Disrupting Transcripts: Spike In “Deflections” on Tesla’s Latest Call

In our final earnings week blog post discussing our new, machine learning-based transcript Smart Summary™, we’re taking a look at the electric car maker, Tesla. The company reported earnings and held its regular quarterly call after market hours on Wednesday, July 24, 2019. The stock dropped immediately upon the results release, and continued dropping during the call and into the next day.

While every quarter at Tesla has many puts and takes (like international shipments, tax credits, specific model volumes), what really stood out for us when we applied the Smart Summary™ was the really serious increase of “deflection” instances quarter over quarter. Our tweet on the topic got almost 10,000 views in a few hours:

 

So what is “deflection?”

Our new transcript Smart Summary™ tool has two underlying mechanisms. One is based on a specially trained machine learning system in which we, mostly former buysider product managers, trained the system to parse and classify sentences into a large training data set across industries and companies. This tool automated what we used to do during earnings season, and will only get better as our users submit direct feedback. Transcript Classifications are not mutually exclusive; a sentence might often be a Guidance and a KPI sentence at the same time. 

The second mechanism powering Smart Summary™ is natural language processing. We have had overall sentiment reports for management and analyst teams, and keyword surfacing applied to transcripts in the product for a long time. What we’re introducing with the Smart Summary™ is the scoring of specific sentences (positive/neutral/negative) across all Classifications (so a sentence might be “positive Guidance” and “positive KPI” at the same time). The other NLP application is what we call “deflection.” We dug through the academic literature and identified a specific lexicon that has shown to lead to future negative corporate events. 

Looking at the Deflection summary view in the Tesla call, we see quite a bit of the usual enthusiasm and grand visions from CEO Elon Musk.

We see the recently appointed CFO discussing the challenges in the quarter-to-quarter business:

It was most interesting to us that the remarks by the outgoing CTO (a surprise departure) were also picking up as deflection:

The final statement on the call by the CEO was also marked as deflection: will Tesla beat the industry economics?

To find out how Sentieo can help with your process, please get in touch

New call-to-action

Disrupting Transcripts: Short Video Example on UPS

Last week, just in time for earnings season, we released Sentieo’s machine learning-based transcript Smart Summary™ feature. In short, we automated what we used to do as buyside analysts: reading through transcripts and classifying information.

This week, we’ve looked at the newest transcripts from Goldman Sachs and Haliburton in detail. Yesterday, we shared an in-depth video discussing Whirlpool: from redlining the 8-K, to looking at the stock price action post-call, to its Smart Summary™. 

Today’s video is shorter, as we only focus on the Smart Summary™ aspect and we discuss UPS. We were able to extract the important information in about 2 minutes.

The Smart Summary™ is one of two major ML-based product releases we had this month: take a look at our synonym suggestion tool as well. Your searches can be a lot more productive with this dynamic addition to our existing large synonym/acronym dictionary.

If you are ready to test out our platform, please get in touch.

New call-to-action

Introducing Sentieo’s New Document Search User Interface

Sentieo users are now able to try the beta release of our brand new Document Search user interface (UI). While it may seem familiar, the UI has been rebuilt using modern tech such as React. This new build enables significantly less memory consumption and greater responsiveness across the board. Plus, we’ve extended Sentieo’s sentiment and natural language processing capabilities to include a host of new AI-powered features detailed here.

AI-Powered Features

 

Smart Summary™

With Sentieo’s new, ML-driven transcript Smart Summary™, we enable analysts to pick up the important information across multiple classifications, such as guidance and KPIs.

To trigger Smart Summary™, users can simply click the “brain” icon on any transcript. To learn more about Smart Summary™, read this post.

 

 

Query Autocomplete

Our data science team has built an algorithm to identify key terms and topics for all companies within our database. Similar to a Google search, when a user starts typing inside of the query box, Sentieo will automatically populate suggested terms for you based on actual terms used by the company. This makes it much easier to find exactly the keywords you are looking for. Autocomplete intelligently populates based on which tickers and watchlists you have selected in the ticker box.

 

Suggested Synonyms

We are very excited to announce a giant leap forward for our Synonyms system. We have been using an NLP + machine learning-based algorithm for the last couple years to help us internally define and expand our deep synonym library. We are now very excited to open up that system to Sentieo users with our new Suggested Synonyms feature.

