A Brilliant Deduction

By Carl Langrock

In solving a problem of this sort, the grand thing is to be able to reason backwards. That is a very useful accomplishment, and a very easy one, but people do not practice it much… When you have eliminated all which is impossible, then whatever remains, however improbable, must be the truth.-Sir Arthur Conan Doyle’s Sherlock Holmes

It’s a key process in the Holmes method, but what actually is deductive reasoning? Let’s skip the formal definition and refer back to the Holmes quotation. The “deduct” part of the word deduction implies the answer. If you subtract all of the information and noise that are not the answer, then what you have left is the answer. It’s not hard; we do it all the time. For example, how do chemists measure the weight of a lighter-than-air gas? One of the methods is deductive measurement. For a trained chemist it’s simple, just as managing a complex direct response buy is simple for a seasoned media buyer.

In March, I introduced a process for tying web performance back to a DRTV schedule. This month I continue on the topic, but highlight a different means to the same end.

DIRECT PARTNERS AND PEOPLE PC
John Walker, senior vice president and media director at Omnicom’s Direct Partners unit, uses a process of continual refinement in creating the right media balance for EarthLink/People PC:

The analytic approach for EarthLink/People PC and our Direct Partners media team has evolved significantly since August 2003, when we started running national cable DRTV in volume. Early on, we were looking at our schedules and results in a way that’s fairly standardized for a DR campaign. We ran specific creatives with specific 800 numbers tied to specific stations. That allowed us to look at the number of calls that we had and to use our media analytics system to help us tie call activity back to the spend. We buy-managed by cost per call for a few months. We established very quickly that the value of the broadcast spending was clear; it was effective enough to stand on its own. That was what I consider the first step on the ladder.

But, as with any campaign, there are all sorts of other factors. When are the spots airing? Is the call center managed properly? Are the operators competent or are they just name takers? Is it a one-step, two-step, or even a three-step sale? We were adding match-backs with direct mail, all sorts of things to further refine the buying process and analyze our results.

A leap forward came when, after being on-air for a significant period, we married our data with the client’s customer database. We then evolved to the point where we could track our advertising to the cost per sale and add in lifetime value-based histories of customers acquired through DRTV. Over time, as we continued on-air, we could go back and see how many sales we captured on a specific station during a specific period. We could see how many of the station’s sales were still customers and we found that certain stations tend to have a more loyal base. When you look at lifetime value, their calls were actually worth more than callers to other stations. A station with a higher initial volume and a lower initial cost per call might actually be less efficient than a station that, for whatever reason, has a customer base that tends to subscribe to People PC for a longer period.

This was the second step: evolving to a point where we could manage our buy based on customer lifetime value. It was like the breakthrough when medicine moved from the two-dimensional view of an X-ray, to the three-dimensional MRI-merging the data made a huge difference. Working with the client to generate this three-dimensional view helped us achieve a very low cost-per-sale plateau that we’ve maintained consistently. aFrom that point, the only thing that wasn’t fully nailed down was tying web sales directly back to the broadcast. The problem is, we can’t corral web response in the same way that we can the phone calls. With telephone calls, there are clear bursts related to the spot airing. But when it comes to web response, our volume of spots is so constant, you can’t see the correlation at a detail level. But we knew that there was an effect.

With People PC, we’ve been on the air consistently for more than a thousand days. We know the media activity and the number of web sign-ups for each of those days. By looking at the data for web activity when we were on hiatus, or when spending was down versus when spending was up, we found a change in the tide, if you will. DR broadcast increases web responsiveness. We saw this repeatedly over time. So, even though we haven’t yet been able to develop a response mechanism that allows us to say, ‘OK, we know that these web sales for this period of time came directly from our advertising on the following stations,’ we have a volume of evidence that the tide of people signing up for service over the web rises with our television advertising.

At a high level, the correlation is direct and clear: as TV spending increases, web responsiveness for People PC increases at a corresponding rate. For example, if we spend 50 percent more in a week than we did in the week prior, we see that web activity increases at almost the same pace. By working with the client and sharing results and customer data, we came up with formulas for spending optimization and for ascribing a percentage of web activity back to the overall DRTV broadcast.

At one point in time, we would assign web response back to specific stations based on relative audience or station schedule percentage. Ultimately, we found that level of detailed ascription to be more cumbersome than helpful. Now we spread the web response evenly across the TV buy, reducing the overall CPL for television. Using this approach, we’ve been able to reduce the overall acquisition cost of television well below the allowable (and very lean) threshold. And the call volume is there. But the evidence is clear that the DRTV advertising is driving customers to People PC’s website.

When the broadcast is cost efficient on that first level-the “call now” level-then it makes the client more comfortable with the idea that the spot is pulling in web activity, as well. But the power of the attraction has a subjective element. So we use the data to the best of our ability to deduce rules of effectiveness. The broadcast buys need to be able to stand on their own, without any web response credit. But we know that our ascription methodology is directionally strong enough because of the quality and stability of the acquisition results.

Everyone knows that ocean tides are affected by the gravitational pull of the moon. But how strong is the effect? The moon is always there. If we could switch off the moon, we would know the effect of the lunar gravity by witnessing the effect of its absence.

Direct Partners and People PC have a proven linkage between their DRTV advertising and web sales. But its high and constant volume of activity obscures the unit-level relationship between a specific spot airing and a web response. How do they know exactly what the effect of DRTV advertising spending on web performance is? By combining their media, response and customer life-cycle data, Direct Partners and People PC are able to deduce the right conclusion. They can factor their total cost of sale into managing their total media schedule. Holmes would be proud-a brilliant deduction!

Next month, I’ll continue this exploration of tying web performance back to the broadcast buy. We’ll train our magnifying glass on a different solution.

Carl Langrock is president of COREMedia Systems in Fairfield, N.J. He can be reached via e-mail at [email protected].