29,000 beauty salons
in one system

Custom web app

Business need

Beauty is a cruel mistress, especially when you need to arrange a last-minute visit to a hair salon. San Francisco alone has 2,900 salons, scattered over dozens of booking platforms. This complicates what could be a very simple process. 

Result

We built an all-in-one booking platform for beauty salons, spas, and barber shops. Our proprietary AI-driven development framework cut the time-to-market by eight weeks. This saved 75% of the client’s budget, allowing them to expand the business into more locations across the US.

Industry
Beauty
Location
USA
USA
Working together since
2024

Salon Tonight is a startup from San Francisco that allows people to book a perfect salon visit in just a few clicks.

The challenge

40 million people in the US alone search annually for a "salon near me". Yet, the booking market is incredibly fragmented. Some sites only allow you to search for spa salons. Others lack half of the available female hairdressers, most barbers, and all massage salons. Gathering the data from all salons into one system is both a challenge and a huge opportunity.

The solution

MindK used our new AI development framework to build the system with a team of one Solution Architect. The application features a monolithic architecture with server-side rendering for SEO. It scrapes data from 12 systems every two hours. This created a great challenge for the application performance that we solved using the AI.

10x times more results than competitors

We designed Salon Tonight to provide up-to-date information about as many salons as possible. This required scraping data from over a dozen aggregators. Rather than having simple API calls, we need to create a step-by-step process with a few workarounds. Development could take a lot of effort since all scrapers required unique tuning to match the data sources.

10x times more results than competitors

We designed Salon Tonight to provide up-to-date information about as many salons as possible. This required scraping data from over a dozen aggregators. Rather than having simple API calls, we need to create a step-by-step process with a few workarounds. Development could take a lot of effort since all scrapers required unique tuning to match the data sources.

With a few suggestions, the AI wrote code for a test run. However, the first version of the scraper parsed a lot of unnecessary data. The next challenge was to identify the data we needed to adjust the parsing. AI can analyze text much faster than humans if it understands the possible data formats. This saved us a lot of time. Even with the full context for the AI, it still took a while to find the best prompt for the task.

Lightning-fast performance with Asynch IO and Redis

With thousands of requests per hour, performance was quickly identified as the application's main bottleneck. The AI suggested creating a separate process for each scraper using a Python library called Async IO. Working with the library is notoriously difficult. It was a great surprise to see the AI use Async IO so well.

The parallel processing of requests also required us to create a message broker using a Redis event broker. It adds a time stamp to each response and automatically schedules the messages. This allows Salon Tonight to run tens of thousands of requests in parallel without tanking the performance.

Intuitive two-steps booking

When analyzing competitors using AI, we were shocked by the sheer variety of booking interfaces. They lacked consistency, required a lot of information and unnecessary steps. To address this problem, we implemented a familiar two-step booking process. This resulted in a 27.3% conversion improvement, according to A/B testing.

Intuitive two-steps booking

When analyzing competitors using AI, we were shocked by the sheer variety of booking interfaces. They lacked consistency, required a lot of information and unnecessary steps. To address this problem, we implemented a familiar two-step booking process. This resulted in a 27.3% conversion improvement, according to A/B testing.

Insights about AI-driven development

AI may already know the API documentation, search the Internet, or parse the provided URL.
It's possible to skip documentation while developing new features.
AI was handy in analyzing the codebase and generating the docs.
AI is much faster at text extraction, normalization, regex patterns adjustment, keywords detection, and can understand the possible data formats.
Adjusting the infrastructure was beyond the competence of AI, as the project has an APP and CMS on the same domain.
Although AI had the full context, it took a while to find the best prompt to successfully optimize our scrapers.

Headless CMS functionality for easy administration

The last piece of the puzzle was an admin panel for easy content creation and editing. The AI created a headless CMS-like functionality without using an actual headless CMS. The SEO optimization will in time allow our client to capture 40+ million annual searchers in the US.

Business Value

The app allows users to choose from 10x more beauty salons than its largest competitor. The "featured salon" functionality allows business owners to advertise their services in exchange for a subscription fee. Dozens of salon chains have already tried out this feature during the first two months after the release. The team is currently adding thousands of new businesses and locations in preparation for the 2.0 version.
1
Beauty salons to choose from
1
Platforms updating information every two hours
1
Performance improvement using asynch processing

More cases

CEMAsys

CSR solution for Forbes Global 2000 companies

Learn more

Aim Analytics

Data-driven TV and radio ad campaign management

Learn more

Bridge

Operating system for global hiring

Learn more