Dataset of A Framework for Developing External Key Performance Indicators Using Google Trends and Multi-Criteria Decision-Making

Datasets Description
To support the development and validation of the EKPI framework, three datasets were collected directly from Google platforms, covering a ten-year period (2014–2024). These datasets were structured in CSV format and include:
1. Google Results Dataset
This dataset captures the number of Google-indexed pages that mention each target brand over a ten-year period (2014–2024). To extract this data, we employed a monthly querying strategy using Google Search’s built-in time filter capabilities. Specifically, the following advanced search operator was used for each brand and month:
"keyword" after:YYYY-MM-DD before:YYYY-MM-DD
For example, the query:
"Brand Name" after:2014-01-01 before:2014-02-01
retrieves the number of indexed pages containing the keyword within January 2014. This approach was applied for each month over the study period and for both brands (“Hamrah-e-Aval” and “Irancell”), resulting in a longitudinal dataset representing the digital content footprint of each brand across time.
All collected monthly results were stored in tabular format, forming the basis for indicators related to Share of Voice(SOV) in Google.

2. Google Trends Timeline Dataset
The second dataset was directly downloaded from the Google Trends platform using its official export feature. The "Interest Over Time" graph in Google Trends provides a normalized time series of search volume for a given keyword or brand. For each brand, data was exported as a CSV file by selecting the appropriate time range (2014–2024) and region (Iran). This resulted in a continuous timeline of search popularity, used to calculate indicators such as:
- Brand Awareness in Google (BAG)
- Growth Rate of Awareness in Google (BAGR)
The exported file includes timestamps and corresponding interest scores (0–100) normalized by Google.

3. Google Trends Related Queries Dataset
The third dataset consists of monthly related search queries obtained from the “Related Queries” section of Google Trends for both target brands. These related queries offer insights into specific user intents and behaviors beyond the main brand search.
For each month between 2014 and 2024, related query data was downloaded as CSV files using Google Trends' built-in export functionality. To enrich this dataset, we manually added a sentiment column to each query. Specifically:
- A subset of 3,762 unique queries was labeled by three independent annotators as positive, negative, or neutral based on domain-specific criteria.
- A domain expert finalized the annotation where disagreements occurred.
- These labeled queries were later used to train and validate a hybrid sentiment analysis model combining lexicon-based[36] and deep learning techniques.
-All 47,000 queries extracted over 10 years were automatically labeled by the trained model.

4. lexicon-dictionary