Technical challenge: building a multi-channel distribution strategy
As a result of the BMW Group's digital makeover with Elastic, the customer buying experience has been enhanced, helping the company drive revenue across the globe. Sales representatives in numerous markets are now able to advise customers and sell vehicles from any location. In addition, they can configure vehicles together with the customer using a virtually shared screen. And sales reps can also give shoppers a virtual tour of the vehicles they have in stock. The automaker also moved the sales process for retail partners largely online via high-traffic websites in key markets.
Before moving to Elastic, there were traditional stock keeping, production pipeline management, and goods management systems keeping track of the raw stock vehicle data. In addition, the BMW Group also had marketing and product information management systems.
The technical challenge was to combine the data of these systems to produce a centralized repository with a search engine on top of it in order to drive a true multi-channel distribution strategy. Without the centralized repository, a lot of manual work was being performed by the dealers and the sales teams to publish the stock vehicles to third-party portals that in fact are in competition to BMW Group’s own sales channels.
Solution: a centralized vehicle search engine powered by Elastic
In response, BMW Group IT built a central stock vehicle repository, with Elasticsearch as the backbone. The Vehicle Search Service indexes all stock vehicles available for online sales, making it easily available for consumers and sales representatives.
Elastic is one of the few products offering weighted searches that are enabling a true similarity search.
Now, when a specific vehicle configuration is searched, similar vehicles with similar configuration options are also shown in the search results. For example, if the customer searches for a black car, the results will not only show vehicles with a color code belonging to the color cluster black, but will also render a white car if the configuration is similar to the search parameters specified. Of course, the white car will not be shown at the top of the search results because it was scored lower.
The scoring of the results can be influenced by setting specific parameters in the search request. This can be very helpful in case a specific result scoring is needed for different sales channels, markets, or audiences.
Another great feature is geographical search. This enables the automotive manufacturer to not only find a good fit for the customer, but a geographically close one if they know the customer's location. Due to the heavy increase of mobile phone usage with GPS on board, this is often a given.
Elasticsearch also scales very well. Being the central repository for a corporation the size of the BMW Group means that millions of vehicles are indexed in Elasticsearch. Vehicles can also be stored in separate indexes per country, which minimizes load for later replication.
Shifting gears to cross cluster replication
In order to guarantee top notch performance, the endpoints must be relatively close to the end customers, meaning that the search has to be available at multiple geographical sites over the globe. This can be done with the Elastic feature of cross cluster replication.
Remote clusters can subscribe to dedicated indices in another cluster and can easily obtain the data from it. In BMW Group's setup, there is a central cluster where all the data processing and enrichment takes place. Once new data is available, the remote clusters get a copy in near real time and can serve customers locally.
With Elastic's cross cluster replication, sales representatives and customers anywhere in the world are now able to find a suitable vehicle more comfortably.
Gauging and managing success with Kibana
Kibana is used as a powerful front-end to browse the data, to execute queries, to manage the environment, to extract reports, and to perform a whole bunch of additional features. It is even possible to train a machine learning model to detect anomalies while data is being processed.
Now, Elastic is used to monitor the data quality of the created stock vehicle documents. BMW Group is able to get a quick overview of data quality in each market, as well as view metrics about missing or corrupt data that prevents vehicles from being displayed on consumer websites and other interfaces.
This dashboard makes use of the same queries and aggregations that the consuming application would use. Elasticsearch's performance allows the dashboard to execute the queries on-the-fly. The contents of the dashboard are in sync with the customers' view.
In addition to all of these valuable features, Elastic provides BMW Group with professional support. Because Elastic is so widely adopted in mission-critical environments, if there are critical defects, they will be fixed promptly.