Prompting with Adscorn

List of prompting strategies and approaches

To help you make the most out of Adscron, we compiled a list of prompting strategies and approaches. Some of these were collected from our team’s experience, and others were shared with us by our community members.

What is prompting? Prompting is the term for textual, natural language inputs used to interact with Adscorn Playground.

Since Adscorn relies on LLM and RNN models, effective prompting strategies can significantly improve its efficiency and accuracy.

Basics: Modes of Operation

The Playground operates in 3 primary modes, each designed to assist with specific types of queries:

  1. Charts: Generate Historical Line or Bar Charts to visualize performance data over time.

  2. Logical Computing: Analyze insights, identify trends, and determine top-performing elements.

  3. Text: Generate creative copy suggestions, explore strategic approaches, or ask any question.

Dataset Levels

The Playground processes data across 5 hierarchical dataset levels:

  • Creatives: Includes assets such as images, videos, or copies.

  • Account: Data aggregated at the account level.

  • Campaign: Focuses on overarching promotions or outreach efforts.

  • Ad Set: Targets specific bundles or audience groupings within campaigns.

  • Ad: Analyzes individual advertisements.

Dataset Auto-Detection

By default, the system automatically detects the dataset level based on your prompt. However, you can force a specific dataset by including relevant keywords in your query:

  • Creatives: Use terms like creative, asset, media, image, video.

  • Campaigns: Use terms like campaign, outreach, promotion.

  • Ad Sets: Use terms like adgroup, adset, bundle, target group.

  • Ads: Use terms like ad, advertisement.

Text Prompting

Leverage the playground to craft compelling ad copy, generate creative ideas, and address common queries with ease.

Simple prompting example

Logical Prompting

Use logical queries to calculate metrics, identify the best or worst campaigns, creatives, and channels, and gain actionable insights.

Metric computing examples

Use action-oriented words like compute, calculate, average, total

  • Compute Total Revenue

  • Compute total AOV, CR in % and CTR in %

  • Calculate conversions value of the best 3 creatives, round 2

Search prompting examples

Use action-oriented words like list, set, find, sort , return

  • Sort campaigns by ROAS and return the best 3

  • List of the top 3 creatives with the highest impressions

List of Useful Metrics

Chart Prompting

Use queries to create historical time-series and bar charts

Chart prompting examples

  • Creatives AOV chart (top 5)

  • Revenue chart

  • A chart of 3 campaigns with the highest conversion rate

Fine-Tuning and Data Transformers

Use Additional Keywords for Data Formats

  • Add %: Retrieve results as percentages (e.g., "CTR in %").

  • Round 2: Round values to 2 decimal places (e.g., "Conversion value, round 2").

  • Sort asc/desc: Arrange data by value (e.g., "Top campaigns sorted descending").

  • Show top/bottom N: Display specific ranges (e.g., "Show top 5 channels").

  • Filter by: Narrow results based on criteria (e.g., "Filter by ROAS > 3")

Last updated