Revolutionizing Local Advertising: Olimaps Bidding System
In today’s digitally interconnected world, social networks have become the cornerstone of communication, connecting individuals from all corners of the globe. However, as our online connections expand, it becomes increasingly crucial to bridge the gap between the virtual realm and the physical world we inhabit. Enter Olimaps, a groundbreaking social network that revolutionizes the way we engage with our surroundings.
Olimaps is not just another social network; it is a dynamic platform that allows users to explore and share the happenings around them in real-time. What sets it apart is its unique feature that merges the power of social networking with location-based advertising: the Olimaps Bidding System.
This ingenious mechanism enables local advertisers to target their promotions precisely, ensuring maximum reach and impact within a specific geographic region.
In this article, we delve into the intricacies of Olimaps’ bidding system, exploring how it harnesses the power of localized advertising. We will examine the benefits it offers both users and businesses, such as increased engagement, targeted marketing, and the potential for seamless monetization. Furthermore, we will analyze the system’s underlying technology and its role in redefining the advertising landscape.
How bids are assigned
When browsing the map, the user will see posts from other users as well as ads from local advertisers. Our system will therefore need to determine which adverts should be shown and adjust the fee that advertisers ultimately pay to be fair.
Let us take a closer look at how the bidding system works from the advertiser’s point of view:
When a permanent job is created, the advertiser can decide how much they are willing to pay each month.
The higher the maximum amount the advertiser is willing to pay per month, the more visible the ad will be from a higher position. If the ad is placed in a competitive area, the advertiser will have to pay more per month in order for the ad to appear against the competitors.
It is important to understand the relationship between the zoom level used by the end-user and the competitiveness of the area; in a competitive environment, ads that do not pay enough will not appear, but if the user is close enough and zooms in, the ad may have a chance of appearing.
Look at this last image, the €200 ad appears next to another that has paid 50€. It’s an unfair situation for the one who pays 200€, so the final monthly fee should be less. But how much less?
Let’s assume a query where these ads are inside the area:
Let’s say that 4 ads are allowed to appear. When a user makes a query for the area, the 4 with the highest maximum bid will be selected, the user will be shown the ads 1, 2, 3, 4, the bid 4 being the one with the lowest value, so we will write down the price of the ad with the lowest accepted bid on these selected ads.
Now let’s imagine another query with a different zoom, covering a larger radius and therefore more likely to find ads with higher bids.
In this case, the four selected are 2, 3, 6, 8, and the lowest bid of these is 1500. So we assign 1500 to each of these selected ads.
For this example, there will be no more queries on these ads, to calculate the final rate we simply have to take the arithmetic mean of all the accepted values.
As we can see, even though ads 2 and 3 have a different price, because they have been exposed to the same circumstances, the final monthly fee is exactly the same.
This gives us two advantages:
- Any advertiser do not pay more than the competition.
- In the event that an ad is surrounded by others with a much higher bid
and that ad does not appear in any searches, the advertiser will not be unfairly charged a monthly fee.
With this method, we will also be able to inform advertisers if your publication is not being visited due to surrounding competition, so they can decide to compete raising the maximum monthly bid.
There is an additional factor to note, if an ad has a much higher bid than those around it, with the bid being much higher than the surrounding ads, with this system it may not seem to make sense to give it a higher maximum bid, but if the user zooms out and sees a larger area, there is a better chance of finding ads with a higher value and it will make sense to give the ad a higher value as it would be visible at larger scales.
Per example, if we zoom out and see the whole region of Spain, we are much more likely to find ads with a higher value. The selected adverts are much more likely to be in large cities such as Madrid or Barcelona, as these cities have more transit areas and are more competitive when it comes to placing adverts. Although there may also be advertisements that are intended to be seen much further afield, it will ultimately depend on the market that the advertiser has in mind, whether it is more local or wider.
Monthly Fee and Visibility Radius Prediction
OK, so now we know the benefits of paying more for a particular ad.
The problem comes when an advertiser has to decide how much to pay for a particular ad to appear in more competitive areas and to appear in much larger areas by zooming out.
The problem comes when an advertiser has to decide how much they should pay to achieve their intended target, if we cannot give them a priori information on the expected reach, they would have to measure the price every month until they achieve their initial target.
The metrics the advertiser is interested in are the following:
- Approximate monthly audience that your publication will receive
- Radial distance that the publication will reach
It must be taken into account that the greater the radial distance, it does not mean that people who are further away will see it.
They will also have to zoom out on the map to see where the ad is placed.
We have a regression problem where we have the monthly quota and the coordinates of the ad that we want to predict, and historical information from publications with information on their bid, audience and radial distance reached.
The information will depend a lot on the closest ads, the closer the ad the more relevant it will be. Another requirement to be taken into account is the constant updating of the information, as advertisements can appear and disappear every day, completely changing the context in a matter of hours.
The context can change in a matter of hours, so we are faced with a situation of concept drift, which we have to take into account for regression.
Another problem for regression is that we have spatial data. The direct use of latitude and longitude coordinates will not add much to the prediction of the model.
Therefore, in order to extract useful information from the coordinates, we will create 3 categories based on the geographical and audience context in which the ad is located.
To define these clusters, we will use as a measure a square whose size will not be determined by the default zoom that is used when entering the home page, as it is not the home page, as it will not be common for users to move around the map.the map. The default zoom is 13.
Level 3: Areas with less than two paid publications in a square area with zoom 13.
Level 2: Areas with two or more publications.
Level 1: Areas of hypercompetitive publication, where there is a high density of publications, there are more than 10 publications in the square area.
Once the delimitation of the map is done, we can know by the coordinates to which category each payment post belongs, having a useful categorical variable for our regression. useful categorical variable for our regression.
For the regression, we can use different types of algorithms. Among the various alternatives, the following stand out for their popularity and predictive capacity: neural networks, random forest or naive Bayes. In line with our agile development methodology, we have chosen a model such as k-NN (k Nearest-Neighbour), since it does not require complex estimation of its hyperparameters and a training phase.
In conclusion, our innovative bidding system has ushered in a new era of localized advertising within the realm of social networking. By seamlessly integrating with the application’s interactive map, this system empowers local businesses to connect with their target audience in a highly targeted and engaging manner. Users benefit from a more personalized experience, receiving advertisements that are relevant to their immediate surroundings, enhancing their exploration and discovery process.
