Marketing has developed an unhealthy obsession with precision.
A dashboard can now show the estimated return of a campaign, a keyword, a channel, a creative execution, a pricing move, a promotion, a customer segment or a single conversion path. That precision is useful, but it has created a dangerous assumption: the more specific the model, the better the decision.
That is not always true.
Some marketing decisions require a narrow response model. If a retailer needs to estimate how a 5% price cut affects weekly sales, a price response model is appropriate. If a media team needs to understand how much incremental sales a TV campaign generates after accounting for carryover, an advertising response model is useful. If a brand wants to optimise promotional depth, a promotion response model is the right instrument.
But not all marketing decisions are tactical. Some are structural. Some concern positioning, competitive defence, portfolio architecture, brand strategy, resource allocation and long-term market dynamics. In those cases, a specific response model can be too narrow. The firm may need a global model: a broader decision-support system that integrates consumer perceptions, competitive positioning, marketing mix effects and strategic objectives.
This is where models such as Defender and Perceptor belong. Their value is not that they estimate one variable more cleanly than a specialised advertising or price model. Their value is that they help management see the market as a system.
And in marketing, systems matter.
The measurement trap
The modern marketing industry often behaves as if every strategic question can be reduced to a response curve.
What happens if media spend increases?
What happens if price changes?
What happens if a discount is extended?
What happens if distribution improves?
These are important questions. They are also incomplete.
A brand is not merely the sum of its advertising elasticity, price elasticity and promotional response. It is also a position in a consumer’s mind, a set of associations, a competitive alternative, a portfolio asset, a margin structure and a future option.
A price model can tell a brand that demand may decline after a price increase. It may not tell the brand whether the price increase damages premium perception, invites a private-label challenge, or changes the brand’s long-term position.
An advertising response model can estimate short-term sales uplift. It may not tell the brand whether the advertising is strengthening distinctiveness, correcting a positioning weakness, or defending against a competitor’s perceptual advantage.
A promotion model can show that a discount lifts volume. It may not show whether the promotion trains customers to wait, erodes reference price, or cannibalises future purchases.
This is the measurement trap: a model can be accurate inside its frame and still be insufficient for the decision.
What global models are trying to do
Global marketing models are designed to support broader managerial decisions. They typically integrate several elements: market structure, consumer choice, brand positioning, competitive response, marketing actions and business outcomes.
Defender-type models are associated with defensive marketing strategy: how a brand should respond when competitors attack, when share is threatened, or when resources must be allocated to preserve a position. The emphasis is not only on one marketing variable, but on competitive posture.
Perceptor-type models are linked to perceptual mapping and positioning. They help managers understand how consumers perceive brands, how close competitors are in perceptual space, which attributes drive preference, and where repositioning opportunities may exist.
The common thread is that these models are strategic rather than merely reactive.
They ask questions such as:
Where is the brand positioned relative to competitors?
Which attributes explain preference?
Which competitor poses the greatest threat?
Should the brand defend its current position or reposition?
Which combination of product, price, communication and distribution supports the desired strategy?
How should resources be allocated when several marketing levers interact?
What are the long-term risks of optimising only for short-term response?
These are not questions a single-variable response model can answer alone.
What specific response models do well
Specific response models are built around a narrower causal or statistical relationship. They estimate how a marketing variable affects an outcome.
Advertising response models estimate the effect of advertising on sales, awareness, consideration or other outcomes. Classic work in advertising response has dealt with carryover, diminishing returns, lagged effects and saturation. The managerial output is often a response curve: spend more up to the point where incremental return declines.
Price response models estimate how demand changes when price changes. They can capture own-price elasticity, cross-price elasticity, reference price effects and competitive price response. Work such as Aschersleben and colleagues’ A Semiparametric Approach to Estimating Reference Price Effects in Sales Response Models (2022) shows how price response can be modelled while recognising that current sales depend not only on current price but also on past prices and reference price formation.
Promotion response models estimate how discounts, displays, coupons or trade promotions affect volume, revenue and sometimes stockpiling or post-promotion dips.
