What Are the Use Cases for Log-Level Data?

The programmatic black box has opened. Ad platforms are beginning to make their log-level data available in the hopes of satiating industry hunger for transparency.

Once considered the “data exhaust” of the industry, log-level data (LLD) has been positioned more recently as a remedy for many of programmatic’s woes.

LLD contains many valuable, impression-level details like geo data, URLs, time stamps, brand safety data, and auction mechanic information that allow buyers to inform bidding practices and prune supply paths.

But that’s not all. There are a number of powerful use cases for LLD that can create value for all parts of the programmatic supply chain—not just buyers. Here are a variety of ways LLD can be used to bring greater transparency, efficacy, and fairness to programmatic advertising.

For the Buyer

Advertisers want access to log-level data for a number of reasons. At a high level, access to LLD increases advertiser visibility into auction mechanics, inventory quality, the competitive landscape, and the users interacting with their ads. A few major buyer use cases are described below in greater detail.

Inform Bidding Strategies

Advertisers want to know why they won (or lost) a programmatic auction. Log-level data can offer insights into the prices at which auctions are being won, as well as the prices other participants are bidding, allowing advertisers to instruct their DSPs to adjust accordingly.

Understanding what bidding strategies, targeting parameters, and price plans resulted in an auction win can help advertisers repeat their steps to success. By the same token, LLD also contains the same details for any auctions lost, allowing advertisers to avoid bidding in the same way in the future.

Relevant log-level data: bid price, floor price, clearing price

Prune Supply Paths

Log-level data can be used to reveal how competitive the bidding ecosystem is so advertisers don’t pay more than they need to (which is especially important with the increasing popularity of the 1st price auction).

LLD can also be used to better understand the performance of various supply sources, allowing advertisers to tailor who they work with based on both price and performance. While fees may play a part, the value a tech vendor can offer with quality inventory, campaign performance, and ultimately return on investment should be determining factors in supply path optimization.

LLD also grants visibility into signals like ads.txt, allowing advertisers to know whether or not they’re buying directly from a seller or indirectly from a reseller.

Relevant log level data: ads.txt, sellers.json, auction type

Confirm Campaign Data (in Real Time)

Unfortunately, advertisers deal with a good deal of inconsistency in how their campaigns are measured and reported. Log level data can act as proof points or “confirming signals” for important campaign performance metrics.

For example, advertisers often deal with discrepant impression counts between their ad server and their DSP. Using log-level data, correlating data points can be matched across the supply chain to determine which impressions successfully delivered and which didn’t. This can be acted upon in real-time before a campaign is over.

Confirmed metrics—in this case, confirmed impressions – can then be used as a reliable and accurate metric of record for that campaign.

Relevant log-level data: auction ID, impression ID, timestamp

Uncover Breakages

Perhaps one of the most significant use cases for log-level data is the ability to uncover breakages that otherwise might go undetected.

A lot of technical errors can happen during a programmatic campaign. There might be cases of broken tags, missed software updates, or poorly handled integrations that cause a campaign to break mid-flight.

Without LLD, advertisers might remain blind to the causes of these issues. However, with LLD, advertisers can stitch together a unified view of each and every impression. Discrepancies between a DSP and an exchange—for example—could point to a possible breakage, which would allow advertisers to optimize away from in real-time.

Relevant log-level data: auction ID, impression ID, timestamp

For the Seller

Like advertisers, publishers have good reason to want access to log-level data. At a high level, increased visibility into the programmatic auction can help publishers better understand why certain inventory might be out-performing others, as well as what buyers are paying for it.

Recently, Index Exchange became the first exchange to make their audit logs available at scale.

A few other major seller use cases are described below in greater detail.

Better Understand the Quality of Their Inventory

Just as bid pricing log-level data can be used by the buy side to inform bidding strategies, that same information (alongside brand safety and viewability data) can be used by publishers to better understand how buyers view the quality of their inventory. Which inventory garners the highest bids? Which inventory (if any) achieves a 90%+ fill rate?

Knowing the answer to questions like these can help publishers best optimize their pricing plan and determine which inventory to drive the most traffic towards.

Relevant log-level data: highest bid price, viewability rate

Create Targeting Personas

User information (non-PII) is also contained within log-level data which can allow for better targeting. Publishers can track this data and create a more holistic picture of the people who come to their site. That persona can be given to the advertisers they work with to help inform their targeting strategies and creative optimization.

Relevant log-level data: geo data, user operating system, device type

For the Industry

Regardless of whether or not you’re a buyer or a seller, there are certain use cases for LLD that benefit the entire programmatic supply chain.

Ensure Auction Mechanics are Honored

Programmatic auctions may be automated, but that doesn’t mean they can’t be manipulated. We detailed for AdMonsters a few of the ways in which auction mechanics could be altered without either the buy or sell side knowing. Second-price auctions secretly becoming first-price auctions, for example. Regardless, log-level data can offer transparency for both sides of the exchange to ensure the auction is run as intended.

Relevant log-level data: auction type, timestamp

Standardize Programmatic Measurement

Standardizing programmatic measurement has seemed a bit of a fool’s errand for some time. Organizations like the IAB and MRC have been at the forefront of creating measurement standards for the industry. But with few ways to technologically enforce these standards, many companies are still free to measure what they want how they want.

This is possible in large part because companies have customarily reported campaign data in aggregate. Measuring in the aggregate creates a scenario where numbers can be calculated differently, resulting in widespread discrepancies between data counts.

Access to LLD changes that. As the granular, impression-level information that is gathered while an ad is displayed on screen, LLD can be used to define how common programmatic metrics should be calculated.

For example, what makes an impression “count” as successfully served?

With access to LLD, both buyers and sellers can codify programmatic measurement according to concrete, specific standards that make “fudging” numbers nearly impossible. The result would be a more consistent marketplace for the buying and selling of ads.

Relevant log-level data: any data point can be used to create a standard