Hack88.Use Geographic Segmentation to Measure Offline Marketing

Hack 88. Use Geographic Segmentation to Measure Offline Marketing

With the growing adoption of multi-channel retailing strategies, Geo-segmentation can now provide valuable insight into offline advertising effectiveness.

Offline advertising can take many forms, including television commercials, radio ads, billboards, print, direct mail, and even guerilla marketing. Compared to the relative ease of measuring Internet advertising, measuring offline advertising is considerably more difficult. Companies have relied on coupon codes, purchase and brand awareness surveys, dedicated 800 numbers, and other costly tactics. Fortunately, because offline advertising is now directing many audiences online, you can leverage web measurement and geo-segmentation to evaluate the effectiveness these efforts [Hack #78]. Most top measurement vendors provide geo-segmentation tools either as standard or as an add-on package; consult your particular vendor for specifics regarding how to get geo-segmentation added for your analysis.

To begin, you must decide which regions you are advertising in. Then decide which regions will serve as control; that is, regions that you are explicitly not advertising in. The control regions are important, because they will be used to remove bias that may occur in your target regions from other, unrelated marketing activities. Once you pick your targets and controls, establish precampaign baselines for traffic, revenue, leadswhatever key performance indicators best highlight campaign success. Without diving deep into statistics, you should at least draw on as much history as the actual advertising campaign will run. For example, if you are running the ad for two weeks, you should have at least two weeks of historical baselines to reflect on.

Once your promotion launches, be sure to keep a close eye on changes in these KPI baselines. If you identify a poor performer early in the campaign, you can save significant costs and increase ROI by reallocating those funds and resources to better-performing regions or initiatives. Of course the response time will vary greatly depending on which marketing tactic you choose (i.e., a catalog or direct mail piece could take six weeks, while a television or radio ad may occur in the same day), so be sure not to cut the strings too early without understanding the particulars of your market and marketing tactics.

6.9.1. Key Measurements

As with other marketing initiatives, standard measurements like response, conversion, revenue per acquisition, lifetime value, profit per customer, and campaign ROI are all relevant.

However, there are also some unique metrics you need to be aware of:

Response lift

This metric is calculated by taking visits from your target regions during the promotion, and subtracting visits from before the promotion. Then compare your control regions in the same way, such that you have net visits from your target regions and net visits from your control regions. Now subtract the two numbers, and you will be left with the response lift for your regions.

Success lift

Similar to response lift, you simply compare success events (such as revenue) for each target region and each control region. By taking the net difference of the two, you can calculate success lift, which in turn will allow you to determine campaign ROI.

6.9.2. An Example

Assume you run a series of radio advertisements in California in November 2004 that directs listeners to your web site. The primary call to action is submitting a lead form, so the most important KPIs are visitors, leads, and leads per visitor.

You decide to run the ads in Northern Californianamely San Francisco, San Jose, Palo Alto, Sunnyvale, and Santa Clara. Furthermore, you decide to use Los Angeles, San Diego, Orange County, and Santa Monica as control citiesregions where you will specifically not run these ads. The first step is to establish traffic baselines from October 2004 data. (For the sake of simplicity, the example excludes lead and leads per visitor baselines in this hack. In a practical setting, this would naturally be a key focus.)

The next step is to track changes in key KPIs over time in response to the campaign. Figure 6-8 illustrates the highlights, and shows how many of the target regions demonstrated strong visitor growth.

Palo Alto jumped 61.7 percent, Santa Clara was up 61.7 percent, and San Francisco was up 11.2 percent. But look at the control cities, where there was also some strong growth there: Los Angeles jumped 10.7 percent, San Diego jumped 43.4 percent, and Santa Monica jumped 69.7 percent. Perhaps those frigid California winters kept more people inside and on the Web in November. Perhaps a competitor in San Diego went out of business; perhaps the marketing group improved local advertising efforts in Santa Monicathere are numerous possibilities.

If you do not reduce the chance of bias from these possible influences, you could overstate or understate your true advertising effectiveness. This is where your control groups come into play. First, sum up visitor traffic for the target cities in October. In the example, this is 3,387 visitors. Do the same for November; the 4,237 for this period in the example suggests a 25 percent gross increase over October. Then perform the same analysis for the control regions, arriving at 1,854 visitors in October and 2,197 visitors in November. This suggests an 18 percent increase.

Figure 6-8. Custom measurement of offline data

Subtracting the 18 percent control growth from the 25 percent target growth yields a 7 percent net increase in visitor traffic from the radio promotion. In other words, the response lift is 7 percent, and it is reasonable to attribute this to the radio promotion.

Depending on the average visitor value, a 7 percent increase in visitors could translate to significant revenues. For example, assuming the average lead conversion rate is 5 percent and the average deal size is $100,000, this 7 percent increase in visitors would yield over $1 million in campaign-driven revenue.

Matt Belkin and Eric T. Peterson