Using Suggested Synonyms, Sentieo is able to parse your query and automatically suggest possible synonyms to expand your search. While all of these synonyms may not be pertinent to your search, the function lets you quickly decide which you would like to add to your search. Just click the suggested synonym and it will automatically be added to your search query:

General Document Search Improvements

The Sources Selector

When we looked at usage analytics, we realized that the Document Type filter was the most used filter by a large margin. To make this filter more accessible and easier to use, we’ve made it a part of the main search interface. Additionally, we have renamed the filter to “Sources” to more accurately represent the functionality.

We have also added a quick “only” selector to the right of each source that lets users quickly focus on a single source. 

Additionally, click the > icon at the far right to load up the more granular filters:

All additional filtering is now available with the Filter Button:

 

Configure and Control Your Results Pane

We’ve added the ability for users to easily resize the results panel. Simply drag the divider or use the maximize/minimize button. As you expand, the application will automatically switch to a table view when enough space is present. 

While in the table view, you can also control what columns are visible. Currently, the list of options is limited but the Sentieo team will be adding to this over the coming months. 

For smaller screens, we’ve also added the option to toggle between split pane and results-only views.

 

Company Summary

If you run a search with a ticker but without a query, the Sentieo initial results page shows you all company documents, as well as a company summary and share price performance. We have taken some key metrics and links out of our EDT Summary and provided them within Document Search. Additionally, users can see a historical stock price graph in the company summary.

 

Improved Document Viewer

We’ve made serious upgrades to our Document Viewer as well. You will notice faster load times for documents across the board. Additionally, we have relocated the highlights and the index panel to the right of the document.

We have also relocated the Redlining function to the top right of the document viewer.

 

To try the new Document Search interface for yourself, please log in, or get in touch!

 

New call-to-action

Disrupting Transcripts: Applying Sentieo’s Smart Summary™ During Earnings Season

After last week’s release of Sentieo’s machine learning-based Smart Summary™ for transcriptsthis week we will be highlighting exactly how to use this powerful new feature to ensure that you don’t miss important information this earnings season, and that you also save time.  

The Smart Summary™ tool classifies a document’s language based on several categories such as Guidance, KPIs, and company-specific key terms (like product names) into a convenient, time-saving format. Further, it applies an NLP sentiment filter to highlight positive and negative language. 

Smart Summary™ is the latest machine learning and natural language processing-based tool released by Sentieo this year. We’ve also released a dynamic synonym suggestions feature for Document Search based on our own, internal synonym search tool. Additionally, we added company-specific autocomplete suggestions to our search. And earlier this year, we rolled out Table Explorer, a flexible and auditable tool that identifies, chains and visualizes tables in SEC filings. 

During the next few days, we will be selecting and publishing our highlights from current transcripts to show how effective Smart Summary™ can be in your research workflow. (Stay tuned).

For example, if we look at the transcript from the Goldman Sachs call last week (below), we can pick up essential corporate and macro information (color-coded as needed), under the Financials and KPIs classification. Scrolling down in the Smart Summary™, we see record assets under supervision: good news in green, versus slower Chinese GDP growth in red, a couple of paragraphs down. The full context is right there as well. 

In addition to viewing by Classification group, users can quickly scroll through all positive and negative highlights, regardless of Classification. 

In the Key Terms field, we catch up on Marcus, Goldman’s growing consumer banking franchise. We can see a mix of positive and neutral statements. 

For the next few days, the blog will be highlighting exactly how we’re applying this powerful new functionality. To try it yourself, please get in touch

New call-to-action

Disrupting Transcripts: Introducing Sentieo’s Smart Summary™

Just in time for earnings season, Sentieo is releasing Smart Summary™, a whole new way to read transcripts. 

We all know the problem that analysts and portfolio managers face four times per year: multiple companies reporting on the same day, which creates impossible scheduling conflicts, work overload, and forced prioritization, which then leads to costly missed information.

Analysts might dial into one or two consecutive high priority calls, and then go on to read another four to eight transcripts later in the day, depending on coverage. These transcripts are always read chronologically, starting with the boilerplate language on top, and ending with the analyst questions. 

Until now. 

