Lost in Translation: Bridging the Chasm Between ‘Search Engine’ and ‘Marketing’ Through Entity Ads


EDITOR’S NOTE: This article, originally published on June 13, 2013, placed at No. 14 in our 20 most popular articles of the year.

ADOTAS — In the world of search engine marketing, there is ironically a large chasm between “search engine” and “marketing.”

A search engine marketer’s (SEMs) day-to-day activities, including keyword selection, organizing keywords into ad groups, etc., have little in common with traditional marketing activities like audience segmentation, messaging, pricing and promotions. As a result, paid search has become the domain of specialists who spend their time translating between search engine jargon and their business. This divide represents economic friction that is depressing the overall paid search market.

The good news is that search is undergoing a massive paradigm shift, moving away from documents connected by hyperlinks and toward knowledge organized around entities and linked by a graph, like Google’s Knowledge Graph or Microsoft’s Satori. In a nutshell, entities make connections between the relationships and concepts that searches are based on, making it easier to categorize, process and, ultimately, provide more accurate results.

Entities benefit consumers because they improve the relevance of both algorithmic and paid search results.  But entities also solve a critical marketer pain point:  the friction that comes from having to continuously translate between publisher constructs and marketer entities. Today, SEMs must use publisher constructs — like keywords, match types, and product targets — to express their paid search campaigns.  If SEMs could use their own entities — the very products they sell, organized in their own taxonomy — massive amounts of friction could be eliminated and the paid search market could grow substantially.

In other words, entity ads are the future of paid search (and all intent-based media).

The Marketer’s Translation Dilemma

The average marketer has an assortment of products (and/or services) they want to sell to consumers.  In order to advertise to consumers via paid search, they regularly encounter two major areas of friction (see Figure 1, below).

First, they must translate the products they sell into publisher (i.e., search engine) constructs:  keywords, match types, ad groups, ads, campaigns, bids, budgets, product targets, etc.  This process creates a lot of friction that is holding the paid search industry back. Way back. Here’s how:

  • It’s is a huge amount of work. For even medium-sized marketers, it requires a team of SEMs, an SEM agency, or both.
  • There’s no synergy between search “surfaces.” For example, constructing a keyword-based ad campaign is an entirely independent effort from constructing a product ads campaign.
  • It’s getting more complicated every day. The number of search “surfaces” continue to multiply with the growing number of device types you can search on (computer, phone, tablet, Google Glass, etc.) combined with search interfaces (typed query, voice, image, location, gesture, etc.) and search-equivalent behaviors.
  • It requires a great deal of highly specialized knowledge about publisher constructs. SEMs need to understand what quality score is, how to best construct ad groups, when to have overlapping match types, etc., etc.
  • It’s lossy. Because there are an infinite number of queries, even experienced SEMs miss valuable keywords/queries.  Missing keywords are bad for marketers, consumers, and the search engine combined.
  • The translation from marketer entities to search engine constructs is too ambiguous. For example, should a marketer’s product category map to an ad group, a campaign, or to an entire account?  The frustrating but correct answer among experienced SEMs is, “it depends.”
  • It’s error prone. Too often, the wrong keyword, match type, ad copy, URL, or bid is deployed, causing either a dive in quality score, profitability, or both.

The Marketer’s Reverse-Translation Problem

If one translation problem isn’t enough, marketers also face a reverse-translation problem.

Marketers have to reverse-translate the performance of their paid search campaigns back to their own business entities, because:

  • A lot of marketing budget depends on manufacturer co-op dollars. If they can’t tie their campaigns back to manufacturer’s brands, the co-op dollars stop.
  • Others in the organization aren’t familiar with paid search. Marketers need to communicate with other departments and upper management about the impact paid search is having on specific lines of business, departments, and product lines. When this communication is absent, paid search budgets wallow.
  • They need to continually optimize their campaigns. Marketers need to make on-the-fly reverse translations to understand the impact of their campaign optimizations and improve upon them. For example, how did a set of bid changes affect the gross margin across a product family? Did a set of negatives have the desired impact across a set of products?

Because paid search campaigns are created using search engine constructs, campaign performance reports typically have almost no relationship to the marketer’s business language or taxonomy.  As a result, SEMs have to spend countless hours massaging performance data to make it comprehensible to the rest of the organization.

Paid Search, the Entity Ads Way

Enabling paid search campaigns to be constructed using “marketing speak” would eliminate a huge amount of friction between the marketing and SEM teams. (See Figure 2.)  Here’s how:

  • There would be no need to translate marketer entities into publisher constructs.  The marketer would primarily use their own entities (i.e., their products) to specify their paid search campaigns.
  • There would be no need to reverse-translate paid search performance from publisher constructs back to the marketer’s entities. Paid search performance data, already in the marketer’s language, would immediately be more understandable and essential to the rest of the organization. Imagine being able to see how a family of business is performing all the way down to the individual SKU – and at every level in between.
  • There would be a ton of synergy between search surfaces. Marketers could take a “write once, deploy everywhere” approach to all forms of paid search advertising.
  • Less specialized knowledge about publisher constructs would be required.
  • There would be no ambiguity or loss associated with the multiple translations. Marketers would know precisely which products need to be sold.
  • It would much less work and be far less error-prone.
  • Marketers and SEMs would spend more time being marketers, not technicians.

With this much friction eliminated from the marketplace, both the cost of paid search marketing would decrease substantially and the ROAS would go up accordingly.  As such, the paid search market would grow significantly.

In other words, the micro-economic forces driving the paid search industry towards marketers’ entity ads model are overwhelmingly compelling. By switching to entity based ads, paid search has the possibility of reaching its full potential and the chasm between “search engine” and “marketing” can be bridged.

Entity Ads:  Hard Problems Need to be Solved

If the economics for entity ads are so compelling, why isn’t there already widespread adoption of entity ads?

The fact of the matter is, there are several hard problems that need to be solved before marketers can use their own entities to specify and manage paid search campaigns.  Specifically:

  1. Entities must be normalized across marketers. Assume two merchants sell the exact same product, Widget A. An entity ads system needs to have normalized entities across merchants, i.e., the system should only contain one entity that represents Widget A.  If the system has multiple entities representing Widget A (e.g., “Widget-A”, “Widget a”, and “Widget-a”), this would be confusing for all merchants on the system. Do “Widget A” and “Widget-A” compete for the same traffic? Should I choose one over the other?  Or should I sell both? Normalized entities reduce friction; conversely, non-normalized entities increase friction.
  2. Entities must be semantically connected via a graph of attributes. For example, if a consumer searches for “blue shoes,” a marketer’s entities probably need to be tagged with two2 attributes, “product type=shoe” and “color=blue” so search engines know which products are eligible for an impression/click. It must know that shoes are a type of apparel (parent), and that there many types of shoes (children). Additionally, equivalences need to be specified in the graph as well.  “Navy” and “azure” should be related to “blue” so that navy and azure shoes are also eligible for an impression/click after a “blue shoes” query.
  3. Consumer intents must be part of the graph. Consumers often issue queries that contain words that are often not found in standard product descriptions and attributes, e.g., “cheap,” “small,” and “cool.”  Behind each of these words is a consumer intent, and these intents can differ across categories. For example, in the queries “cheap digital camera,” “cheap 60″ LCD TV,” and “cheap Louis Vuitton handbag” cheap can mean very different things.

These are difficult, but not impossible, problems to solve.  The party that best solves the entity ads problem will deliver greater value to marketers and reap the associated rewards.

In other words, the entity ads race is on.


Please enter your comment!
Please enter your name here