It’s a question as old as Adwords: is my branded search spend truly incremental? Branded search ads are those that are attached to brand-name keywords, like “Tide” instead of “detergent”. Ads that show when someone searches for a brand name tend to (at least seem to) perform very well, but it stands to reason that people searching for the brand itself may have already been intending to purchase that brand, and the ad at the top of the screen was simply the easiest way to get there.
It gets worse: for an advertiser to spend on branded keywords, consumers have to first search for their brand on Google (a similar dynamic is at play on other search engines, and in app stores).
That means that the amount you spend is not directly under your control — in order to spend more, you first have to induce people to search for your brand. Even if you could nail down the incrementality of the channel, recommendations to spend more are helpful only to a very limited extent, because they don’t tell you how to spend more.
Commission-based affiliate channels are an even more vexing challenge. In this case, there is no doubt: the spend itself is not incremental, because it comes after the sale.
In fact, it is the sale that causes the spend, not vice versa. It is the affiliate’s other activities: their content, their SEO, their trust in the audience, etc., that is incremental. Simply including affiliate spend in any regression based approach, without sophisticated adjustments, will give a misleading picture.
Recast solves these challenges with a concept we call “lower funnel” channels. Lower funnel channels are those channels whose spend is caused by other channels (which we call “upper funnel”), and therefore whose budget is not directly under the control of the marketing team. For example, a TV ad may lead someone to search for the brand on Google. They may then click on the ad — leading to increased branded search spend — or they may click on the affiliate link a few places below — increasing affiliate spend. The branded search ad or the affiliate may yet have had a truly incremental effect; it’s possible that the TV-watcher would not have bought the product had it not been for either of those two placements. Perhaps, for example, a competitor was bidding on the branded term, and the ad recaptured a potential customer that was on the cusp of being lost.
In this case, the TV spend is “upper funnel” and the branded search and affiliate spend is “lower funnel”. Recast handles these kinds of interactions by modeling them directly. For example, if we imagine that every dollar of TV spend leads to one dollar of branded search spend and one dollar of sales, and every dollar of branded search spend leads to two dollars of sales, a naive model that doesn’t account for the induced spend will understate TV’s ROI by 25% and overstate branded search’s ROI by 50%. Recast solves this by using a sophisticated statistical method that simultaneously estimates both the relationship between TV and branded search spend (more TV spend causes more branded search spend), the relationship between TV and sales directly, and the relationship between branded search and sales directly. The Recast platform can then calculate both the direct and indirect effects of TV on sales.
When you upload a budget into Recast’s forecasting tool, it’s not just predicting how much revenue you’ll earn, but also how much you’ll spend on branded search. (In fact, it’s first predicting how much branded search spend there will be, and then how much revenue you’ll earn, because that depends on the effect of the branded search spend). We have a strong perspective about how to properly handle these types of channels when validating the performance of your model.
Modeling the dependence of lower funnel spend on upper funnel activity helps Recast properly model — and account for — the effect of all channels. In Recast, an extra dollar spent on TV will: (1) lead to increased sales directly; (2) lead to increased spend on branded search, which (3) leads to increased sales indirectly.
The total ROI of TV spend should therefore include both the direct and indirect sales, but also the direct and indirect spend. All of this is explicitly accounted for in the Recast platform.
