Know-How Exchange : Research/Metrics

Hi Steve,

in this world of multiple (offline and online) touchpoints, i am mostly looking for tools that help me correctly attribute results to specific campaigns, accounting for indirect and feedback effects. For one, paid search research has “taken the advertising out” of paid search advertising – it is treated as a direct marketing channel and not as advertising in the traditional sense. Thus, paid search is investigated with respect to direct clicks and conversions only. The assumption is that if a click can not be linked directly to a sale it is considered as not generating revenue.

Typical paid search data, e.g., Google’s AdWords, contains daily information on the keyword-level. For example, say a firm bidding on the keyword “Widgets” has generated 100 clicks (with a cost of $50) and 10 sales today. These 100 clicks are framed as a cost of $50 and compared with all sales (i.e., 10) that this keyword has generated today. Thus, the generally accepted performance metric cost-per-sale for this keyword is $5.

This approach to performance measurement does not account for a future sale tMoreover, it does not give credit to all the other marketing actions that went into the consumer's decision to click through. i have quantified such indirect and long-term effects by vector-autoregressive (VAR) modeling, as recently applied online, eg to compare the net impact of WOM marketing versus events and PR for a social media site (Trusov, Bucklin and Pauwels Journal of Marketing 2008).

Steve -

There is a lot of discussion about these questions right now, and not surprisingly, there are differences in terms of industry, sales force size, and unique competitive situation. That said, there are two resources that I think would provide you with some insight, not just in terms of CPG as it relates to resource allocation, spending and ROI, but across different industries and customer type (B2B vs. B2C).

One is https://www.marketingnpv.com/, which not only highlights some current thinking on different models but also may provide you with some useful tools (or, at least, the key aspects used in these tools). For example, one technique discussed is the response-curve approach.

Another resource is Admap - https://www.admapmagazine.com/index.asp. You can access back issues on their website.

Hope this helps - good luck!

I'll add a three points if I may

1) Re econometrics, I’d suggest that use of econometrics is evolving to be far more responsive to short term needs of brand managers / planners meaning an increase in interpretive tools and a reduction in the requirement for specialist analysts input - i.e. automated or semi-automated modelling which better reflects the short term realities of the business concerned. Furthermore the only real value in econometrics is as an input to other processes so in terms of what's new, I'd suggest planning and scenario tools that can immediately take econometric (and other) inputs and may be used by non-specialists for planning.

2) Agent based models that simulate the behaviour of individuals within a population rather than treating all customers as being homogenous and alike. I'm very interested in this branch of analysis since it appears to offer a significant leap in terms of insight when compared to methods such as econometrics. It's something that I know a few people have experimented with (such as PepsiCo) to answer questions such as "how many vending machines do we need and where do we site them".

3) Risk based analysis - building on statistical modelling techniques (including stochastic simulation techniques), actuarial techniques offer marketers a greater insight into the likely outcome of plans and enable trade-off decisions which aren't purely based on ROI but also take into account the broader objectives of the firm
Hope these are interesting insights,
John