Initial Setup and Calculation

Define Channels

Define “channels” for grouping your marketing activity into. Each channel should:

• target substantially different audience,
• target audience with consistent propensity to buy,
• represent comparable amount of spend and/or audience. As a rule of thumb, spend should not vary between channels by more than say 10 times.

Channels do not necessarily correspond to different media types. For example, it could be:

• Facebook
• Digital programmatic
• Paid search
• TV based on reach-and-frequency
• Thematic programs on TV and channels on Youtube

It is expected that promoting in thematic context would deliver audience with higher propensity to buy, at a premium CPM. To answer whether this premium is worth paying you to split out this activity from the rest of promotion into a separate marketing input.

This is the initial definition which you can change in the course of the analysis.

InContekst software helps you collate all the necessary marketing data and easily group it into channels.

Select Metrics and Grain

Decide which metrics, and at what level of grain, to use for marketing inputs and outcomes. To do this, run separate test campaigns on key channels or re-analyse previously run single-channel campaigns. The objectives are:

• to study the effects of the campaigns on various metrics – sales, response, social media, website traffic.
• to determine how quickly the effects of marketing fade.
• to identify metrics that respond to marketing quickly yet are predictive of future sales.
• to identify grain at which to monitor data.
• to calculate carry-over. This parameter tells how much of one activity in one time grain, e.g., week, still affects the target metric in the subsequent one.

The narrower the data grain, the quicker and the more accurate results you can get. Typically, sales revenue as a function of marketing and advertising is monitored at the weekly level, requiring a few months’ worth of data for analysis. But if you use daily grain, perhaps with response metric, you can get the same number of datapoints within a fortnight. Similarly, you can reduce the delay in analysis if you can link both marketing inputs and outcomes to geographical areas.

InContekst software makes it easy to put inputs and outcomes metric side-by-side, find correlation between them, experiment with different scenarios, calculate carry-over.

Run Initial Estimation

Perform initial effectiveness estimation based on selected metrics and data granularity. List the factors that could affect the target metric in the specified period and run Model Calculator.

Within InContekst software, this means picking and choosing desired metrics from its extensive library, then pressing a button and waiting a few minutes. The system reports on:

• Outcomes attributable to marketing inputs. For example, how much sales occurred due to advertising activity vs. those that would have occurred anyway (baseline).
• Full and marginal effectiveness for each channel. For example, if the marketing outcome is sales revenue and marketing inputs are spent by channel, effectiveness is equivalent to ROMI (Return on Marketing Investment). If the outcome is the number of units sold and inputs are the number of impressions, effectiveness is measured in the number of additional units sold per thousand of impressions.
• If a channel is under- or over-invested. A channel has a limited audience, and it is possible to use it so much, that at the margin you will spend more to generate a sale than the sale is worth to you. On the other hand, it is also possible to under-invest in a channel so that your message simply goes unnoticed.
• Model error: This is the combination of factors that have not been specified, for which data has not been collected, or simply random factors that are always present in life and we cannot account for.

« Previous: Basic Principles

Next: Optimise Marketing Mix »