Market Brew

Release Notes

New features, bug fixes, and content you might've missed.

may 21, 2024
  • Modeling Google's AI Overviews. The SGE Visualizer has been updated to model Google's AI Overviews.
may 15, 2024
  • Modeling Site-Level Parasite SEO. A new algorithm, based on embeddings, is introduced into the search engine model.

    Called Top Cluster Similarity (SUBDOMAIN), or shown as "Top Cluster (Site)" on the SGE Visualizer, it is site-level algorithm that looks for the embedding chunk with the highest cosine similarity for the query, and updates the similarity depending on whether or not that embedding chunk is also found on the site's top embedding cluster.

    Pages that aren't related to the site will receive lower scores because of this algorithm. This can be considered a site-level Parasite SEO filter.
  • Site Cluster #. A new metric is being captured for all new crawls. After converting content to embedding chunks, each embedding chunk now receives a site cluster #, depending on the k-means calculation of all embedding chunks on that subdomain. In addition to using this in the Top Cluster Similarity (SUBDOMAIN) algorithm, users can also view this metric on the Website Embedding Listing screen, under Result Cluster.
may 8, 2024
  • Modeling Page-Level Parasite SEO. A new algorithm, based on embeddings, is introduced into the search engine model.

    Called Top Cluster Similarity, it is page-level algorithm that looks for the embedding chunk with the highest cosine similarity for the query, and updates the similarity depending on whether or not that embedding chunk is also found on the page's top embedding cluster.

    Content chunks inserted into a page that aren't related to the page will receive lower scores because of this algorithm. This can be considered a page-level Parasite SEO filter.
  • Page Cluster #. A new metric is being captured for all new crawls. After converting content to embedding chunks, each embedding chunk now receives a page cluster #, depending on the k-means calculation of all embedding chunks on that page. In addition to using this in the Top Cluster Similarity algorithm, users can also view this new metric on the Website Embedding Listing screen.
may 1, 2024
  • Spotlight Visualizer. A FREE limited version of Spotlight has been released to the public. This is the second visualizer representing the entity-based semantic classifiers in modern search engines, with the SGE / Embeddings Visualizer being the first.
  • Spotlight Author v2.0. An updated version of Spotlight Author has been released. This tool is also available to the public.
april 10, 2024
  • Ask Calvin v2.0. Thousands of real Market Brew customer service responses have been now included in Calvin's context window. Ask any question you have about Market Brew or SEO in particular.
  • SGE Author. Users can now test their HTML in real-time to see which version works the best with embeddings. This testing tool is one of the core SEO tools for beginning to explore how search engines include your cntent ( and their embeddings) in its ChatGPT-like (Large Language Model) response called the Search Generative Experience (SGE).
March 11, 2024
  • First Input Delay Deprecated. Following Google's decision to remove First Input Delay (FID) and replace it wit Interaction to Next Paint (INP), Market Brew has removed FID and Domain FID from the list of available algorithms in all new Ranking Blueprints.
February 22, 2024
  • Content Blended Similarity. Following the deployment of the Content Similarity Score and its algorithm inside all Ranking Blueprints, we now introduce a different scoring methodology: an average across all embedding chunks. The SGE Visualizer has been updated with this new metric (Avg. Similarity):
  • All new Ranking Blueprints include Content Blended Similarity. Tasks will automatically show the breakdown of the score for your target page vs. the outperformer.

