Leveraging AI technology has revolutionized data analysis, offering unprecedented insights and efficiency. However, the precision of these insights depends on the clarity and detail of the prompts provided to AI Analysts. While prompts do not have to be lengthy, incorporating key elements can greatly enhance the relevance and accuracy of the AI's outputs. Here are the key components of an effective prompt:
Specific Metric(s): The specificity of the metrics you inquire about plays a crucial role. Instead of broad queries like "how was performance," which grants AI considerable interpretative freedom, specifying the metric of interest, such as "click-through rate," enables the AI to focus and deliver more precise responses.
Dimensions Filtering: Guiding the AI to the target data set improves the quality of analysis. By specifying the dimension (e.g. platform, campaign, ad set, ad) you're interested in, you enable to filter data more effectively. For instance, indicating a preference for "retargeting campaigns across Facebook and Google Ads" allows the AI to apply logical filtering, identifying campaigns likely associated with retargeting. You can also provide more specific filtering in your prompts, like "use campaigns that have '2024' in the name" or "filter for ad sets that mention 'BOFU.'"
Date Range Definition: Clarifying the time frame for your analysis is essential for accuracy. Whether you provide specific dates (e.g. January 12 - 24, 2024) or use natural language (e.g. last week, last year, December 2023), defining the period of interest helps Avian focus its analysis on the relevant data.
The Impact of Detailed Prompts
While Avian is capable of operating with minimal information, the detail and clarity of your prompts directly influence the efficacy of the AI's analysis. Providing more specific information reduces the need for assumptions on the AI's part, leading to more accurate and actionable insights.
That said, the power of Avian lies not only in its ability to handle detailed prompts but also in its flexibility to interpret and act on simpler, natural language requests. This design philosophy encourages experimentation with different types of prompts, guiding users to discover the most effective ways to access and analyze their data. By focusing on brevity and clarity, you can achieve optimal results while taking full advantage of Avian’s natural language processing capabilities.
Troubleshooting Tip: When crafting queries for Avian, it's important to be mindful of the inherent limitations present in the APIs of the data platforms we connect with. These limitations can impact the combinations of dimensions and metrics that can be returned in a single query.
For example, the Meta (Facebook) API does not support fetching combined data for "Age and Gender" breakdowns alongside "Device" data in the same query. This restriction means that, despite Avian's powerful analysis capabilities, we are unable to retrieve certain combinations of data that the native platform APIs restrict.
For data insights that require combinations of dimensions and metrics not supported in a single query by the source API, consider breaking down your request into separate queries.
If you've adjusted your prompts and are still encountering issues, check out this article: Data Configuration Issues.