larger packs always have a lower unit price?

Usually they offer better unit value, but the optimal ladder depends on pack role, margin, consumption, channel, and strategic intent. What a ladder should never contain is an unexplained value inversion.

When should a company downsize a pack?

When protecting the absolute price point genuinely serves affordability and the change survives scrutiny on shopper value, trust, occasion fit, disclosure requirements, operational cost, and portfolio effects — not as a quiet substitute for a price conversation.

How should retailer pass-through be handled?

Model several pass-through assumptions rather than one, and monitor the actual shelf price after implementation. Demand responds to the shelf price, not the invoice price.

Can the agent use competitor pricing?

It may use legally approved public, historical, or independently sourced market information under company policy. Qualified legal teams define permissible sources and uses, and unverified future competitor pricing must never be stored or operationalized.

How do we estimate elasticity when price rarely changes?

Hierarchical models, pack and category analogues, cross-market evidence, and controlled experiments — with the model level and uncertainty disclosed rather than hidden behind a single decimal.

What is the best first pilot?

One category, one market, a limited number of packs and customers, reliable data, clear financial definitions, and an engaged RGM owner with the authority to act on the output.

What is the biggest AI mistake in RGM?

Building an automated price recommender that ignores gross-to-net realization, pack switching, channel effects, retailer economics, shopper value, governance, and implementation.

Revenue Growth Management is usually described as a set of commercial levers. The deeper reality is that it is a system of connected choices: raise price and volume changes, pack switching changes, mix changes, retailer economics change, promotion response changes, channel behaviour changes, and portfolio contribution changes. That is why pricing cannot be managed one SKU at a time, or through cost-plus logic, or through a historical elasticity, or through a dashboard. The real decision is which shopper proposition, in which product and pack, at which absolute and unit price, through which customer and channel, under which commercial terms, will create the strongest sustainable value for the shopper, the retailer, and the manufacturer.

The Agentic Revenue Growth Manager creates the decision layer that answers that question. It does not replace elasticity models, optimization engines, finance systems, RGM experts, customer teams, or legal controls — it connects them. It identifies what must be known, retrieves the correct evidence, selects the appropriate models, keeps regular-price and promotional response apart, maps own-price and cross-price effects, tests pack and channel switching, calculates gross-to-net realization, compares price, pack, promotion, assortment and mix scenarios, makes uncertainty visible, prepares decisions for authorized humans, monitors what the market actually does, and converts outcomes into learning. The PRICING Method walks the route: pin down, resolve, infer, construct, integrate, navigate, generate.

The defining question is not “how much price can we take?” It is how to redesign the revenue architecture so the portfolio keeps winning across affordability, value, differentiation, retailer economics, and manufacturer profit.