You must answer several questions before starting to implement Optimize.

An Optimize session works at a single audience level, so depending on the audience level—what contactable entity such as customer, prospect, account, or household—you choose to optimize over, you may need more than one Optimize session.

Once you have decided on your audience level, one of your first questions should be, "By what metric or goal do I want to optimize?" For example, whether you are maximizing profit, revenue, or ROI. This will determine what "score" you will use for optimization—dollars of profit, revenue in dollars, probability of response, expected profitability in dollars, profitability in dollar over cost per offer, and so on. Optimize maximizes the sum of this score in the optimized (surviving) transactions. All marketing campaigns and flowcharts participating in the same optimization session must use the same meaning of "score" for optimization.

Once you know your optimization metric, you should ask "What rules or constraints does my business have that drive the optimization process?" The following describes different types of business problems you may want to address through optimization.

Channel capacity optimization—with fixed channel capacity, what channels should be used to communicate with which customers and when?

Customer optimization—with limited opportunities to communicate with my customers, which offers should I give to which customers while managing contact fatigue?

Offer optimization—which customers should get offer X, when you have only a limited number to distribute?

Time optimization—which offers conflict with each other? Which ones should be given only if another offer has already been presented? Which offers are most valuable to present to a particular customer first?

Budget optimization—how should I allocate my money across different campaigns, offers, or customers?

You can create optimizations that address many of these issues. You should also consider these rules and constraints for use across your organization, including global suppressions (such as ensuring customers who have elected to not receive email are not contacted by email) and rules you will use to manage contact fatigue (by segment, by channel, time periods, or specific offers).

Once you have determined the rules and constraints that are appropriate for you organization, you should ask yourself "for what period of time should I optimize?" The larger your time window—the amount of time you are looking into the future across proposed contacts that have not yet been delivered—the more opportunities there are to optimize. However, too large of a time window may be prohibitive for a number of different reasons. For example, trying to use a six-month time window would require that all your campaigns be designed and implemented at least six months in advance.

Optimize also uses scores as a measure for choosing one proposed transaction over another when solving the optimization problem. Therefore, you need a method for generating scores. Some common methods are as follows:

A constant per-offer or offer-cell combination

A computation per individual, offer, or offer-cell combination

A predictive model (based on individual, offer, channel, or time)

As you ask yourself these questions, you may want to start small, with only a few campaigns, one product line, or one business division. You should also start with simple optimization problems. Although you may want to maximize profit while working with offer, budget, and customer optimization, you may want to apply fewer rules and constraints at first, and analyze the results before adding more. Remember, each time you add more constraints, you reduce the optimality of your results. You should therefore focus on those that are most important to you first and weigh carefully the business benefit of additional constraints. After studying the results of your initial optimization sessions, you can add more rules and constraints to add complexity if desired. Figuring out the right set of rules and constraints to use for your business that will likely take some testing; you will likely need to refine your rules over time.