Seasonal Campaign Optimization is the Holy Grail for the multi-channel merchant. The closer you come to this ideal, the higher your revenues and profits. However, the path to Seasonal Campaign Optimization is challenging. It requires not only marketing professionals, but also decision scientists who provide the quantitatively-derived insights that drive the entire process. The decision sciences piece, although important, is outside the scope of this article. Instead, we will focus on the "nuts and bolts" of implementation:
Establishing a Longitudinal Analytical Framework
We begin where an earlier article ended ""Calculating Incremental Response," a cover story in the May 1, 2005 issue of Catalog Age. The main theme was the difficulty of accurately measuring performance in today's multi-channel world of overlapping promotions. The recommended solution was to create a "Longitudinal Analytical Framework" to support the continuous creation of knowledge. The article defined a Longitudinal Analytical Framework as:
- The creation of a robust repository of complete customer history; that is, a best-practices marketing database.
- The implementation and subsequent analysis of continuous waves of seasonal test treatments [emphasis added], in order to clarify the complex interactions of channels and promotions.
It is these "continuous waves of seasonal test treatments" that are so essential to Seasonal Campaign Optimization. Here is why:
Creating Seasonal Test Treatments
Creating a seasonal test treatment is analogous to making a movie where each scene is crafted carefully to tell an unfolding story. Each scene "that is, an individual customer contact "is coordinated with the previous one, and plays a role in telling the entire story over time. Generally, a separate movie "an across-time test treatment "should be made for each major season.
This multi-channel, multiple impression approach is far more effective than the traditional "one off" promotions that are conceived and executed in a vacuum. This is particularly true in complex environments such as when direct channels interact with brick-and-mortar retail; or in B-to-B "and especially those situations that contain the added complexity of a field sales force.
Creating a compelling movie "a seasonal test treatment "requires the following five questions to be answered:
First, which messages should extend with consistency across the season, and which need to unfold as the season progresses?
For example, "appropriate gifts for the holidays" can act as a season-wide theme, while "rapid shipment for last minute shoppers" can be overlaid at the appropriate time.
Second, what should be the level of marketing intensity for each customer segment?
The objective is to match marketing intensity with each segment's potential. A primary method is to adjust the number of catalogs, including re-mails. Additional ways to affect marketing intensity include:
- Supplemental contacts such as emails, including "priming" efforts in advance of "or reminders just after "catalog drops. Sometimes, outbound calls are appropriate, and especially in B-to-B environments.
- Increased page counts and the judicious use of more expensive postage.
- When applicable, efforts by the field sales force.
- For companies with a retail channel, in-store tie-ins such as postcards announcing specials limited to best customers.
Marketing intensity can be defined more broadly than just promotional touches; for example:
- Special 1-800 numbers for best customers.
- Increased investment in the resolution of service problems that involve best customers.
- More elaborate packaging and expedited shipping.
- Loyalty and points-based programs.
Third, what is the timing of the various types of promotions?
For example, should the second catalog drop be closer to the first or to the third? And, should emails be sent in advance of the catalogs to announce their impending arrival, or should they be positioned as a trailing reinforcement?
Fourth, what audiences should be selected for each promotion?
For example, how deep into the file should we go for each of the catalogs? And, which poor-performing customer segments should receive only emails?
Audience selection requires that promotional cost/benefit analysis be done. It is important for all such analysis to incorporate the estimated "baseline order activity." This is the revenue that would have occurred in the absence of any promotional activity, and generally differs dramatically by customer segment (see the May 1 article). These estimates are established through ongoing testing and analysis. Failure to include this baseline order activity generally results in the over promotion of certain customer segments and the misallocation of marketing dollars.
Fifth, should the messages be personalized? If so, then how?
For example, should the catalogs be mass-customized with the selective binding and digital messaging capabilities offered by today's sophisticated printers? If so, how should the business rules that drive such efforts be defined? Likewise, should the emails be tailored to individual customers?
