Samsung | Case studies

Samsung Ads uses SurrealDB to replace three data stores and cut query time from hours to seconds

Samsung Ads is the advertising arm of Samsung Electronics. It reaches audiences across Samsung Smart TVs and mobile devices, and its campaigns run on the content, user and device data those screens produce. The team builds real-time knowledge graphs on SurrealDB to plan and execute those campaigns.

Challenge

Samsung Ads plans campaigns against one of the largest first-party audiences in consumer electronics: the viewing, app and device behaviour of Samsung Smart TVs and phones. Turning that behaviour into an audience a marketer can target is the whole business, and it depends on asking connected questions of the data quickly.

The data that answered those questions lived in three legacy stores. Customer profiles sat in one system, behaviour in another and content metadata in a third, so a single audience question became a multi-step query stitched across all three. Insight arrived hours after it was asked, and audience targeting was tuned against a picture that was already out of date.

The silos also kept the team from a unified view of any one audience. Segments were coarse because the joins that would have sharpened them were too slow to run at campaign pace, and coarse segments meant ad spend went on impressions that returned little. The analytics platforms on top were rigid as well: as marketing strategies changed, the stack lagged behind them.

Solution

Samsung Ads moved the audience data on to SurrealDB and modelled it as a knowledge graph. Profiles, behaviour and content became nodes and edges in one graph, so the relationships that used to be assembled at query time across three systems are stored once and traversed directly.

Because SurrealDB is one engine for several data models, the same records are also reachable as documents and as key-value lookups. The team used the key-value path for the hot reads that campaign tooling makes constantly, and query times fell from hours to a quarter of a second. Marketers retrieve an insight in the time it takes to ask for it.

On that foundation the team built dynamic audience segmentation and real-time campaign tracking. Segments are recomputed as behaviour arrives rather than on a nightly schedule, and a campaign can be adjusted while it is still running. The three legacy stores and the analytics platforms that read from them were retired in favour of the one SurrealDB platform.

Results

Precise targeting lifts campaign ROI by 25%

Sharper segments put impressions in front of the audiences most likely to respond. Campaign ROI rose by a quarter, which at Samsung Ads' scale saves millions in annual ad spend.

Decisions in seconds, not hours

Queries that took hours across three systems return in around a quarter of a second from one. Campaign adjustments that used to wait for the next day's report now happen in the moment.

One platform, 30% lower operating cost

Retiring the legacy stores and the analytics platforms on top of them cut operational costs by 30%. There is one system to run, one to secure and one to train people on.

A 360-degree view of the customer

With content, user and device data in one graph, the team sees each audience whole. That view is what makes deeper personalisation possible, and it is the basis for the campaigns Samsung Ads runs today.