Earnings season compresses a large amount of information into a short review window. Research teams must track changing company narratives, analyst opinion, management developments, and emerging risks across multiple companies before results are announced. Reviewing every signal manually can make it difficult to determine which upcoming earnings events deserve immediate attention.
News sentiment can help reveal how the narrative surrounding a company is changing, but it is unreliable when treated as an independent investment signal. As discussed in Why News Sentiment Alone Is a Terrible Investment Signal, positive coverage does not necessarily indicate improving business expectations, while negative headlines may reflect risks that analysts have already considered.
Historical analyst-grade sentiment provides a second perspective. FMP's Historical Stock Grades data shows how the distribution of Strong Buy, Buy, Hold, Sell, and Strong Sell ratings changes over time. Comparing consecutive monthly observations can reveal whether analyst consensus became more positive or negative entering the pre-earnings review period, without treating that change as a prediction of an earnings beat, miss, or subsequent stock-price move.
The more useful review signal appears when these inputs are evaluated together. News sentiment may improve while historical analyst-grade sentiment becomes more negative, or sentiment may deteriorate while the analyst-grade balance improves. These conflicting movements create divergence that can help research teams identify companies requiring deeper pre-earnings investigation.
This article demonstrates how Financial Modeling Prep data can be used through Claude MCP to compare company-specific news sentiment with changes in historical analyst-grade sentiment before earnings. FMP supplies the underlying news and Historical Stock Grades records, while Claude applies the defined relevance filtering, deduplication, sentiment classification, grade-balance calculation, and rule-based comparison. Headline, source, publication date, and URL evidence is preserved for retained news records so the sentiment judgments remain auditable.
The resulting analysis classifies the available evidence as aligned positive, aligned negative, divergent, or review required, and then assigns a pre-earnings review-priority level. The thresholds and confidence rules used in this demonstration are author-defined operational rules applied consistently across the selected companies. They are not statistically validated cutoffs, probabilities, or measures of expected volatility.
The objective is not to forecast an earnings surprise, post-announcement price direction, or realised volatility. Instead, the analysis helps prioritize limited research time around companies where the public narrative and historical analyst-grade sentiment are reinforcing each other, moving in conflicting directions, or lack enough evidence for a reliable conclusion.
Key Takeaways
- Pre-earnings monitoring is more useful when company-specific news sentiment and changes in historical analyst-grade sentiment are evaluated together rather than as isolated signals.
- Divergence can appear when news sentiment improves while the historical analyst-grade balance deteriorates, or when sentiment deteriorates while the analyst-grade balance improves.
- FMP supplies the underlying news and Historical Stock Grades records, while Claude performs relevance filtering, deduplication, sentiment classification, grade-balance calculation, and rule-based comparison.
- Mixed or ambiguous news remains Neutral unless there is a clearly dominant company-specific near-term implication, while the underlying headline, source, date, and URL remain available for review.
- The analysis classifies evidence as aligned positive, aligned negative, divergent, or review required, then assigns an analyst-review priority.
- Classification thresholds and confidence labels are author-defined operational rules for this demonstration. They are not statistically validated measures of probability, volatility, or predictive accuracy.
- The output is designed to prioritize pre-earnings research. It does not predict earnings surprises, post-announcement stock direction, realised volatility, or trading outcomes.
Building a Pre-Earnings Divergence Framework
The analysis begins with companies scheduled to report earnings within the selected monitoring window. For this demonstration, the candidate universe consists of US-listed operating common stocks, while ETFs, mutual funds, preferred shares, warrants, blank-check companies, acquisition vehicles, and duplicate share classes are excluded.
Company selection occurs before alignment or divergence is calculated. Candidates are sorted by earnings date and ticker, then evaluated in that fixed order. A company qualifies for the final demonstration only when both the news signal and historical analyst-grade signal provide sufficient directional evidence. Screening stops once four companies qualify.
This prevents companies from being selected because they happen to produce an interesting aligned or divergent result. The screening output should show which candidates were evaluated, which failed the evidence requirements, and which four ultimately qualified.
The thresholds below are author-defined operational rules used consistently throughout this demonstration. They are not statistically validated cutoffs, probabilities, or measures of expected volatility.
Measuring pre-earnings news activity
FMP supplies the underlying company-specific news records, while Claude performs the relevance filtering, deduplication, and sentiment classification.
The news layer covers articles published during the 30 calendar days preceding the run date. For every retained article, the following evidence is preserved so the classification can be audited:
- Headline
- Publication date
- Source
- URL
- Article snippet, where available
Exact duplicate URLs, duplicate headlines, syndicated copies of the same underlying story, broad market summaries without a material company-specific development, and incidental company mentions are removed before calculating sentiment.