Measuring the Effectiveness of Seasonal Test Treatments
Determining the effectiveness of seasonal test treatments requires a shift in measurement from by-the-promotion to by-the-season. This is because, in a multi-channel world of overlapping promotions, the allocation of purchase activity to specific promotions is a somewhat arbitrary exercise (see the May 1 article).
A seasonal test treatment employs multiple contacts and channels. Therefore, it is impossible to produce the optimal "movie" "that is, "Seasonal Campaign Optimization" in the parlance of the direct marketer "on the first attempt. However, one can evolve towards optimization with a "best guess" starting point and then testing one's way to improvement.
The following is a generalized four-step approach to Seasonal Campaign Optimization that you can adapt to the specifics of your business:
First, start with your existing "control" promotional strategy for each of your customer segments, including the sequencing. Generally, this will be the approach that you employed during the same season last year. For example, assume that you are a hybrid B-to-C/B-to-B marketer with three channels: a catalog, an ecommerce site and a national retail presence:
- For this hypothetical company, the existing (control) strategy for its peak Holiday season is to mail between one and three catalogs to each of its active customers based on the prediction of upcoming purchase activity; that is, an initial drop followed by as many as two re-mails. Only the best customers receive all three catalogs.
- The catalogs are supplemented by two emails, just prior to the first and third catalog.
- Prior to the Holiday season, outbound calls are made to B-to-B customers to update contact information and ship-to addresses.
- Baseline order activity is modest at the beginning of the season and builds steadily as December 25th approaches.
- The catalog mailings have increasing effect as December 25th approaches.
- Order activity drops dramatically after December 25th.
- B-to-B orders are mostly ship-to gifts.
- For reasons of operational complexity and expense, management is willing to create no more than two test treatments a season.
Second, create multiple seasonal test treatments with different marketing intensity compared with the control strategy. In our example, assume the creation of two panels as follows:
- Panel #1 adds a third email to the control strategy, plus a post-card that advertises a retail tie-in.
- Panel #2 adds a fourth catalog to the very best customers, plus a third and fourth email, plus a post-card that advertises a retail tie-in.
The following is an illustration of Panel #2, which has the greatest marketing intensity compared with Panel #1 and the Control. For the sake of simplicity, we will not consider strategies beyond promotional touches; for example, special 1-800 numbers for best customers:
Note that, within the limitation of two seasonal test panels, many of the possible permutations are not being tested; for example:
- The addition of a third email only.
- The addition of a retail tie-in only.
- The addition of a fourth catalog.
- The addition of a fourth catalog and a third email.
- The addition of a fourth catalog and a third and fourth email.
- The addition of a fourth catalog and a retail tie-in.
Therefore, it is important to employ judgment when determining which treatment permutations are most likely to be successful.
Third, measure the effectiveness of each seasonal test treatment as well as the control strategy. This is done as follows:
- Calculate total revenue across the season.
- Subtract the operational costs of sourcing, handling and shipping the merchandise associated with the revenue.
- Subtract total promotional costs. The result is Total Contribution
- Normalize Total Contribution as necessary, such as per-starting-customer.
The panel with the most favorable Normalized Contribution is the winner and will act as the Control going forward.
Fourth, focus in future seasons on fine-tuning the Control by methodically altering the inputs that define the seasonal contact strategy: message, promotion, timing, audience and personalized content. It is here that additional previously-untested permutations can be examined. In this way, constant progress can be made towards Seasonal Campaign Optimization.
When analyzing customer behavior, it is essential to understand the sequence of promotional events that led to that behavior, and not attribute it to a single promotion. Evolving towards the right sequence is an optimization problem that requires the intense interaction of marketing professionals with the decision scientists who provide the quantitatively-derived insights to drive the process. Reasonable seasonal test treatments must be developed with differential levels of marketing intensity. The most profitable treatment acts as the Control going forward, and then is refined over time in a constant evolution towards Seasonal Campaign Optimization.