    Clicking on the Content Blended Similarity percentage will take you to the SGE Visualizer, whereas clicking on the Embedding Chunk Offset will take you to the individual Embedding Chunk screen, complete with a breakdown of extracted entities and topic cluster for that specific chunk.
February 17, 2024
  • Website Embeddings. From the Website Dashboard screen, users can now search embeddings and get back a similarity score and a cluster group #. Clustering can be adjusted from 3-10 groups, and URLs can be filtered so that the embedding search / clustering is only done on that filtered group of URLs. If embeddings are not yet enabled, click "Actions" and "Re-Crawl".
  • Centerpiece Annotation. All webpage embeddings now have a particular set of entities and topics, as well as a topic cluster. Just click on the eye icon to view. Future derivative algorithms will use this data.
February 12, 2024
  • Heading Vector Algorithm. Google's purported Heading Vector Algorithm has been modeled and added to the list of default algorithms with each new Ranking Blueprint. Each H1...H6 tag is scored based off of a cosine similarity score for that tag and the query. H1 tags are weighted more so the similarity counts more. An average score is taken based on all H tags on that page. This new algorithm can be added by creating a new Ranking Blueprint and ensuring all sites have been re-crawled.
  • META Title Embedding. All META Titles now have an associated similarity score. This is a cosine similarity between the META Title field and the query. This new algorithm can be added by creating a new Ranking Blueprint and ensuring all sites have been re-crawled.
January 31, 2024
  • SGE on Market Focus and Spotlight Focus screens. The SGE Visualizer tool is now available directly from both Market and Spotlight Focus screens.
  • Embedding Footprint. The SGE Visualizer tool has been upgraded to include a visual footprint of each embedding cluster. Each embedding cluster has a footprint. To create a unique footprint, we visualize the embedding vectors in a lower-dimensional space, projecting them onto a 2D plane using a technique called Principal Component Analysis (PCA).
January 24, 2024
  • Embedding Cluster View. The SGE Visualizer tool has been upgraded to include a cluster map of each page. Each embedding block has been assigned to one of three groups, called the embedding cluster. This cluster is viewed as a group in many embedding search algorithms, and this cluster footprint is sometimes used during the machine learning process to recognize good or bad sites.
January 19, 2024
  • Ranking Blueprint. The Analysis Group has been renamed to Ranking Blueprint.
  • % Associated Link Flow. The Market Focus and Spotlight Focus algorithm screens now display % Associated Link Flow for a better understanding of what the incoming anchor text distribution looks like.
  • Free SGE Visualizer Tool published. Use any URL to explore the beginnings of Search Generative Experience (SGE) modeling. It begins with a embedding visualizer that allows users to begin to understand how search engine LLMs might choose your content based on embeddings.
January 10, 2024
  • Cosine Similarity Inspector Tool By Webpage. The Cosine Similarity tool has been added to the Webpage Scoresheet screens, under "Content Analysis" -> "Cosine Similarity Tool". Enter any search query and you will be taken to the Cosine Similarity Inspector Tool for that webpage, for the given query.
  • Cosine Similarity Inspector Tool. Added top similarity score.
January 7, 2024
  • Cosine Similarity Algorithm. The Cosine Similarity Algorithm has been added to the baseline search engine model. This algorithm models Sentence-BERT embeddings and Google's concept of Centerpiece Annotation. A cosine similarity score is calculated against segments of content on each page. Create a new search engine model to include this new algorithm.
  • Cosine Similarity Tasks. Like other algorithms in the model, each landing page in the SERP is evaluated to input into the calibration process and receive its own bias/weight settings (called Boost Factors). When there is a high correlation, this task will be shown and both the target page and outperformer cosine similarity scores will be displayed, each with links to the Cosine Similarity Inspector Tool.
  • Cosine Similarity Inspector Tool. An inspection tool has been created to analyze the embeddings.
December 30, 2023
  • Location Algorithm Updated. Improvements to latitude / longitude coordinates for short queries.
  • Core Web Vitals not set. CWV data that is not available for a particular URL will now be represented as the maximum range for that Core Web Vital. This was previously assumed to be the minimum range, causing tasks to erroneously show not set URLs as the best performing.
November 10, 2023
  • Introducing Calvin, the first AI search engineer assistant in the world. Calvin is immediately available to all existing users. Calvin can analyze your search engine models, or any of the data on the Market Brew platform.
August 30, 2023
  • One-Time Calibration is now default. When using the setup wizard, the one-time option is now selected as the default. This aligns with the best practices of our users, as most users want to manually re-calibrate the models.
  • Performance Enhancements. Infrastructure investments continue to improve build time.
August 1, 2023
  • Missing PSO Scores. Particle Swarm Optimization (PSO) scores that are missing on the Analysis Group Listing screen now have an informational icon and popup to warn users that these Analysis Groups have no PSO score assigned. This can happen if the Analysis Group is currently being setup, or if there was an issue with calibration.
  • Model Rankings. The Model Rankings column on the Analysis Group screen was previously showing the rankings since last calibration. They now depict both the current rankings as of the last re-calculation / website re-crawl as well as the rankings at the time of calibration.
JuLY 27th, 2023
  • Interaction To Next Paint. This Core Web Vitals metric has been added to the list of Boost Factors. Additionally located on the Website Dashboard and Webpage Scoresheet screens. All Core Web Vitals metrics are now pulling from Google's CrUX system, using the 75th percentile measurement, aggregated from the last 28 days.
  • Password Reset. A bug preventing the password reset has been fixed.
JuLY 13th, 2023
  • White Label Added. The Market Brew user interface is now able to be privately labeled (white label). Contact your representative for details.
June 10th, 2023
  • Custom SERP. Market Brew uses 3rd party ranking data to calibrate against. Sometimes that ranking data is incomplete. In those cases, select “Custom SERP” under the TSEE options, and provide a list of valid URLs that the Analysis Group should use to calibrate against.
Custom SERP
June 8th, 2023
  • Bulk Analysis Group Deletes. A performance issue was fixed when requesting bulk deletes.
  • Similar Words. Market Brew's model of Google's Hummingbird system now shows the list of similar words in the alert emails during Analysis Group setup.
Similar Words Email Alert
May 16, 2023
  • Task Simulator. You can dial in the % completed for each task, forecast rankings / traffic against your search engine models. For instance, if you are unable to complete a certain task, or only partially able to complete it, you can see what will be necessary to achieve your desired new ranking.