Marketing mix models estimate the contribution of multiple channels and variables, often using aggregate time-series data. They sit somewhere between global and specific models: broader than a single advertising response model, but still often focused on decomposition of sales and ROI rather than full strategic positioning.
Specific response models are powerful because they are operational. They help answer “how much?” and “what if?” questions around a defined lever.
Their strengths are clarity, actionability and measurable accountability.
Their weakness is that they often assume the strategic frame is already correct.
The rise of MMM has made the distinction more important
Marketing mix modelling has returned to the centre of measurement debate because privacy changes, cookie loss and attribution limitations have weakened user-level tracking. Industry sources such as Gartner’s Marketing Mix Modeling: A CMO’s Essential Guide for 2025 frame MMM as a core tool for measuring the impact of marketing investment and improving ROI. Circana’s 2026 discussion of MMM similarly presents it as a way to measure incremental ROI, account for external factors and optimise budgets.
This renewed interest is justified. MMM can help brands understand how channels, price, promotions, seasonality, distribution and macro factors contribute to business outcomes.
But MMM does not automatically solve the global versus specific model question. A marketing mix model may cover multiple variables, yet still be designed mainly for budget optimisation. It may tell a CMO to move spend from one channel to another. It may not tell the business whether the brand’s underlying positioning is wrong, whether category structure is shifting, or whether the company should defend share differently against a new entrant.
This is why model purpose matters.
A global model is not simply a bigger regression. It is a decision-support framework for strategic choices.
When a global model is the better choice
A company should favour a global model when the decision involves interdependence, strategic uncertainty or market structure.
The first context is positioning. If the question is “what should the brand stand for?” or “how should the brand be repositioned against competitors?”, a variable-specific response model is too narrow. Perceptor-type models and perceptual mapping approaches are better suited because they connect consumer perceptions to preference and competitive space.
The second context is competitive defence. If a competitor launches a new product, cuts price, increases advertising or attacks a brand’s core benefit, the response cannot be evaluated one variable at a time. A Defender-type framework can help assess whether to defend through price, communication, product improvement, distribution, segmentation or some combination.
The third context is portfolio strategy. If a company manages multiple brands, product lines or tiers, a price model for one SKU may create local optimisation and portfolio damage. A global model can account for cannibalisation, positioning overlap and segment coverage.
The fourth context is market entry or repositioning. When a brand enters a new market, historical response data may be limited or misleading. A global model that incorporates consumer perceptions, competitor positions and simulated scenarios may be more useful than a narrow response model trained on past data.
The fifth context is long-term brand building. Advertising response models often focus on short-term sales effects because those are easier to observe. But brand strategy also involves mental availability, differentiation, salience and future demand. A global model can help evaluate actions that do not immediately convert into sales but shape future response.
The sixth context is major strategic planning. Annual planning, category strategy, brand architecture and investment allocation require a system view. Specific response models can feed the process, but they should not dominate it.
In short, a global model is preferable when the question is not “what is the effect of this lever?” but “which levers should be used, in which combination, to defend or improve market position?”
When a specific response model is the better choice
A company should favour a specific response model when the decision is narrow, measurable and lever-specific.
If the question is “what happens if price increases by 3%?”, use a price response model.
If the question is “how much incremental sales does advertising generate over the next eight weeks?”, use an advertising response model.
If the question is “what promotional depth maximises contribution margin?”, use a promotion response model.
If the question is “what is the saturation point for paid search spend?”, use a channel response model.
Specific response models work best when the company has sufficient historical data, the marketing lever is clearly defined, the outcome is measurable and the decision horizon is short to medium term.
They are also useful for experimentation. A/B tests, geo-tests and incrementality studies can estimate variable-specific effects with more causal confidence than broader models, provided the design is sound.
The danger comes when companies use these models for decisions they were not built to answer.
A price elasticity model should not define brand strategy. A media response curve should not decide positioning. A promotion model should not determine whether a brand has pricing power. A short-term ROI model should not be allowed to erase long-term brand investment.
Specific response models are instruments. They are not the orchestra.
