DOOH Client-Side vs Server-Side Measurement

This article is the fourth in a series on DOOH Audience Impressions which explores what the factors are that can accelerate DOOH towards the “holy grail” of cross-platform media compatibility. In it, we will discuss a robust protocol for client-side ad measurement and its benefits and uses. We will also contrast it with server-side techniques. This article is also an excerpt from a larger work, Quividi’s DOOH Audience Impressions White Paper which can be downloaded here.

Client-Side Measurement

A key aspect of Quividi’s measurement mechanism which makes it cross-platform is its local APIs for integration with the content management system which is playing the ads. This gives Quividi the ability to measure audience impressions for a specific ad as it is playing and also fulfills the requirement from online that ad measurement be performed locally.

Local CMS Integration

Quividi’s measurement system, when integrated with a 3rd Party CMS, is able to scope the audience measurement to the airtime of a single ad play with sub-second granularity. The protocol of live data exchange between the CMS and Quividi that permits this is illustrated below.

The protocol of live data exchange between the DOOH CMS Player and Quividi VidiReports.If the CMS supports receiving audience reporting (step 3), its reporting can later be verified against Quividi’s.

Content Intelligence

A key advantage of CMS integration is access to audience notice metrics such as attention. This permits marketers to have unprecedented levels of content intelligence data. Without live data from the DOOH player, an audience measurement solution could only provide screen-level analytics, where content analytics would only be available by averages.

Marketers that have access to content-level analytics data for their content have a distinct advantage over those with screen-level data, due to the fact they can see changes in audience dynamics like second-by-second attention drop-off. This isn’t even possible online.


Here we illustrate how Quividi helped Aldo see the different performances of ad creative executions compared to their targeted audience.

Live API for Programmatic (Pre-Bid)

Another important client-side usage of audience impressions is for programmatic real-time bidding (RTB). Programmatic SSPs can also make use of Quividi audience data in the construction of their RTB bid requests. This is provided via a dedicated API which is built around the dynamics of DOOH programmatic.


The protocol of live data exchange between the DOOH SSP and Quividi VidiReports. the DOOH SSP uses the audience report from Quividi to build its bid request for programmatic RTB auction.

For example, most RTB auctions are for the current ad slot, but in DOOH it may be auctioning the next slot in the playback sequence. Quividi’s API allows for these types of near-time queries. The pre-bid audience report is generated using a local database of current and historical data and is far more dynamic (sub-second granularity) than server-side equivalents (hourly granularity).

 

Server-Side Measurement

Not all CMSs and SSPs have integrated with Quividi’s client-side measurement APIs and instead they employ a server-side approach to audience reporting. We explore these approaches in this section.

Server-Side Data Stitching

Data stitching involves matching the raw playlog data from the CMS and raw audience data by times and locations. This approach is not employed by Quividi due to its challenges (huge server computing power required and inaccuracies in the output).

Audience Export For Programmatic (AUA)

The preferred technique of server-side audience measurement is to use historical averages of a location across a time period as an “audience impression multiplier” on the served impression.

The Average Unit Audience (AUA) Export is provided on a per location, per week-hour basis and it is further segmented into genders (male, female) and each gender into age groups(<17, 18-24, 25-34, 35-44, 45-54, 55-64, 65+). It depends on the content duration.

In the above example, a 15 second advertising running at 10AM on screen 48338 will likely generate 2.314 impressions when it will run. This number can further be broken down by gender and age group.

Sampling and Server-Side Extrapolation

With a sufficient sample size and proper random selection methods, measured locations’ data can be extrapolated to non-measured locations to give them audience coverage as well. The extrapolated locations audience are modeled after the averages of the sampling locations. This approach, with the right sampling size, guarantees that the extrapolated results for each screen will be +/-20% of reality in 95% of the cases. For more information on Quividi’s own sampling and extrapolation offering, click here.

Extrapolation is often used in post-export pre-import data engineering to fill in missing data such as for new deployments. 

 

Conclusion

In this article, we illustrated a robust client-side ad measurement protocol which has the side benefit of providing rich content intelligence to marketers. We demonstrated how server-side methods have many disadvantages compared to Quividi’s client-side approach. We also discussed how Quividi has programmatic support both client-side with the Live API (pre-bid) and on the server-side with the AUA Export.

EnglishChina