Once computing power became available, serious analysts built forecasting models that projected results for future periods and let marketers explore the impact of changes in their policies.

Marketers in other fields might have liked similar information. But with no way to track purchases by customer, they lacked the foundation. Knowing the relation between ad spending and sales, or even between specific promotions and specific purchases, isn’t enough. You need to know how many sales came from repeat customers to generate a meaningful estimate of lifetime value.

Modern databases give non-DMers ways to measure purchases by customer. Perhaps just as important, the availability of these databases has made non-DMers familiar with lifetime value concepts and reinforced their awareness of retention’s importance. Thus many are now ready to consider the sort of business forecasting models long familiar to direct marketers.

Marketing Budget Allocation Software (Marquant Analytics, 310/471-8979, www.marquantanalytics.com) is one of several modeling tools from Marquant that combines business forecast modeling with optimization — that is, finding the allocation of limited resources that produces the best result. In MBA’s case, the limited resource is the marketing budget and the desired result is maximum discounted cash flow, or net present value. MBA determines the level of total marketing spending and the division between acquisition and retention programs that yield the greatest net present value. It gives an objective answer to the eternal question, “What should my marketing budget be?”

This sort of modeling can be done at many levels of detail. MBA works at a very general level, simply distinguishing between expenses for acquisition and retention. Other Marquant products break things down further by allowing for multiple customer segments and for cross purchases. But none work at the individual customer level, so they cannot select specific names for actual promotions.

Despite its general nature, MBA considers the diminishing returns on incremental marketing investments. This is one critical feature that distinguishes marketing economics from many conventional optimization methods, which assume items such as unit costs are static. But marketers, at least in theory, make their most effective investments first. Thus, acquisition and retention costs per customer usually rise as the budget increases and marketers make more marginal investments. (This isn’t always the case. In a high-fixed-cost situation, such as an expensive TV ad or Web site, the average cost per customer may decline as the initial investment is spread over more names.)

Gathering realistic statistics for these values is a challenge that Marquant doesn’t really address: The system either generates smooth curves based on a few data points or accepts more precise inputs provided by the user. This simplification may bother detail-oriented users, but it is appropriate for the big-picture decisions that Marquant aims to facilitate. It also means that Marquant tools cannot optimize the mix of specific marketing programs.

In keeping with this approach, MBA asks for a relatively small number of inputs: numbers of prospects, converted customers and retained customers; transaction margin (i.e., profit contribution before marketing or fixed costs) per new and renewed customer; acquisition and retention budgets; and maximum acquisition and retention rates. The system converts the acquisition and retention budgets to cost per customer using the incremental cost curve already described. Profit contribution per customer is assumed to be constant, rather than declining as increased marketing brings in less-qualified customers. This is another oversimplification that is probably adequate for Marquant’s purpose.

MBA assumes that input values stay constant for up to 60 periods. It uses the values to estimate cash flow by period, taking into account new customers and retention of existing customers. Output reports show the net present value of spending, revenue and profit contribution for acquisition, retention and in total. Users also can view detail by period. The system compares cash flow for the current spending policies against flows from optimal policies and for other scenarios such as a fixed marketing budget, fixed increase per year or maximum spending level. It also can run elasticity analyses showing the effects of changes in the inputs themselves. Reports provide tabular and graphical outputs intended to be understood by non-technical viewers.

The market segment and cross-sell modules add some complexity to the forecast models but remain fairly simplistic. For example, the market segment model treats each segment as separate from all others rather than permitting customers to migrate from one segment to another — a common way to simulate the customer life cycle. The cross-sell model calculates the number of customers with each product combination in one period who buy each product combination in the next period. This could simulate customer migration, except that it is limited to three products. It serves Marquant’s goal, which is to consider cross-sales when calculating optimal marketing budgets.

Marquant’s limited functions are reflected in its prices. Costs range from $20,000 to $50,000 for an engagement, which is considerably less than most marketing optimization products. The price includes software plus help in preparing the initial models. Clients then retain the software to run as they please. Marquant’s products, introduced in 2004, evolved from a consulting practice begun 10 years prior. The software has been sold to about a half-dozen very large companies, where it is used for general planning by senior management rather than as a tactical marketing tool.