The hidden cost of model fragmentation
Many organisations now have too many models and too little integration.
The media agency has an attribution model.
The pricing team has elasticity models.
The CRM team has churn models.
The finance team has forecasting models.
The brand team has tracking data.
The insight team has segmentation.
The e-commerce team has conversion models.
Each model may be useful. Together, they may still fail to answer the strategic question.
This fragmentation creates several problems.
First, models optimise against different outcomes. Media may optimise sales, pricing may optimise margin, CRM may optimise retention, and brand may optimise consideration. Without a global framework, trade-offs become political.
Second, models operate on different time horizons. Paid media looks at days or weeks. Brand tracking looks at months. Pricing may affect immediate demand and long-term reference price. Strategy may require years.
Third, models use different units of analysis. Some are customer-level, some brand-level, some SKU-level, some market-level. This makes integration difficult.
Fourth, local optimisation can create system-level damage. A price cut may improve volume but weaken premium positioning. A performance campaign may improve acquisition but attract low-value customers. A promotion may lift this quarter’s sales while depressing future baseline demand.
Global models help reduce this fragmentation by forcing the organisation to define the system.
The academic roots: decision support, not dashboard decoration
The value of global models lies in decision support. That idea has a long tradition in marketing science.
Little’s Models and Managers: The Concept of a Decision Calculus (1970) argued that marketing models should be understandable, robust, controllable, adaptive and complete on important issues. This remains a useful standard. A model that is statistically elegant but managerially unusable is not enough.
Urban and Hauser’s work on product and marketing strategy, including models related to positioning, preference and market response, also emphasised the practical use of models in managerial decisions. Perceptual mapping and preference modelling were designed to help managers understand how consumers see markets and where strategic opportunities exist.
Advertising response research, from Vidale and Wolfe’s early sales response model to later work on advertising carryover and dynamic effects, helped formalise how communications investment translates into sales over time. Price response and reference price literature similarly advanced understanding of how consumers react to price levels, promotions and past prices.
The point is not that global models replace response models. The literature suggests a hierarchy. Response models estimate components of market behaviour. Global decision models integrate those components into strategic choices.
The practical decision rule
A useful way to decide between global and specific models is to ask five questions.
First: is the decision about one lever or several?
If one lever, a specific response model may be enough. If several levers interact, a global model is safer.
Second: is the decision tactical or strategic?
If the decision concerns next month’s price or media spend, use a response model. If it concerns positioning, defence, brand architecture or long-term resource allocation, use a global model.
Third: is the outcome direct or mediated?
If price directly affects unit sales, response modelling works. If advertising affects awareness, consideration, preference, pricing power and future demand, a broader model may be needed.
Fourth: is the competitive context stable or changing?
In stable contexts, historical response models are more reliable. In changing contexts, global scenario models can help managers think beyond the past.
Fifth: is the risk local or systemic?
If a wrong decision affects only one campaign, a specific model is acceptable. If a wrong decision affects brand equity, market position or portfolio economics, use a global model.
The answer is often not either/or. A mature company should use both.
How global and specific models should work together
The strongest marketing decision systems use global models as the strategic frame and specific response models as components.
A global model defines the market, the competitors, the positioning, the objectives and the strategic alternatives. Specific models estimate the likely effects of advertising, price, promotion, distribution or product changes within that frame.
Consider a premium brand facing a low-cost entrant.
A price response model can estimate what happens if the premium brand cuts price.
An advertising response model can estimate whether increased media spend can defend demand.
A perceptual model can show whether consumers see the entrant as a true substitute.
A Defender-type global model can help decide whether the brand should cut price, reinforce premium messaging, launch a value sub-brand, improve product benefits, change channel strategy or accept some share loss to protect margin.
The strategic answer may be: do not cut price, because the entrant is not perceived as equivalent among core buyers; increase communication around quality and launch a controlled value offer for price-sensitive segments.
No single price model would produce that answer.
The danger of overusing global models
Global models also have risks.
They can become too complex. A model that tries to represent the entire market may become impossible to explain or validate.