    Query Score, Ranking, and Traffic Improvements are instantly updated by adjusting the completion sliders for each task. Click the Raw Query Score improvement to view the Combined Search Listing screen to visualize how implementing the tasks will effect your new rankings.
Task Simulator screen
  • Task Simulator link added to Optimization Tasks screen. All optimization tasks screens that are filtered by Analysis Group will now have an option to forecast rankings / traffic.
Forecast Rankings button
May 03, 2023
  • Calibration automatically turned off after Analysis Group setup. Auto self-calibration option has been removed. Now, each Analysis Group will turn off calibration automatically after setup. This workflow was used 95%+ of the time for users, as this was a necessary first step to begin A/B testing after calibration.
  • Manual Calibration Option Added. After A/B testing cycle is completed, select "Calibrate Model" from the Refresh dropdown to re-calibrate the model against the updated rankings.
  • Stale calibration alert added. Market Brew will bubble an alert if the Analysis Group hasn't been calibrated in more than 90 days.
  • Tasks and Target Page's Query Score monitoring added. The overall task progress and the target page's query score improvement in the model rankings are now tracked on the Analysis Group screen, showing the improvements since last calibration.
  • The Task Listing Screen has replaced the completed field checkbox with a progress bar. The progress bar will show how much each algorithm score has improved since last calibration. 100% means the gap has been completely closed between your target page and the outperformer page for that algorithm. Updates each time the Analysis Group is re-calculated. Progress bar will start updating once Analysis Group has been calibrated again.
  • % Occurrences Column added to Market / Spotlight Focus screens. New column showing % occurrences can now be sorted/filtered.
april 17, 2023
  • Pagination added. On Market Focus and Spotlight Focus screens, pagination has been added to the top of the lists.
  • Similar Words alerts update. Similar Words alert emails were being sent each time a re-calibration was triggered. Now this email will only be sent once per Analysis Group.
april 12, 2023
  • Social Links updated. Improved algorithm calculations in the Paid Link, Advertorial Link, and other areas. Links to will no longer be counted in these penalty calculations.
  • Preferred Pages. Preferred URL "reset to default" option has been added on the website edit screen. This option will lookup every target page added to an Analysis Group and reset your preferred pages to this list.
Reset to Defaults - Preferred Pages
april 6, 2023
  • Link Neighborhood Tasks. Improved recommendations for all Link Neighborhood Tasks. Each task will now show both the outperformer's and target's incoming/outgoing links alongside their link flow values, for easier comparison. To pick up these new changes, re-crawl the sites that you wish to get this new detail.
March 30, 2023
  • Link Neighborhood Penalties. Landing pages that had no incoming links were being incorrectly promoted in the model as having a good Link Neighborhood. We've adjusted this so pages with no incoming links will be worse than pages with incoming links. To pickup this new change, initiate a re-crawl for the site for which you wish to see this update.
  • Advertorial Penalties. The advertorial penalties were showing the wrong percentage amounts in the task summary. The penalties were valid, but the percentages were not the values that were used in calculating the algorithm. Values have been updated and all new tasks generated will contain the updated figures.
March 24, 2023
  • Spotlight Focus / Expertise Task Enhancements. Topics and keywords are now suggested to improve both the Spotlight Focus and Expertise scores. This is now viewable in all new generated optimization tasks and Conversation Reports. The Task Summary screens (Analysis Group -> Website Data Dropdown -> Task Summary) will also generate these on-demand.
Spotlight Focus suggested topics
March 23, 2023
  • Market / Spotlight Focus UI Enhancements. Users can now access the Link Flow Finder directly from the Anchor Text tab in both the Market Focus and Spotlight Focus screens.
  • Link Neighborhood Recommendations. The Link Neighborhood tasks and recommendations are now updated with more detailed descriptions of the Link Neighborhood.
  • Query Score Breakdown. The Query Score Breakdown now shows 0-100% instead of 0-1 for all algorithm scores. The Query Score Breakdown screens can be viewed from the Combined Search Listing screen (Analysis Group -> Actions Dropdown -> Simulate a Search) by clicking the Query Score percentage next to any ranking result.
March 16, 2023
  • Market / Spotlight Focus UI Enhancements. Users should see their Basket tab open by default now. The # of occurrences on the 1st tab is now sorted by default.
  • Analysis Group Setup Emails. All users will now receive notices of Analysis Group setup issues. Previously only Solutions Engineers received these emails. These can be turned off on the Account Settings -> Email Notifications tab.
  • Similar Words System Now Widely Available. Our Similar Words system, which models Google's Hummingbird (query expansion), is now editable by all users. This can be enabled / disabled and edited on the Analysis Group -> Settings tab.
Februrary 7, 2023
  • Major Spotlight Performance Enhancements. Users should see their Analysis Group setup times decrease by 60%-80% due to a major performance enhancement in the Spotlight algorithm, which taps into new software and hardware to generate topic clusters at scale.
January 20, 2023
  • Delete All Teams + Analysis Groups: on both the Team Listing and Team screens, there is a new option to delete not only the Team, but also all underlying Analysis Groups. Useful for cleaning up Analysis Groups that were only used on a one-time basis.
January 14, 2023
  • Inbound / Outbound Link Neighborhood Algorithm: previously added as Boost Factor. Now available on Task Listings and Analysis Group radar plots. Measures avg. link flow per page. Low values are showing correlation with decreased rankings.
  • Advertorial Algorithm: previously added as Boost Factor. Now available on Task Listings and Analysis Group radar plots. Measures ratio of external / total editorial links. Designed to catch signatures of guest blogging or advertorial posts that are designed only to produce link flow share to a specific external site. High values are showing correlation with decreased rankings.
  • "Paid" Links Algorithm: previously added as Boost Factor. Now available on Task Listings and Analysis Group radar plots. Measures ratio of external non-editorial / total non-editorial links. Designed to catch signatures of paid links or link selling. High values are showing correlation with decreased rankings.
