Bugeting Allocation Revisited
David Merrick, Business Economics Limited
Karl Weaver, Data2Decisions Limited
In recent years numerous articles have been written about new methods and techniques for budget allocation. A fair proportion of these
have been published in Admap, and have highlighted some sharp divisions in approach. Dyson ([1],[2]), for example, describes a
mathematical process for allocating budget across a portfolio of brands based on advertising response curves, and outlines the financial
benefits as well as the pitfalls.
By contrast, Harper and Bridges ([3]) argue that the response-curve-based approach to budget allocation is too complicated and 'black box'
for clients to understand. As such, it constrains the budget-allocation process. Instead, they advocate a scoring-system approach.
While both sides underline the importance of client involvement, at first sight there appears to be little else in common. Our purpose in
writing this article is to contribute to the debate by examining these approaches to see whether they are as different as they appear to be
and to investigate to what extent it is possible to use the best features of each. We take as a starting point the advantages of the responsecurve
approach.
It is widely accepted in the literature as a way to describe how advertising works (for example, Batra, Meyers and Aaker ([4]), East ([5]),
and Rossiter and Percy ([6]), to name but three).
Karl Weaver, Data2Decisions Limited
In recent years numerous articles have been written about new methods and techniques for budget allocation. A fair proportion of these
have been published in Admap, and have highlighted some sharp divisions in approach. Dyson ([1],[2]), for example, describes a
mathematical process for allocating budget across a portfolio of brands based on advertising response curves, and outlines the financial
benefits as well as the pitfalls.
By contrast, Harper and Bridges ([3]) argue that the response-curve-based approach to budget allocation is too complicated and 'black box'
for clients to understand. As such, it constrains the budget-allocation process. Instead, they advocate a scoring-system approach.
While both sides underline the importance of client involvement, at first sight there appears to be little else in common. Our purpose in
writing this article is to contribute to the debate by examining these approaches to see whether they are as different as they appear to be
and to investigate to what extent it is possible to use the best features of each. We take as a starting point the advantages of the responsecurve
approach.
It is widely accepted in the literature as a way to describe how advertising works (for example, Batra, Meyers and Aaker ([4]), East ([5]),
and Rossiter and Percy ([6]), to name but three).
- It is widely understood and used by media agencies in planning and analysing campaigns.
- It has rigorous and well-defined links to profitability that can form the basis of a dialogue between marketing and finance departments.
- The quantifications can be based on actual performance through econometric analysis.
- But equally, the approach can be based on the judgmental views of management, acquired through a structured process.
We believe this last point is central to the debate, because it enables a scoring and ranking approach to be embedded in the responsecurve
methodology while retaining the other advantages.
Response-curve budgeting
Before looking in detail at methods of implementation, here is a brief recap of how the response-curve-based approach to budget allocation
works.
There is no mystique about a response curve. It is simply a way of describing how sales (typically) respond to advertising spend which is
something that arguably we ought to have a view on anyway. Surprising though it is, at its simplest we need just two numbers to describe
the response of a brand in a market to advertising spend:
methodology while retaining the other advantages.
Response-curve budgeting
Before looking in detail at methods of implementation, here is a brief recap of how the response-curve-based approach to budget allocation
works.
There is no mystique about a response curve. It is simply a way of describing how sales (typically) respond to advertising spend which is
something that arguably we ought to have a view on anyway. Surprising though it is, at its simplest we need just two numbers to describe
the response of a brand in a market to advertising spend:
- the maximum increase in sales that can be achieved by advertising (money no object)
- the spend required to obtain half of this maximum increase.
Consider the simple example of a single brand operating in a single market (Box 1).
The calculations to obtain these results are very simple. At most, they require the simplest of spreadsheets, and could be done with a hand
calculator.
There are a number of assumptions embedded in the example, and many more complicated situations that have to be handled. We look at
these later, but it is the simplicity of the underlying idea that we want to stress here.
The concept of the response curve has been around formally in economics for over 100 years. Known as Gossen's second law of marginal
utility, it is used to explain how consumers trade off their income across goods and services.