They can hide weak assumptions. If inputs on consumer perception, competitive response or long-term brand effects are poor, the model may produce elaborate but fragile scenarios.
They can become slow. Strategic models are often less useful for rapid tactical decisions.
They can become political. Because global models influence major resource allocation, teams may challenge assumptions that threaten their budgets.
They can create false integration. Combining many variables does not guarantee understanding. A global model must be grounded in theory, evidence and managerial purpose.
So the case for global models is not a case for bigger models. It is a case for better-framed models.
The danger of overusing specific response models
The risks of specific response models are more common.
They can overvalue what is easy to measure. Short-term sales responses often receive more attention than brand effects, customer quality or future demand.
They can encourage tactical over-optimisation. A business may become excellent at optimising promotions while destroying margin discipline.
They can understate interaction effects. Advertising can change price sensitivity. Price can change brand perception. Distribution can change advertising effectiveness. Promotions can change baseline demand.
They can become backward-looking. Models trained on past response may fail when consumer behaviour, competitive structure or media environments change.
They can mislead strategy. A channel with high short-term ROI may not build future demand. A price cut with strong volume response may weaken the brand.
This is why specific models must be governed by strategic context.
What this means for CMOs
The CMO’s job is not to choose between global and specific models. It is to build a measurement architecture where both have a clear role.
The CMO should ask: which decisions are we trying to improve?
If the organisation needs to allocate media budget, MMM and advertising response models are central.
If it needs to set prices, price response and elasticity models are necessary.
If it needs to defend share, reposition the brand or enter a new segment, global models such as Perceptor or Defender-style frameworks are more appropriate.
If it needs to understand long-term effects, the model architecture must include both short-term response and longer-term brand or customer equity measures.
If it needs to align finance, marketing and sales, the model must be explainable enough to create shared confidence.
The CMO should also prevent model imperialism. No team’s model should become the entire truth. Media attribution is not brand strategy. Price elasticity is not customer value. Brand tracking is not sales forecasting. MMM is not a complete theory of growth.
A good organisation treats models as decision tools, not belief systems.
The final verdict
A company should favour a global model when the decision is strategic, multi-variable, competitive, positioning-led or long-term.
It should favour a specific response model when the decision is tactical, lever-specific, data-rich and outcome-focused.
The mistake is to confuse specificity with superiority.
A specific model may be more precise, but only within its frame. A global model may be less precise, but more relevant to the real decision. The most dangerous model is not the one with uncertainty. It is the one that answers the wrong question with confidence.
Marketing does not need fewer models. It needs better model discipline.
That means starting with the decision, not the dataset. It means knowing when to zoom in on price, advertising or promotion, and when to zoom out to market structure, positioning and competitive defence.
Because the real question is not whether a model is global or specific.
The real question is whether it helps the business make the decision that matters.
References
Aschersleben, P., et al. (2022). “A Semiparametric Approach to Estimating Reference Price Effects in Sales Response Models.” Journal of Business Economics.
Bass, F. M. (1969). “A New Product Growth Model for Consumer Durables.” Management Science.
Hanssens, D. M., Parsons, L. J., & Schultz, R. L. (2001). Market Response Models: Econometric and Time Series Analysis. Kluwer Academic Publishers.
Little, J. D. C. (1970). “Models and Managers: The Concept of a Decision Calculus.” Management Science.
Lilien, G. L., Kotler, P., & Moorthy, K. S. (1992). Marketing Models. Prentice Hall.
Naik, P. A., & Raman, K. (2003). “Understanding the Impact of Synergy in Multimedia Communications.” Journal of Marketing Research.
Urban, G. L., & Hauser, J. R. (1993). Design and Marketing of New Products. Prentice Hall.
Vidale, M. L., & Wolfe, H. B. (1957). “An Operations-Research Study of Sales Response to Advertising.” Operations Research.
Gartner. (2025). “Marketing Mix Modeling: A CMO’s Essential Guide for 2025.”
Circana. (2026). “Marketing Mix Modeling Insights: Driving ROI in a Complex Landscape.”