december 3, 2022
  • Competitor Backlink Reports now include "First Seen": available on Task Listings, for any AHrefs Domain Rank or Semrush Domain Rank optimizations, directly above Task Descriptions, now sortable by each link's first seen date.
  • Canonical URLs now generate Tasks in each Analysis Group: Analysis Groups with a target URL that points to a canonical URL will now generate tasks.
November 17, 2022
  • Competitor Backlink Reports now available in a clean XLS format: available on Task Listings, for any AHrefs Domain Rank or Semrush Domain Rank optimizations, directly above Task Descriptions.
  • 22% increase in Particle Swarm Optimization accuracy: we've improved the machine learning models which calibrate your search engine models (Analysis Groups).
August 19, 2022
  • Ahrefs Domain Rank: the existing Domain Rank Boost Factor was renamed to Ahrefs Domain Rank.
  • Semrush Domain Rank: added Semrush Rank for Website as a new Boost Factor.
July 2, 2022
  • Fixed Related Entities Bug: related entities in the Spotlight Focus screen were missing links to the Knowledge Graph.
  • Added Ignore Backlinks Option: typically used when you are doing A/B testing and have a test site that is included, but also has lots of external incoming links from the same TLD production site. Found on the Website Settings screens.
June 25, 2022
  • Added dedicated internal DBPedia servers: this performance enhancement will greatly increase the capabilities of the entity-based knowledge graph algorithms within your search engine model.
  • Performance Improvements on Analysis Group Screen: this performance enhancement will greatly decrease the load times of Analysis Group screens.
June 21, 2022
  • Added Delete button on Analysis Group: you can now directly delete any Analysis Group on the AG screen, located under the Actions drop-down.
June 13, 2022
  • Added Billing Emails: on the Account Settings screen, on the Invoices tab, you can now specify additional billing emails that will receive a copy of the quarterly invoices.
  • Performance Enhancements: we've added servers and enhanced the queuing algorithms to better serve our users.
  • Activation Bug: we've fixed a bug related to activating new Market Brew accounts.
June 1, 2022
  • Added No Pricing Tier Option: when creating Market Brew accounts, Resellers now have the option to disable any pricing tiers, and instead just pre-load the crawls / credits into the account.
April 30, 2022
  • Added Analysis Group Button: on the Team and Account Settings screens, an "Add Analysis Group" button has been added.
  • Added TSEE Ranking: on the Analysis Group screens, on each site data row, the Target Search Engine Environment (TSEE) ranking will now be updated to show both the the current ranking position and the ranking position at time of calibration.
April 30, 2022
  • PSO Score on Analysis Groups: on all Analysis Groups, Analysis Group Listings, and Team screens, the Particle Swarm Optimization (PSO) score can now be found. A scale from 0 to 100% was created to inform users how well the model was calibrated compared to other models in the Market Brew system.
April 23, 2022
  • Allow Manual Setting of Target URL: needed for A/B testing where you have a Test Analysis Group that is copied from the Production Analysis Group. Target URL can now be set once the production site has been swapped out for the testing site.
March 22, 2022
  • Notification Preferences: we have updated the notification preference system, now found on the Account Settings screen under "Email Notification Settings"
February 15, 2022
  • Added Location Algorithm: we have added a Location Algorithm as a Boost Factor. This will be enabled whenever there is a location entity detected within the query, and each page will detect location entities and create a centroid (Lat / Lon) based on those entities. This is particularly helpful with local SEO situations.
december 1, 2022
  • Evergreen Chrome JavaScript Engine: we have updated the JavaScript parsing engine to automatically follow the upgrade path of the the public Chrome browser. Upgraded rendering as well as new Web Core Vitals scripting are the biggest features, with internal performance enhancements as well.
  • UI Refresh: Market Brew has updated its user interface to a new, modern, high-tech dashboard, leveraging the latest in client-side JavaScript and asynchronous processing. Default settings have been chosen: dim mode + slim-plus menu, top-bar theme colored, green, profile mode popup. Users will be able to select their own settings, like dim or dark mode.
  • Task System Refresh: Market Brew’s current task system will be expanded significantly and pushed to the Customer Home (top screen). While the previous task system relied on static alerts (shown on the SEO Alerts tab on Dashboard, Scoresheet, et al.), the new task system will rely heavily on the Biggest Weaknesses engine that drives the Simulation and Conversation Reports, but will also draw off of the existing Predictive Alerts and other subsystems. Tasks will introduce the concept of Categories, which will enable matching to different Team Roles (see below).New screens: Task -- shows the task, the category, the Query Potential, and options to complete the task. Workers will be able to mark the task complete, and also log how many hours they worked on the task. Market Brew may optionally add estimated hours to complete the task as well. Task Listing -- shows the list of tasks per Analysis Group. Includes Task hours / completed / assigned. Filterable by Category and Snapshot Date. Sortable by Query Potential. Team Tasks -- shows the list of all tasks per Team. Includes Task hours / completed / assigned. Filterable by Category and Snapshot Date. Sortable by Query Potential.
  • Market Brew Teams: Teams enable accounts to set up and manage a task-based system that will be shown on the Home Screen of the new UI dashboard. Parent accounts will be able to create the following: Teams -- defined by Analysis Groups (AGs), each Team will give access to a list of AGs to its Workers. This will replace the current authorization scheme of assigning individual authorizations per AG. Existing workspaces will be converted to a Team. Teams will also be able to define specific Tasks that will be filtered out. Workers -- all child accounts will be converted to Workers. Workers can be assigned to multiple Teams, and will be assigned to one or more Roles. There will be a special type of Worker called Team Lead which will be assigned to a Team, and will have admin privileges for that Team (add/edit/remove Tasks/Workers) Roles -- Predefined Roles will exist out-of-the-box (Off-Page, On-Page, etc…) but additional Roles can be created and assigned as necessary. Parent Accounts or Team Leaders can assign Worker Roles as necessary, but only Parent Accounts will be able to create / edit Roles.