With Sentieo’s new, ML-driven transcript Smart Summary™, we enable analysts to pick up the important information across multiple classifications, such as guidance and KPIs, while skipping over the “great quarter, guys!” parts. The segments based on these classifications can either be seen highlighted in the transcript itself or extracted, with context, in a special field. The text is also run through an NLP screen and color-coded in green or red, if needed, with enabled clickable user ratings that will continue to improve the accuracy of the learning system. In addition, Smart Summary offers keyword extractions, enabling readers to view text related to these surfaced key terms.

These features empower analysts to get the essential information in minutes, saving valuable time during the busiest periods of the year. The benefit of the ML-driven Smart Summaries does not end with faster information processing or expanded coverage; it also ensures that analysts are not missing important information on their last few transcripts at the end of the workday. 

Below we are looking at Facebook’s latest quarterly call transcript, and scrolling down to the Guidance classification text:

Below we are looking at the same Facebook transcript, but reviewing the Whatsapp key term text:

Another aspect of Smart Summary is “Table View,” which enables active sorting of sentences by the Classification score (ie. “which sentences have the most guidance language?”). This adds another layer of improved accuracy and productivity to the analyst workflow. 

The Table View is also available for surfaced keywords. Analysts can see useful statistics, like whether a keyword is new, or whether it got dropped. They can also see whether usage, by either management or analysts, is going up or down. This type of quantitative insight for keywords was previously so labor-intensive that it was rarely done, and only for one or two key terms at most.

 

How did we build Smart Summary™?

We spent countless hours training a machine learning model on a very large number of transcripts across all industries, effectively automating what an analyst would do. Once the model was up and running, we spent a lot of time on corrections and fine-tuning for this official release. 

Smart Summary™ is a blockbuster new addition to the Sentieo platform that serves our overarching mission: augmenting human decision-making through the latest technological tools. The release is a part of the latest ML/NLP update to our industry-leading Document Search, which now also includes the surfacing of synonym suggestions based on a machine learning tool trained on millions of corporate documents. (We will discuss this in an upcoming blog post).

 

To see how Sentieo’s complete workflow solution can work for you, please get in touch

New call-to-action

Missed Our “Ask Us Anything” Webinar? Here Are Your FAQs About Sentieo’s Dashboard, EDT, Screener, and Mosaic

Last month at Sentieo’s New York office, our Head of Research, Nick Mazing, hosted a two-hour “Ask Us Anything” live webinar. Nick answered questions from Sentieo’s current and prospective users, covering the many functionalities of the platform. In this post, we will share a few FAQs about Sentieo’s Dashboard, EDT, Screener, and Mosaic features.

Dashboard FAQs

How do I customize my Dashboard to fit my personal research workflow?

The Dashboard is the highly customizable screen that a Sentieo user can utilize to:

  • Monitor multiple financial and alternative data metrics for a list of stocks
  • See internal work being done on positions
  • See any documents coming in (such as new press releases or sell-side reports)
  • Get alerts on their saved searches

Extensive customization allows this screen to be just the way you need it to be, and it lets you save time with one quick look over everything that matters to you.

Mosaic FAQs

Mosaic is Sentieo’s alternative data module. Sentieo takes several data sets and does correlations against past performance to determine the optimal weights of these sets in the “basket.” We are completely transparent: you can see past performance (R-squared and hit rate), and you can see the basket weights.

Can you show us a couple examples on how to use Mosaic?

In this example, we looked at the individual brands mapped for Capri Holdings (formerly Michael Kors), Abercrombie and Fitch. We also looked at how investors could have avoided a 90%+ loss in Blue Apron shares by monitoring the Sentieo Index.

How did you come up with the Sentieo 11 stock picks in January?

Nick went over the popular eleven stock picks that we released in early January, which did really well. We look at the performance of the list, the alternative data that was used, along with the fundamental “overlay.”

Equity Data Terminal (EDT) FAQs

Sentieo’s EDT has everything you’d expect in terms of data for an individual equity, plus a lot more.

How do I try online “consensus” income statement models in EDT?

Nick showed how users can quickly enter their own assumptions about future growth, margins and multiples, and compare against the consensus. This function enables fast idea pre-qualification, saving you and your team time to focus on ideas that could “work” at a first pass.