Imagine a world where you can only buy Mars bars and apples with your earnings, and they cost the same. If you have a sweet tooth you
might buy a Mars bar, eat it and enjoy it more than you would have the apple. Eventually you will get fed up with Mars bars, either because
you have been greedy and eaten several in quick succession, or you have just got bored with eating them for every meal. The additional
pleasure ('utility' in impassive economics-speak) you get from another Mars bar is less than from eating an apple. So you buy an apple
instead.
The theory goes that if humans behave in a rational way, we are 'optimising' all the time between Mars bars and apples, apples and
oranges, oranges and a new settee (for sitting on, not for eating!), work and leisure, spending and saving, and so on. As we consume we
travel along the response curve. What is important is where our point of saturation is and how quickly we arrive at it. These factors are
essentially describing the shape of the curve.
Incorporating reality
The example given above is deliberately simple: real-world problems are significantly more complex. For example, we could be dealing
with many brands in many markets. We could, of course, repeat the above calculation for all of the brand and market combinations, add
them up and find what the overall budget is for maximum profitability. This might indeed be a useful exercise, but it might also give us a
nasty shock. The optimum budget could be larger perhaps substantially larger than the amount available.
We can handle this by allocating a fixed budget optimally across the brands and markets with the use of a spreadsheet to carry out the
calculations.
A further issue is how we define profit. In the above example, we used a 'constant profit margin' approach. This takes sales revenue less
variable costs less advertising expenditure as our profit measurement. In general, the profit margin for a particular brand in a market is a
significant factor in the calculation illustrated. And it follows, logically, that as the margin derived from sales of a brand increases, more can
be spent on advertising and it can still remain profitable.
The 'constant profit margin' approach is better than using 'net sales revenue' (sales revenue less advertising costs), which ignores variable
costs completely and effectively assumes a gross margin on sales of 100%.
However, the 'constant profit margin' approach is still unsatisfactory, as it ignores the fixed costs in the business. A better approach is the
standard accounting 'fixed costvariable cost' model. In fact, as shown by Shaw and Merrick ([7]), the 'fixed costvariable cost' approach will
always give an 'optimum' advertising spend that is higher than the 'constant profit margin' model. This is ironic, given how widely used the
'constant profit margin' model is in marketing departments and media agencies. Ignoring fixed costs is an assumption borne out of
convenience and is always an approximation.
To illustrate this, we have reworked the optimisation in the first example, assuming that the fixed costs of the business are 50% of the total
costs of the business. The results are given in Box 2.
The effect of introducing fixed costs into the equation is to double the estimate of the optimum amount to be spent on advertising. The
estimated effect of optimal advertising on the profit margin also increases, from 1 percentage point to 4 percentage points.
Equally, of course, the painful truth is that there is no guarantee that an optimisation will support the case for any advertising expenditure
at all. If, in the above example, the profit margin were only 25% of sales revenue and the fixed costs were only 10% of total costs, we
would be in the situation where any spend on advertising would reduce profitability.
This illustrates a general and important point about the response-curve approach. Some key numbers the profit margin and the percentage
of fixed costs will come from the finance department. As such, they can be regarded as given. Although we still have to estimate the two
numbers that describe the response of sales to advertising, the calculations are not totally dependent on these estimates and, in fact, are
often less sensitive to the estimates (within a sensible range) than might be thought.
Other factors that will affect the allocation of budget and therefore should be considered are as follows.
- Media costs As the cost of media for a brand in a market increases, logically you might expect it to take a smaller share of a fixed total budget. This is true after a point. However, if the incremental profits derived from advertising remain relatively high, the spend on the brand in the market can increase, even though the volume of advertising bought might fall.
- Carry-over This is often mentioned in connection with how advertising works, but sometimes overlooked in budget allocation. If there is a carry-over effect from last year's advertising, this effectively raises the level of advertising that already exists in the budgeting period which, in turn, will reduce the budget allocated to that particular brand. Clearly this impact depends on the level of spend as well as the carry-over rate. It will also depend on when the money was spent during the past year. Similarly, when assessing the impact of this year's budget, we should take into account any effects carried over into the following year(s), unless the objective is solely concerned with this year.