    The following Widgets will be shown per Worker on their Customer Home screen:

    Target Sites -- links to Website Dashboards

    Top Tasks -- for all Teams that Worker is assigned to, one widget will show # of Tasks available, and another will show top list of Tasks, with links to the individual Task screens

    Tasks Completed (Progress Bar) -- shows how many tasks have been completed during current Snapshots. Tasks will be archived during each new Snapshot, and progress bar will reset according to new tasks that will be created. Filterable by Date Range.

    Tasks Completed (Snapshot History) -- by default, shows any tasks that have been marked completed in the current Snapshot. Tasks will be archived with each new Snapshot, and the completed list will be reset. Filterable by Date Range.

    All of the above can be filterable by selecting one or more Teams.

    New Screens: Add / Update / Delete Teams, Add / Update / Delete Workers, Add / Update / Delete Roles
November 15, 2021
  • Time Zone Selection: all accounts can now select their preferred Time Zone, located on the Account Settings screen.
  • LD+JSON: on the Spotlight Screens, a new (Schema) tab can be found that automatically generates the LD+JSON schema for that page.
  • Spotlight Author: on-demand version of Spotlight added to main menu and Spotlight screens.
  • Particle Swarm Optimization 2.4: PSO 2.4 continues enhancements to provide speed and reliability to the self-calibration process.
October 24, 2021
  • Particle Swarm Optimization 2.3: PSO 2.3 continues enhancements to provide speed and reliability to the self-calibration process.
October 15, 2021
  • Particle Swarm Optimization 2.1: PSO 2.0 introduced the concept of broad solution tree exploration, and with that came the valued added opportunity that a user could monitor this process and manually add and/or remove sites to and from the Analysis Group (AG) with ease. However, users were not yet able to see which solution tree branches have been tried, and which ones worked the best. PSO 2.1 introduces this ability. Along with it, it gives the users the ability to view Boost Factors, Domain Correlation and PSO score for each solution tree branch evaluated.
  • Tried Combinations: Users can now access the list of solution tree branches that were evaluated in the self-calibration process. From the Setup screen → View Failed Sites Screen, users can now click on “View All Tried Combinations”. When a user clicks on “View All Tried Combinations”, a list of each combination, ordered by PSO Score, that was tried during the self-calibration process appears. The list is filterable by Website. Websites and Boost Factor names are ordered alphabetically. Along with the ability to see which combinations bring out the best boost factors, users can select a specific combination by clicking “Force This Combination”, and immediately reconfigure the AG.

