Data mining and restaurant location search heuristics: almost the same problem as with free parking
Red dots - there are no tables, green - there are places.
The problem of getting to restaurants in Moscow during rush hours is well known : you can not sit at a table for lunch at a business lunch or just can not find a place on Friday evening.
At the same time, as you probably know, many well-known and constantly busy establishments are inferior in quality to less well-developed analogues. The main problem is that visitors simply do not have the opportunity to learn about such places if they are not specifically interested in them. At the same time, there is usually no time to engage in the selection of establishments, the choice is already on the street. The traditional mechanics of restaurants - the reservation system - are failing here.
I continue to talk about crowdinvesting projects that are on the Smartmarket (I remind you that we are a kickstarter-type platform, only in the Russian Federation and with the purchase of a share in the company's profit). The solution to this problem is the essence of bocco startup . The idea of the project is to display the workload of restaurants in real time. A resident of a large city just takes out a smartphone and says to his companions: “But behind this house is a free cafe, which is quiet and cozy. Let's go there!". They go around the line and go to a place where they can sit quietly.
The first question is the suitability of the idea. The team has been developing the project for a long time, and during this time an extensive feedback was received on the need for such a service. As we see this situation, we: the key issue of success in the quality of implementation is that information about workload should accurately reflect reality.
So, the solution is free mobile applications for different OS and site. The service shows the loading of the nearest establishments and helps to find a table. To keep it “yours”, right at the touch of a button from the application, you can book it.
The most difficult moment is obtaining data on congestion. There are four vectors here:
- From restaurant software that knows how many tables are occupied.
- From restaurateurs themselves through a software solution from the service.
- From users of mobile applications (as in Yandex.Traffic: you solve your problem and share data with others). Beta, by the way, is already working.
- From open sources, for example, from the chekins of social networks.
Plus there is still a vector - the service constantly trains an expert system that allows you to evaluate workload. Various external public parameters are introduced into it, the load history (beta testers are already collecting it) and data from restaurants. As a result, the main gap closes - the lack of data on points that do not know about bocco. The load in them is calculated algorithmically, and the accuracy of the algorithm is constantly growing.
As mentioned above, life is quite simple for the user: he sees a free restaurant and just goes there. If he chooses an institution in advance, then he can reserve a table using the button. The application itself will do everything you need. But in restaurants, the booking procedure now is not a gift, and requires quite complex operations. The team also makes a piece of software that automates the reception of armor, which greatly simplifies the work of staff. This part of the service will be integrated with restaurant CRM systems and their software.
Now there are agreements with manufacturers of restaurant software, with restaurants (more than 10 large) and several partner sites. The team completes the legal formalities, after which it will be able to sign official agreements with partners and name them openly.
The team decided that this is the best way to raise funds for the current stage of the project. There is already a fully functional beta, many parts of the product are in the final stages of development. Sowing is no longer very relevant, while the service is not running, so there are no sales, so the team is not ready for Round A. Smartmarket is a great opportunity to bridge this gap.
Plus, given the need for an active feedback, crowdinvesting will provide an opportunity to get at least a hundred invested activists who will help improve the algorithm for determining workload, as well as make feature requests and write about bugs.
What kind of team?
In the field of IT, each team member has serious experience. The participants worked in mail.ru, ABBYY, wildberries, CMX.
Assessing the competitiveness of the project, we immediately asked what would happen if 4sq or Altergeo fastened the functionality literally at the moment after the release of the service. Everything is interesting here:
- For the correct display of congestion on chekins, it is necessary that a significant percentage of visitors to the restaurant check in. Now this situation is very, very far away. Checks from 4sq are used as one of the parameters of the algorithm - in the vast majority of establishments, the number of people at the same time based on the checks is not more than a dozen during the day. Therefore, even as one of the parameters, chekins are suitable only for a small number of institutions popular among the corresponding audience. As the only criterion in the foreseeable future, the Chekins will be irrelevant.
- The load mapping service is a very local product. The current algorithm has more than 20 parameters, most of which are unique specifically for Moscow and specifically for restaurants - in other cities or for other infrastructure objects completely different parameters will be required, other hypotheses will be applied, etc. 4sq is a global product, for them such an approach is impossible. Altergeo could theoretically repeat this work, but for them it would mean the need to catch up, so the team is not scared.
Interestingly, bocco has no plans so far to sell the service of one of the social networks.
The next question was about the relevance of obtaining data . It was important to understand how the owners of the establishment are objectively interested in providing data on workload, because the hype with the queue can be beneficial. In practice, it turned out that this is a fairly common misconception. The benefit of the service for underloaded establishments is obvious to everyone. Popular institutions, of course, benefit from a full load, but not a stir. Excessive loading reduces the level of loyalty to the restaurant: if the client is refused, the institution loses the location of such a visitor, because the next time when choosing a place he remembers that you can not sit there. Therefore, in the long run, an honest display of congestion is the most profitable strategy for a restaurant.
At the same time, it is clear that not everyone will conscientiously give data on a full load. Accordingly, the task of the algorithm is to correctly predict the overabundance of people in large restaurants and give actual data from sources in restaurants with less traffic.
The critical mass of users that is needed for up-to-date data will be typed in about six months after launch. At the same time, since user feedback is not the only factor, in the aggregate it will take only a couple of months to get started - and this is just the beta test period.
To recruit an audience, it is planned to promote applications in the stores, as well as receive traffic from partners with whom we already have preliminary agreements.
- Server part - all the functionality is implemented.
- Website - design developed, implemented static layout.
- Application for Android - all the basic functionality is implemented, except for booking.
- Application for iOS - designed, implemented a prototype with the main screens.
- Product for restaurateurs - a working prototype has been implemented.
- 990 thousand rubles have already been raised from the required 2.4 million, that is, slightly less than half the funds.
More details on the investment side of the project are on the Smartmarket page , plus there is information about the team, details and a link to the site where you can subscribe to the beta.
Only registered users can participate in the survey. Please come in.
Will it fly or not?
- 51.8% Yes, free tables need to be looked for 99
- 29.8% I don’t know, I did not come across the question of finding tables 57
- 18.3% No, even cool data mining will not help here (or, perhaps, I will write another reason in the comments) 35