- Phasing In a similar way, the allocation will depend on the phasing of the media. This is a direct result of the diminishing- returns effect of a response curve. If a lot of spend is loaded into a short time then you are more likely to 'max out' on the curve for that particular brand. Other things being equal, this will mean that the marginal return from spending on another brand or market is likely to be higher. By contrast, if the spend is more evenly spread out, this will lead to more money being allocated to that brand.
- Long-term effects It is often the case that calculations such as those described focus on the short-term effects of advertising. If long-term effects are thought to be important in a particular case, the approach can be adapted to take these into account. For example, we can incorporate relevant brand equity measures, tracking research or sales effects over longer periods of time.
Describing the response
The key issue that remains is where we get the numbers from to describe the response of sales to advertising. Basically, there are two
approaches.
The key issue that remains is where we get the numbers from to describe the response of sales to advertising. Basically, there are two
approaches.
- Econometric analysis Given good data, this has many advantages, not least that it is (or should be) objective and therefore provides an impartial view on what is working and what is not. It also gives realism to what can be achieved for the brands in question. It requires time and effort but will, as a side effect, also produce quantifications of other marketing activities such as pricing and promotions, allowing the different expenditures to be compared.
- Judgment This approach is based on ranking and scoring the brands and media in a structured way, and then analysing the data to obtain the inputs needed for the optimisation.
There is extensive literature on econometric analysis (for example, Hays ([8]) and Greene ([9])) and we will not dwell on it here. Instead,
we focus on judgmental estimates, because in our view this is where the two approaches described at the beginning of the article
converge.
The simplest and most direct approach is to ask those familiar with the brands and markets to provide views on the two key parameters in
the example given earlier either directly or by scoring or ranking the brands of interest. Even where managers are not confident of their
responses, a Delphi process of feeding back the aggregate results and inviting revised views can provide good guidance. Of course, where
there are no quantitative data, such as a brand launch or repositioning, employing judgment in this way is inevitable.
A similar approach that we have found useful combines the views of managers with benchmarking data. In some cases a client already has
relevant benchmarks available from previous analyses of its own brands (see Foley and Kloprogge ([10]) for a recent example). In other
cases it is possible to find relevant reference points, for example, from published literature.
The process is best carried out in several stages, including:
we focus on judgmental estimates, because in our view this is where the two approaches described at the beginning of the article
converge.
The simplest and most direct approach is to ask those familiar with the brands and markets to provide views on the two key parameters in
the example given earlier either directly or by scoring or ranking the brands of interest. Even where managers are not confident of their
responses, a Delphi process of feeding back the aggregate results and inviting revised views can provide good guidance. Of course, where
there are no quantitative data, such as a brand launch or repositioning, employing judgment in this way is inevitable.
A similar approach that we have found useful combines the views of managers with benchmarking data. In some cases a client already has
relevant benchmarks available from previous analyses of its own brands (see Foley and Kloprogge ([10]) for a recent example). In other
cases it is possible to find relevant reference points, for example, from published literature.
The process is best carried out in several stages, including:
- collecting the data.
- analysing the data and modelling optimum allocations under various assumptions
- reviewing the results and refining the assumptions.
In our view, these steps are best done in a series of workshops with the client. This enables more 'realism' to be incorporated through the
inclusion of operating constraints such as minimum spend levels, or specific sales targets. Workshops are also a forum for developing a
deeper understanding and challenging the assumptions behind the model.
Generally speaking, if the data and time are available then the objectivity and rigour of econometric analysis would make this the
preferred approach.
Whichever approach is adopted, in the end we would not expect the output of the model to be followed exactly, but it is a useful guiding
tool and a structured process for arriving at a solution that can be understood and implemented. Invariably the budget is not allocated in
exactly the way the computer calculates. Several iterations are required to incorporate qualitative judgments about factors such as longterm
strategic importance, distribution issues, competitor actions and other business considerations. These issues may not align with the
results of a short-term profit-maximising analysis. What is critical is to ensure that the process is as transparent as possible.