    Note: One of the many PSO 2.1 performance improvements short circuits the PSO process once at least 3 combinations with a sufficient PSO score (this is constantly being updated alongside the PSO score calculations updates described below) and a 100% Domain Correlation are evaluated. Not all combinations are guaranteed to be tried, so a careful inspection of this list may reveal a better combination to try. Market Brew continues to evaluate this algorithm and may elect to broaden or narrow the tree search in the future.
  • PSO Score Calculations Updated: PSO 2.0 had combinations that scored better without all URL paths matched, simply because they had more sites in the model. PSO 2.1 instructs the self-calibration algorithm that ALL domains and if possible ALL URL paths are matched, at the expense of shrinking the AG, with a minimum # of sites set at 3.
October 24, 2021
  • Market Brew Spotlight 1.0: Market Brew Spotlight will bring together a number of new Natural Language Processing (NLP) technologies and a Market Brew Knowledge Graph to provide the user enhanced content abilities within the model. Among other things, Market Brew Spotlight will perform named entity extraction, including entity detection, name resolution, and entity disambiguation.
  • Market Brew Knowledge Graph: The Google Knowledge Graph is a knowledge base used by Google and its services to enhance its search engine's results with information gathered from a variety of sources. The information is presented to users in an info box next to the search results. These info boxes were added to Google's search engine in May 2012, starting in the United States, with international expansion by the end of the year. Google has referred to these info boxes, which appear to the right (top on mobile) of search results, as "knowledge panels". The information covered by Google's Knowledge Graph grew quickly after launch, tripling its size within seven months (covering 570 million entities and 18 billion facts). By mid-2016, Google reported that it held 70 billion facts and answered "roughly one-third" of the 100 billion monthly searches they handled. By May 2020, this had grown to 500 billion facts on 5 billion entities. 

    There is no official documentation of how the Google Knowledge Graph is implemented. According to Google, its information is retrieved from many sources, including the CIA World Factbook, Wikidata, and Wikipedia. Market Brew will take a similar approach and use data from these sources to build its own in-house Knowledge Graph. The Market Brew Knowledge Graph will be built using Lazy Initialization, giving the platform the ability to maintain the same precision as Google without having to scale all at once. The Knowledge Graph will be accessible from Spotlight screens, linked whenever a corresponding entity is highlighted.
  • Market Brew Topics: Using Spotlight, a Market Brew Topic will be created per Webpage. In the future, Market Brew Topics may be broken down into both per-Webpage as well as per-paragraph, to mimic Google’s Centerpiece Annotation. In addition, a future Topic Focus Score may be created to show users how focused their content is, based on the overall Topic Focus. The Market Brew Topic will be shown on the Spotlight screen for each Webpage.
  • Market Brew Topic Clusters: Using the Webpage’s Market Brew Topic and leveraging the new Market Brew Knowledge Graph, the system will provide the user the opportunity to discover new content that should be added to their Website. This content will include a proposed snippet that should be included within the Webpage, as well as anchor text that can be used to link to proposed additional or existing Webpages about these other Market Brew Topics. Market Brew Topic Clusters will be shown on the Spotlight screen for each Webpage, as well as in a dedicated “Missing Topic Clusters” screen on a per-Website basis.
  • Expertise Algorithm (E-A-T): An Expertise algorithm has been created that can score a particular Webpage based on the perceived expertise of the author. Based on the Market Brew Topics, the Market Brew system can collect all of the details from every related entity and compare the writing on that Webpage to the body of content from all related entities. The more coverage that the content has, the higher the Expertise score will be. The Expertise score will be shown on the Spotlight screen for each Webpage, and will be included as a standard Boost Factor.
  • Spotlight Focus: After each Webpage is structured using Spotlight, the Market Focus algorithm has been upgraded. Instead of using a keyword density “bag of words” approach to generating initial meaning from a Webpage, the Spotlight results will be used. The incoming anchor text part of the calculation would remain the same, with the final Spotlight Focus Basket being similar to the Market Focus Basket. A Spotlight Focus Score would be generated wherever a Market Focus Score is generated, and used as a new optional Boost Factor. The Spotlight Focus screen will be accessible anywhere the Market Focus screen is currently accessible.
October 21, 2021
  • Particle Swarm Optimization 2.2: PSO 2.2 continues to refine some things, notably it’s PSO scoring calculator, some performance optimizations, and some screen optimizations.
  • Tried Combinations: The Tried Combinations Screen is now updated. It has more information to make better decisions on which one to pick.
  • Self-Calibration Report: The self-calibration report will now show only one row (the row that the PSO process picked). Previously it would show all of the failed attempts before it. Since the last iteration of the PSO process is now selecting which Tried Combination to use, there are no failed attempts before it.
July 22, 2021
  • Adding New Boost Factors: some algorithms have been added to the list of Boost Factors that were previously hard-coded and shown on the Webpage Scorecards.
July 2, 2021
  • Added Proxy Servers: on the Website Settings screens, users can now enable proxy crawling and select various proxy regions.
  • Core Web Vitals: We are releasing Core Web Vitals as additional ranking signals to track. We are able to do this at scale (so you can see how your competitors data compares) by injecting the core web vitals scripts on every page that Market Brew crawls.
June 30, 2021
  • Added Similar Words System: using neural networks, we are now detecting and injecting similar words in queries. New screens are now available for Resellers to adjust this system, including adding/removing new similar words and adjusting the similarity between those words.
may 29, 2021
  • Added Chrome's Blink Rendering Engine: we are now using Chrome's own Blink Rendering Engine, which enables crawling of all JavaScript driven sites.
March 13, 2021
  • Escaped Fragments: Market Brew has done a major upgrade in how it crawls escaped fragment sites. Market Brew is now able to handle a site like, where the page doesn’t reload, but various parts do. Market Brew typically throws away fragments in URLs (the “#” and anything to the right of it), as has been the conventional wisdom and guidance from Google for 15 years. The reason is that this part of the URL is never sent to the servers. This part is considered client-side only, but with technologies like ReactJs and Gatsby, these fragments are now driving a large part of the state of a user’s interaction with a website.