We show a specific example, Royal Carribean, where we use our own estimates for sales growth, gross margin, and multiples to arrive at an internal price target above the current.

We also looked at the estimates/surprises and seasonality of returns for Royal Caribbean (RCL).

How do I use Plotter to visualize this?

Nick went to Sentieo’s data visualization tool, Plotter, to analyze further. He explored revenue and EBIT seasonality, compared it to the actual underlying cash flow, and then he calculated EBIT margin along with some descriptive statistics.

Revenue, earnings and other estimates are a popular feature in Sentieo’s Equity Data Terminal (EDT). We go over one example of the beats/misses function along with the classic “estimates revisions” page that shows changes in estimates by year and rolling NTM. Finally, we show how to export these to Sentieo Plotter, where users can overlay many other data sets.

Screener FAQs

Sentieo’s Screener is very powerful. There are thousands of possible options to screen on: not just by widely available financial metrics, but also by CAGRs and percentile ranks inside these metrics. Nick showed his “Large Cap Quality and Momentum” saved screen to illustrate.

Want to try using Sentieo yourself? Request a trial here.

Introducing Table Explorer: Enhanced Table Extraction and Analysis Across Time

At Sentieo, we are always striving to give fundamental managers the tools to make better decisions faster. Table Explorer (TableX), the next iteration of our suite of table analysis tools, was developed with this vision in mind.  

Model-building is an essential part of the investment process of fundamental managers. Models afford them the ability to develop a deep and differentiated understanding of the machinery of a business, via the interplay between revenues, expenses, and capital accounts. Models also allow for a parameterized flex of assumptions for establishing a differentiated view on a company’s earnings prospects and — by extension — the value of the business overall.

While modeling is critical to the process, the investment world is increasingly driven by algorithm-enabled trading and the large data science departments of multi-billion-dollar hedge funds. Therefore, managers are obligated to pay close attention to the areas where modeling and data visualization can happen faster, and more accurately.

At Sentieo, we believe in an augmented investing approach that uses machine learning and data science to add value to a human-driven investing process.

We’ve rebuilt our table suite from the ground up using a machine learning driven approach validated by trained financial analysts. The new TableX enables Excel-style visualization directly from company filings in a dual-pane window, showing both the Excel data and the source document side by side, enabling an immediate audit of every number.

We’ve added the ability to quickly plot discrete line items in the same window, enabling simple thesis exploration either as a complement or antecedent to further model work. We’ve also added inline calculations to impute fourth-quarter numbers from 10K data, common size analysis, and YoY or QoQ percentage changes. And we show a table of contents of all tables in a document to speed up navigation through company filings.

table explorer table extraction

Then, we integrated TableX with our Plotter app, to further facilitate regression analysis of fundamental metrics like revenue with other metrics like exogenous macroeconomic data or website traffic.

We also integrated TableX with our Notebook app to facilitate thesis sharing among a team or when reporting to a portfolio manager. These workflow tools help money managers develop ideas faster and stay more organized with their ideas.

Lastly, we facilitate export to Excel so that the audited financial data can be used as a backing database for your own models.  We include links back to the document in the export, and in future releases, we will enable “one-click” refreshing of base data sheets in models.

The end goal of Table Explorer is to give fundamental money managers the tools to observe-orient-decide-act faster, in order to win in a high velocity investing landscape. Get a free trial of Sentieo here, or tell us what you think of the new TableX!

4-Minute Videos on How to Use Table Explorer:

Navigating Table Explorer:

Common Size Analysis in Table Explorer:

Get a free trial of Sentieo here, or tell us what you think of the new TableX!

New call-to-action

Introducing Sentieo Capture: Clip and Save Document Content into the Sentieo Notebook

We recently introduced an enhancement to our document markup functionality: Sentieo Capture.

Within documents in Sentieo, you’ve had the ability to highlight, annotate, and tag your colleagues within important passages of text. Oftentimes, you’ve also come across important graphics such as charts or tables that you may have wanted to collect and save.

With Sentieo Capture, you can now:

  • Capture any portion of a document (such as a capex breakout slide from a company’s investor presentation).
  • Mark up the capture with shapes, arrows, and lines of text.
  • Save the captured area as an image (along with the source document) directly into your Notebook.
  • As with highlights, you can also add label(s) and annotations to your captures.