Conclusion
There is no doubt about the importance of allocating advertising budgets effectively. We believe the response-curve approach provides a
rigorous framework for this key activity, but it need not be tied to opaque 'black box' techniques. The concepts are simple and clear, and a
range of techniques exists for identifying the performance characteristics of the brands and markets needed for the optimisation. The
response-curve approach can be supported with econometric analysis, taking advantage of the rigour and objectivity of this technique.
Where time and data are not available, it can be based on judgments and benchmarks, or often in practice it is a combination of both
approaches. Using response curves does not replace business planning expertise, but our experience is that, used in the correct way, the
approach can greatly assist the budget-allocation process by providing specific and valuable guidance. Furthermore, it enhances
management's understanding of how the brands and markets relate to each other, and promotes a rational and constructive debate of the
issues.
inclusion of operating constraints such as minimum spend levels, or specific sales targets. Workshops are also a forum for developing a
deeper understanding and challenging the assumptions behind the model.
Generally speaking, if the data and time are available then the objectivity and rigour of econometric analysis would make this the
preferred approach.
Whichever approach is adopted, in the end we would not expect the output of the model to be followed exactly, but it is a useful guiding
tool and a structured process for arriving at a solution that can be understood and implemented. Invariably the budget is not allocated in
exactly the way the computer calculates. Several iterations are required to incorporate qualitative judgments about factors such as longterm
strategic importance, distribution issues, competitor actions and other business considerations. These issues may not align with the
results of a short-term profit-maximising analysis. What is critical is to ensure that the process is as transparent as possible.
Conclusion
There is no doubt about the importance of allocating advertising budgets effectively. We believe the response-curve approach provides a
rigorous framework for this key activity, but it need not be tied to opaque 'black box' techniques. The concepts are simple and clear, and a
range of techniques exists for identifying the performance characteristics of the brands and markets needed for the optimisation. The
response-curve approach can be supported with econometric analysis, taking advantage of the rigour and objectivity of this technique.
Where time and data are not available, it can be based on judgments and benchmarks, or often in practice it is a combination of both
approaches. Using response curves does not replace business planning expertise, but our experience is that, used in the correct way, the
approach can greatly assist the budget-allocation process by providing specific and valuable guidance. Furthermore, it enhances
management's understanding of how the brands and markets relate to each other, and promotes a rational and constructive debate of the
issues.
[1] P Dyson: How to manage the budget across a brand portfolio. Admap, December 1999.
[2] P Dyson: Setting the communications budget. Admap, November 2002.
[3] G Harper and D Bridges: Budgeting for healthier ROI. Admap, July/August 2003.
[4] . R Batra, J Meyers and D Aaker: Advertising management. Prentice Hall, 1996.
[5] . R East: Consumer behaviour. Prentice Hall, 1997.
[6] J Rossiter and L Percy: Advertising communications and promotion management. McGraw-Hill, 1996.
[7] R Shaw and D Merrick: Marketing Payback. Prentice Hall (to be published mid-2004).
[8] W Hays: Statistics. Harcourt Brace, 1991.
[9] W Greene: Econometric Analysis. Macmillan, 1993.
[10] T Foley and P Kloprogge: How channel planning tools can deliver ROI. Admap, March 2004.
[2] P Dyson: Setting the communications budget. Admap, November 2002.
[3] G Harper and D Bridges: Budgeting for healthier ROI. Admap, July/August 2003.
[4] . R Batra, J Meyers and D Aaker: Advertising management. Prentice Hall, 1996.
[5] . R East: Consumer behaviour. Prentice Hall, 1997.
[6] J Rossiter and L Percy: Advertising communications and promotion management. McGraw-Hill, 1996.
[7] R Shaw and D Merrick: Marketing Payback. Prentice Hall (to be published mid-2004).
[8] W Hays: Statistics. Harcourt Brace, 1991.
[9] W Greene: Econometric Analysis. Macmillan, 1993.
[10] T Foley and P Kloprogge: How channel planning tools can deliver ROI. Admap, March 2004.
© Copyright World Advertising Research Center 2007. Reproduced with permission of Admap, the world’s leading source of strategies for effective advertising, marketing, and research. Learn more at www.admapmagazine.com.
