Quickstart guide to MapXP available at the bottom of this page
Map is based on the TEXT of Search Results (organic, ads, shoppoing, news)
Keywords were indentified through Sistrix:
- Highest search volume keywords from the keywords that top U.S. Banks rank on:
[Bank of America, JP Morgan, Citi, Wells Fargo, Goldman Sachs, PNC, Capital One, Regions, Key, Navy Federal, Ally]
To determine how similar firms' positionings are on Google, we used Natural Language Processing (NLP) and unsupervised machine learning.
Retail Banking topics are discovered with semantic text analysis of search results. Prevalence of noun-chuncks in each topic are visualized below.
Noun N-Gram Clouds for each Topic (mouse-over to see):
Individual firm clouds are available in the Attribute Pop-ups when you mouse-over their respecitve bubbles on the map.
Firm clouds are topic specific - a firm that is positioned in multiple topics will have a different cloud in each topic.
expresses how prevalent a firm is within a topic as a function of how many keywords associated with a topic the firms appears on, and on which rank.
expresses the extent to which a firm is key player within a topic. It is a function of the number of firms within a topic that the focal firm is very similar to
- Explore Map by KPI according to bubble color and size using drop-downs at top
- Mouse-over Bubbles to see most similar firms (when slider at bottom is set to > 0%)
- Zoom-in or out with mouse wheel or slider on right
- Click+Drag moves map (can be used in combination with zoom)
- Show Names of firms in map with check-box at top
- Adaptive Name Size (check-box at top) varies size of firm names in map according to selected size KPI (drop-down)
- Summary of firm KPIs on mouse-over when check-box "Attribute Pop-up" is checked
- Mouse-over Legend to highlight firms in map corresponding to legend criteria (e.g., on heat-map mouse over dark green color to show all matching firms)
- Search-Bar with autocomplete helps you find a firm in the map faster when you enter its name or part of its name
- Similarity-Slider at the bottom allows you to view similarities among firms' positionings as arcs (curved edges): arc-weight indicates degree if similarty (the fatter and darker, the more similar)
- Rotate & Reflect map using controls on the left of this text box: enter degrees (-360 to 360) and click button to rotate; reflect vertically by clicking button
- Double-Click on map to reset zoom and center map
Map projection made with stochastic neighborhood embedding.
Topics identified using manifold approximation and high dimensional density clustering.
Firm proximity in map based on similarity of firms' texts in search results (embedded using transformer based sentence encoding).
Word clouds based on N-grams of nouns identified by word dependency parsing and part of speech tagging.