Why Should I Use Sentieo Capture?

While performing a deep dive analysis on a company, investment professionals are likely to face information overload, which can make it challenging to effectively manage all the information being consumed. Ultimately, effectively managing information and data becomes one of the key skills of a successful investment professional:

  • How do you collect information snippets from the documents you are reading, and then create a summary report?
  • How do you record the link (URL) to remember the source of the document along with your information snippet?
  • How do you organize the information you’ve collected so that you can easily retrieve it at a later time?
  • When collaborating with others on your team, how do you efficiently share and update this information?

You have probably used various tools to tackle these questions, including the Windows Snipping Tool, MS Outlook, MS Word, MS OneNote and Evernote. These are all great tools and have their own particular strengths, but they do not solve the a major problem in the investment professionals’ information capture flow: having to source, organize, remember, and be able to retrieve your research.

At Sentieo, we are building features to replace the fragmented information capture workflow to become a one-stop solution for managing, sharing, and retrieving the information you’ve gathered on a company or topic. That way you can come up with a meaningful conclusion using all of the information you’ve gathered, and not just bits and pieces. In this post, we’ll nudge you into trying a slightly different approach when doing your company research in Sentieo.

 

How Do I Use It?

The Capture feature can be triggered via the photograph icon located in the upper-right corner of the document that you are viewing. There is no real setup required other than making sure that you have installed the Sentieo Chrome extension, which you will be prompted to do if it is not already installed.

To enable capture mode, simply click the “Capture” button on the toolbar while viewing any document within Document Search or your Notebook:

 

Begin your capture by selecting the area you would like to capture and clicking and dragging:

A new screen with your captured area will appear, allowing you to:

  1. Name your capture
  2. Mark up your capture with the annotation tools menu
  3. Copy to clipboard, e-mail out, or save the image to your computer
  4. Add a label or annotation to your capture
  5. Save or discard your capture

Where is this information being stored, and what can I do with it?

All of your bookmarks, highlights, annotations (and now captures) are also stored and searchable within your Sentieo Notebook. This also means that your information is available from any computer with Google Chrome and an internet connection.

Like highlights, all of your captures are automatically sent to your Notebook and populated with the source document title and filing date, any relevant company tickers, and any labels and annotations. This automatic tagging makes retrieving your highlights and captures a breeze in the future.

Additionally, you can share selected research with the rest of your team and people outside of Sentieo. You can also view what others have shared with you.

 

We hope that Sentieo Capture saves you more time and keeps your research more organized. Give highlighting and Capture a try; you won’t want to do your Sentieo research without it! Please message us via customer support chat, or email nauman@sentieo.com with any feedback on the functionality that you would like to see added to the new capture feature! Haven’t used Sentieo yet? Get a free trial.

New call-to-action

sentiment

Introducing Sentieo’s Game-Changing Transcript Sentiment and Linguistic Analysis Features

Over the last year, the Sentieo team has been investing heavily in NLP & data science. We were very excited to release the Sentieo’s Transcript Intelligence Report, which is the first of many features built on a powerful new technology that lets us dissect documents and help you find insight.

There is signifcant amount of academic research that supports the idea that changes in sentiment among those in company management, as well as those working as analysts, do indeed have some relationship with stock returns.

Management is typically a negative predictor of stock returns. When management seems overly excited about the future, this may indicate that the stock will underperform. (See “Manager Sentiment and Stock Returns,” Journal of Financial Economics). On the other hand, when investor sentiment is low, subsequent returns are relatively high on a subset of stocks (high volatility stocks, unprofitable stocks, non-dividend-paying stocks, extreme-growth stocks, and distressed stocks), consistent with an initial underpricing of these stocks. (See “Investor Sentiment in the Stock Market,” Journal of Economic Perspectives).

Our Transcript Intelligence Report provides automatic analysis on how sentiment and keywords for a given company have changed across quarters. The analysis easily splits out Management Sentiment vs. Analyst Sentiment. Let’s use Netflix as our example.