    There are now a large amount of sites that are using these escaped fragments (the “#!” in the URLs). Because Market Brew has invested into building a state-of-the-art Javascript rendering engine, built upon the popular Selenium open-source project, it was well-positioned to take advantage of the understanding of these escaped fragments. Market Brew will now treat any URLs that have a “!” (exclamation mark) directly after the anchor (#) as a separate page.

    Here’s an example of this in action:!/menu/category/viewall/

    Notice that you can click on “Chicken” and the page doesn’t reload (even though the entire user experience does). In the past, Market Brew would not treat these URLs differently, but now it does. Able to render the page exactly like a user would see in Chrome, Market Brew’s headless Chrome browser brings our rendering investment full circle.

    Note: any sites that are affected by this upgrade will take advantage of these new enhancements immediately after re-crawl.
  • Particle Swarm Optimization: Market Brew is running a new type of calibration algorithm, currently in shadow mode. Users will not see this, however, over the next few weeks, we will be collecting information on how well the new algorithm performs and will fine tune it before releasing it into the system.

    Currently, we use the no duplicate rankings algorithm as the correlation set. This has a nice balance between removing/adding sites, and exposing exact landing pages in the models.

    Recently some of our clients have expressed the desire to have more exact landing pages show up. This means that the models would potentially cause more removing/adding sites during the (slower) calibration process. The way to fix this is to have the correlation use the duplicate ranking algorithm, which is a more fine-grained correlation set. It will be more “picky” with which sites are in the group, because the constraints will be tighter. The duplicate ranking algorithm is much more processor intensive, as you can imagine, with orders of magnitude more data to compare.

    To this end, we’ve been working on performance optimization with respect to our Particle Swarm Optimization system. We’re happy to announce that we’ve achieved speedups of 10x over the past few months, and because of this, the ability to try a more intensive optimization algorithm became available.
  • Workspace Screen Fixes: There were a number of bugs related to the workspace screen filtering and adding/removing of sites. These have been fixed. We’ve also added a few new filters and will look for feedback before adding new features to this screen.
  • Lucene Filters: We’ve made some great enhancements to the Lucene Query Parser engine. One situation, where a client was trying to calibrate against the keyword “balisong”, surfaced an irregularity that we never saw before. 

    A site that was ranking very well for this keyword, was not showing up at all in our models., as it turns out, had only one type of spelling of this keyword: “bali-song”. The Standard Tokenizer in Lucene was taking the hyphen and treating it as a delimiter, making the matching of “balisong” against “bali song” impossible (note that it had no problem the other way around, matching “bali song” to “balisong”).

    So we introduced Lucene’s Word Delimiter Filter in place of the standard tokenizer. Our early tests are showing a remarkable improvement in these situations.
  • JavaScript Frameworks: Our JavaScript Chrome rendering engine now supports the following five largest JavaScript frameworks: React, Vue, Angular, Ember, Backbone.js. In addition to the specific support of escaped fragments, which are seen in some of these frameworks, Market Brew is now rendering the snapshot views of these sites at a much higher success rate. A tremendous amount of enhancements have been made in this area, and Market Brew continues to be at the forefront of JavaScript development.

