Matchbacks for e-commerce, email, direct mail, retail marketing campaigns.

By Susan Pizzano

Matchbacks were once more science than art. When customers called or mailed in their orders, or bought in stores, accurately matching back sales was simple. Now the process requires you to blend technology and marketing for statistical validity.

Because marketers want to create more effective and efficient marketing plans with limited budgets, and ensure high degrees of confidence before implementing costly testing and roll-out strategies, they must be sure to do the following:

allocate demand to appropriate campaign/promotional effort/marketing strategy (e.g., catalog/phone, paid and organic search, affiliates, email, in-store, bouncebacks, Facebook, Twitter, mobile, other nonmail);
allocate accurate marketing costs to each order, since variable costs continue to rise; and
show return on investment accurately by campaign/promotional effort/marketing strategy.

The matchback process seeks to understand customers' paths to purchase by source and channel. In order to be most effective, the process must do the following:

1. Determine which orders that don't match source codes should be included in matchbacks, namely: retail, internet, email, paid search, mobile and affiliate orders.

2. Create business rules that allow the allocation of credit for a sale among multiple contact vehicles as appropriate to understanding the number and mix of contact. For this, do the following:

determine if you want to match email orders back to a catalog or to the email that triggered the orders;
determine how much you want to allocate to supporting mediums (e.g., blogs, Facebook fan clubs); and
tailor the process to the specific brands' business needs and practices, with each brand having unique characteristics (contact strategy, merchandising and business objectives).

The explosion of media and contact methodologies today makes it essential that marketers understand the interaction of all consumer touchpoints and each one's role in delivering the final sale.

One retailer developed a Facebook page to determine its effectiveness as a store driver. It used an email campaign to send recipients to the Facebook page, then offered promotions on Facebook. It continued to email the customer list as well. Of the customers who received the email invitation to Facebook, 32 percent signed up as fans.

Using these matches, the retailer hit these names against its system to see if they'd purchased in a specific time frame. This helped determine if the Facebook page had enticed them into stores to buy. These names were layered into the retailer's matchback for the dates of the promotions offered on Facebook as a separate key code. In a three-month period, the Facebook names performed at a 50 percent greater rate than the email-alone names.

Best practices for performing matchbacks include the following:

1. Be obsessed about capturing source codes on the front end to minimize your unallocated pool.

2. Make sure all customer service reps understand that need, and are trained to ask for the code whenever an order is placed.

3. Make it easy to enter the code at checkout on the internet.

4. Include all orders from all channels in the business rules; offers and order channels are interrelated.

Use all of the data available to build business rules for the process:

promotion/contact history (mail/email/Facebook visits/contact file);
on- and offline circulation plans;
responder file (transaction data);
nonmailed source codes (package inserts, paid search, affiliates, bouncebacks); and
campaign metrics (order curves).

Validate logic quarterly using percentage thresholds to identify when the rules need deeper investigation. Perform hygiene on contact file(s) and transaction data before the matchback process begins for accuracy. This will increase matches.

Consider item-level codes to be able to identify results at the SKU level when source codes aren't available to determine the best contact when more than one contact vehicle is live. Include all active contact vehicles in the process. Use order curves to determine appropriate time frames for inclusion.

The methodology will not be the same for all marketers. Business goals and marketing objectives will shape the process. What will most likely be true for all marketers, however, is that some sort of fractional allocation will be needed.

Customers are exposed to multiple contacts and potentially influenced by each. Each company must determine weighted, channel-specific business rules based on the pool of potential matches, all open campaigns, the value of the marketing program to the business and the reach of the campaign.

The most straightforward method is matching back names to specific programs without considering the contact vehicles' overlap. Or you can base it on the life of the marketing program, creating a weighted average to allocate each week's results.

Determine all promotions that may have triggered a purchase, looking at the order curve for each. Include all contact vehicles that are within 90 percent of projected completion when identifying relevant consumer touchpoints. Plot the order receipt dates against order curves. Calculate allocation percentages based on completion percentages to create the weighted averages.

The averages must be recalculated each week based on the live contact vehicles and order curve status. This method recognizes that multiple touchpoints can affect a given purchase.

Susan Pizzano is executive vice president of Marketsmith Inc., an integrated strategic marketing firm (spizzano@marketsmithinc.com).