First, we open up Sentieo Document Search and search for Netflix transcripts. We choose a transcript in the left hand navigation, and see a “View Linguistic Intelligence Report” button at the top of the document in the right hand pane.

sentiment

 

 

We’re taken to the full Transcript Analysis page, and can now dive into the analysis!

sentiment

 

Sentiment Scores

Sentiment scores are generated by tagging each word in the document as positive, negative, or neutral using a dictionary designed for financial documents and applying a formula to generate a score for each section. The formula for sentiment is below:

The number (typically a decimal value below 0.1) is a percentage that should be considered on a relative basis, rather than absolute.

Management Sentiment

The following chart shows sentiment for Management commentary and Q&A answers on eight quarters of earnings calls. Sentiment is calculated as positive word % of total words – negative word % of total words.

management

 

 

 

Analyst Sentiment

The following chart shows sentiment for Analysts from the Q&A section of eight quarters of earnings calls. Sentiment is calculated as positive word % of total words – negative word % of total words.

 

 

 

Management-Analyst Sentiment Spread

The following chart shows the historical spread between Management and Analyst sentiment.

The spread between the two would suggest divergences in sentiment. For example, if the spread turned negative, then it is possible that management has turned less bullish while analysts have remained bullish, thereby signaling the potential for management to disappoint in future periods.

 

 

 

Conference Calls Sequential Comparison Word Cloud

The comparison word cloud shows the biggest changes in average keyword frequency between the 2018-04-16 conference call and the 2018-01-22 conference call for NFLX.*

 

 

 

 

Analyst vs Management: Keyword Differences

This is another comparison cloud showing the major average differences between management and analyst keywords on the 2018-04-16 conference call.

 

Management Top 15 Keywords

This table shows the top 15 keywords by frequency referenced in the past eight quarters of earnings calls. This provides a visual view at a glance into what is being discussed on the earnings calls over time.

 

 

 

Analysts Top 15 Keywords

 

 

 

Sign up for a free trial of Sentieo to get the sentiment report for Netflix and any other ticker you follow.

New call-to-action

crypto interest rates Sentieo

From Crypto to Interest Rates: A Sentiment Analysis of Q1 2018 Earnings Calls

Today we’ve published our second Sentiment Analysis Report, which summarizes last quarter’s top keyword searches and provides detailed sentiment analysis across all industries. We used Sentieo’s brand new Transcript Sentiment Analysis feature to analyze earnings call transcripts and discover which topics companies discussed the most last quarter, versus the same quarter in 2017. (See our previous report here).

We also compared the sentiment of management and analyst sections of transcripts, and graphed these data points so you can easily see trends or discrepancies between the two. We publish these reports every quarter, so you can stay updated on information that could impact your investment decisions this year. Here are some interesting themes that came up in our research:

 

Sentiment Analysis

Management versus investor sentiment is diverging.

Our sentiment analysis on transcripts shows that a decoupling is taking hold between the language from company management and market participants. Management continues to be upbeat during earning calls and presentations, while sell-side analysts and investors are taking a more cautious stance. To learn more, download the full, free report.

 

Keyword Mentions

Two substantial highlights from the various themes we cover in this report are related to cryptocurrency and Trump.

While mentions of crypto have continued to ramp up, two companies in particular had a surge in references: IBM and Overstock.com.

IBM, with its Watson program and early involvement in the emerging fintech scene, is a recurring leader in the category. What’s new is that as concepts are maturing, bigger and more influential ecosystem players are now making moves: IBM recently revealed that it has been meeting with executives from commodities trading platforms, large corporations, and perhaps most importantly, central banks, to explore cryptocurrencies and blockchain in their operating models. (CoinDesk)

Overstock.com also made the headlines as possibly the first $1bn+ listed company dipping into crypto funding with an ICO (Initial Coin Offering). The stock fell sharply year to date (-40%), in part in reaction to the Securities and Exchange Commission starting an investigation on Overstock.com’s subsidiary that did the ICO. (Investopedia)

With this report, we are starting down the path of quantifying linguistic data. This report is a real use case of the exciting new features we recently released, like our Transcript Sentiment Report function, which is part of Sentieo Document Search.

Below is a sneak peek of the report: a sample page about the information technology sector.

To learn more about the companies, industries, and regions where crypto and other themes are being most discussed, download the full report, which covers this sector and many more. To find out more about how to run your own sentiment analysis with Sentieo, sign up here for a free trial.

 

New call-to-action