OrangeFlow

Challenge

The paid media team faced problems with delivering insightful funnel reports to clientes at scale. The team was using Funnelytics to create the reports. This tool is great for creating funnel maps, but we found that the "forecast" feature was not very user-friendly, and lacked a few features that we needed. Also, the reports were not very customizable, and the team was spending a lot of time creating them manually. Furthermore, the a recent price increase in Funnelytics made it unfeasible for us to keep using it.

Solution

Using a javascript React library, I created a custom visual interface that allows the team to build funnel maps and reports quickly and easily. The interface is user-friendly and allows for easy customization of the reports. It also allowed for metric forecasting, which was a feature that we were underusing in Funnelytics.

With ownership of the code, we saw the oportunity to implement a few features that we needed. By integrating our application with LLM tools like GPT and Claude, we were able to create a system that generates reports and insigthful analysis with one click.

The system uses the context of funnel maps created in the interface to generate reports that are tailored to each client's needs. The reports include insights on the performance of each stage of the funnel, as well as recommendations for optimization and, more importantly, a forecast of the results that can be achieved with the current budget.

Features

The application includes several features to enhance the reporting process. The main features are:

1. Flow (Funnel Map Creation and Forecasting):

The application allows the user to create funnel maps using a simple drag and drop interface. The user can add stages to the funnel, and then provide the conversion rate for each step. The application will automatically calculate the result, based on the conversion rate and the budget.

2. Report & Analysis Generation:

The application uses LLM tools to generate reports and analysis based on the funnel maps created by the user. On the image, we can see a few examples of the reports that can be generated.

The application generates a report that includes insights on the performance of each stage of the funnel, as well as recommendations for optimization. The report also includes a forecast of the results that can be achieved with the current budget. On the image, we can see a full Page Analysis, that:

1. Access the URL of a selected page, from a given funnel step, and visually analyzes a Screen Shot of it's first section;
2. Fetches the page's full text content and analyzes it using LLM tools;
3. Obtains the full funnel context, with the conversion rates and budget, and analyzes it using LLM tools;
4. Generates a report that includes a forecast of the results that can be achieved with recommended on-page optimizations, for the current budget.
5. Finally, the user can generate an implementation plan, with a step-by-step guide on how to implement the recommended changes.

3. Chat with AI Assistants:

The application includes a chat feature that allows the user to chat with AI assistants. The assistants are powered by LLM tools and can answer questions about the funnel maps. The assistants can also provide recommendations for optimization and answer questions about the performance of each stage of the funnel. The assistants are provided with context of the selected funnel map, so they can provide accurate and relevant answers.