    Note again that if you suspect that one or more Analysis Groups are affected by these enhancements, I would highly suggest that you take advantage of the new features as soon as possible.
February 23, 2021
  • Re-Crawl Entire Analyis Group: we've added the ability to mass re-crawl an entire Analysis Group. Simply click on the new action in the Actions dropdown.
  • Anchor Text Distribution: We have greatly enhanced our keyword research tools. To select the Anchor Text Distribution screen, click the link in the data dropdown button underneath the site you wish to research. Every word in every anchor text in the site is evaluated and associated with a specific amount of link flow. We call this “Associated Link Flow Share”. This is the relative amount of ranking power that each link transmits to a target page. Some links are more important than others, and this metric gives us that meaning. Click on the “Show Competitors” checkbox and you instantly switch from the target site’s keywords to it’s competitor’s keywords. Any site that appears alongside the target site, in any Analysis Group, is included. Therefore, the more Analysis Groups the better the keyword research becomes.
  • Link Flow Finder: As you know, one of the most important features of Market Brew is the Link Flow Distribution analysis. We've had a lot of success stories from brands utilizing the Link Flow Distribution analysis to boost traffic to large portions of their sites, but this process has been far from automated. One of the biggest hurdles for this that we've seen is having the knowledge on the team to know where to strategically add Link Flow to certain pages in a site. With our new Link Flow Finder feature, this hurdle should be a thing of the past.

    The Link Flow Finder works like this: for each target page, the system takes the already calculated Market Focus Basket of phrases assigned to that URL, and uses those phrases to do a proximity search on all the other pages in the site. Each resulting text fragment / snippet, with these key phrases highlighted, is then presented to the user with the corresponding page that it is on. Some Market Brew magic happens when our models find that the target site’s Market Focus Basket is NOT the best Market Focus Basket. When this occurs, it is the outperformer site that is used to harvest the keywords to search against. This is particularly helpful when a target site doesn’t know what the correct keywords are yet. Therefore, the more Analysis Groups the better the Link Flow Finder becomes.

    The user only needs to click "Add Link Flow" under any page in the Link Flow Distribution screen, and then add links to the target page using the highlighted snippets. The snippets are even ordered by the most potential Net Link Flow Share, making this process a very powerful approach to shaping the Link Flow Distribution and Balance.
  • Domain Rank Finder: Another huge enhancement in the model is our new Domain Rank guidance. The Domain Rank Finder works like this: backlink sites that link to two or more of the target site's "competitors" but NOT to the target site itself, are suggested as potential link partner sites. What is unique about the way we do this is that we use the other sites that accompany the target site in each Analysis Group as the "competitor" list. This allows us to pick sites that sometimes are not a direct competitor (and therefore more likely for a link from the same source to be easier to acquire). Therefore, the more Analysis Groups the better the Domain Rank Finder becomes. These potential links are ordered by "exclusions" first (how many "competitor" sites get links from these sites) and then they are sorted by the Net Link Flow Share that a new link on that site would provide (how powerful the backlinks on that page are).
  • Updates to Existing Reports: The Market Brew Conversation report is now generated immediately after Analysis Group setup! No need to re-crawl each site at least twice before generating the report. This will further automate the process of prospecting, and for our A La Carte customers, it will generate a link to the report immediately upon purchasing the product. All trend analysis on the Market Brew Conversation report is removed for the first report (as there is only one snapshot of data at this point). All competitor “Biggest Weakness” and “Trend Analysis” sections have been removed: the Conversation is solely focused on the target site now, resulting in more focused and less noisy content. Both the Link Flow Finder and the Domain Rank Finder recommendations will also be listed on the Simulations screens as well as the automated Conversation reports, when applicable. You should see these start to show up as your sites are re-crawled and new reports are generated. We believe that these two improvements to the Market Brew product make it both incredibly powerful and easier to use. We are constantly striving towards full automation and these enhancements make good on that promise. In addition to the two major enhancements mentioned above, we have added detailed analysis of the HTML scoring, including featured text fragments, giving users transparency into the underlying Lucene Query Parser scoring. We have also added detailed analysis of the URL Host and URL Path much in the same way we show the META Title analysis, complete with highlights and recommendations. As you will see, there have been some major cosmetic upgrades to both the Simulations screen and the Conversation report, and improvements in rendering on both mobile and desktop browsers.
  • Additional Rank Tracker Databases: We have added additional databases that have become available from SEMRush’s API for Analysis Group setup. In addition, we have added the country name to the dropdown list, before we were only specifying country code.
  • Updated Chrome JavaScript Engine: we have updated the JavaScript parsing engine to the latest stable version of headless Chrome (Google Chrome 88.0.4324.182). Upgraded rendering as well as new Web Core Vitals scripting are the biggest features, with internal performance enhancements as well.
  • Two different account types now available: Both per-crawl and per-Analysis Group crediting are now available to Resellers.

Previous release notes from 2006-2020 have been archived and are available on request.