The 30-day period is divided into two equal windows:
- Earlier period: July 1 through July 15, 2026
- Recent period: July 16 through July 30, 2026
Each retained article is classified as Positive, Neutral, or Negative according to its likely near-term implications for company operations, demand, revenue, margins, financial position, management, regulatory position, or the upcoming reporting period.
A story is classified as Neutral when its implications are mixed, ambiguous, primarily factual, routine, or when favorable and unfavorable elements do not produce a clearly dominant near-term interpretation. Share-price performance and broader market direction do not determine company-specific sentiment.
For each period:
|
Sentiment Balance = (Positive Articles − Negative Articles) ÷ Total Relevant Articles |
Then:
|
Sentiment Change = Recent Sentiment Balance − Earlier Sentiment Balance |
Sentiment direction is classified as:
- Improving: Sentiment Change is at least +0.20
- Deteriorating: Sentiment Change is −0.20 or lower
- Stable: Sentiment Change is greater than −0.20 and less than +0.20
- Insufficient evidence: fewer than five unique relevant articles are available across the complete period, or either half-window contains fewer than two relevant articles
News activity is classified as Accelerating only when both conditions are met:
- the recent period contains at least two more relevant articles than the earlier period; and
- the recent-period article count is at least 1.5 times the earlier-period count.
Otherwise, news activity is classified as Not accelerating.
Only companies with sufficient news evidence and an Improving or Deteriorating sentiment trend proceed to the analyst-grade comparison.
Measuring historical analyst-grade sentiment
The second evidence layer uses FMP Historical Stock Grades rather than historical EPS-consensus snapshots.
Historical Stock Grades provides dated distributions across Strong Buy, Buy, Hold, Sell, and Strong Sell ratings. Because the available observations follow a monthly cadence, this demonstration compares the two consecutive observations dated June 1 and July 1, 2026.
For each observation:
|
Analyst Grade Balance = (Strong Buy + Buy − Sell − Strong Sell) ÷ (Strong Buy + Buy + Hold + Sell + Strong Sell) |
The change in analyst-grade sentiment is then calculated as:
|
Grade Sentiment Change = July 1 Analyst Grade Balance − June 1 Analyst Grade Balance |
Analyst-rating direction is classified as:
- Positive: Grade Sentiment Change is greater than 0
- Negative: Grade Sentiment Change is less than 0
- Stable: Grade Sentiment Change equals 0
- Insufficient evidence: either required monthly observation is unavailable or the underlying rating distribution cannot be compared reliably
Only companies with a Positive or Negative analyst-rating direction qualify for the final demonstration set.
The analyst-grade comparison and the July news analysis intentionally use different observation windows because they reflect different data cadences. The grade signal captures the latest available monthly change in analyst consensus entering the July monitoring period, while the news signal captures how the company-specific narrative develops through July. They are therefore treated as complementary pre-earnings evidence rather than simultaneous measurements.
Classifying sentiment and analyst-grade alignment
Once both directional signals are available, each qualifying company is classified as:
- Aligned positive: sentiment is Improving and analyst-rating direction is Positive
- Aligned negative: sentiment is Deteriorating and analyst-rating direction is Negative
- Divergence: sentiment is Improving while analyst-rating direction is Negative, or sentiment is Deteriorating while analyst-rating direction is Positive
- Review required: either signal cannot be classified reliably
Divergence indicates that the recent company narrative and historical analyst-grade sentiment are moving in opposite directions. It does not establish which signal is correct or predict an earnings surprise or subsequent stock-price move.
Assigning pre-earnings review priority
The final classification indicates analyst-review priority, not predicted volatility.
For this demonstration:
- Elevated: the company shows Divergence, or it is Aligned negative with Accelerating news activity
- Moderate: the company is Aligned negative with Not accelerating news activity, or Aligned positive with Accelerating news activity
- Low: the company is Aligned positive and news activity is Not accelerating
- Review required: either underlying signal is unsupported or non-comparable
These labels are author-defined triage rules designed to prioritize research effort before earnings. They are not statistically validated measures of risk, probability, expected return, or volatility.
Assigning evidence confidence
Confidence describes the depth of evidence supporting the rule-based classification. It is not statistical confidence, predictive probability, or model accuracy.
For this demonstration:
- Low confidence: total relevant news volume is five or six, or either half-window contains exactly two relevant articles
- High confidence: at least ten relevant news articles are available and each half-window contains at least four relevant articles
- Medium confidence: all other valid classifications
- Review required: the underlying review-priority classification cannot be supported
Both required Historical Stock Grades observations must be available before any confidence label is assigned.
These confidence cutoffs are also author-defined operational rules rather than statistically calibrated measures.
FMP Data Inputs and Their Analytical Roles
The analysis combines three FMP data inputs to identify upcoming earnings events, evaluate changes in the surrounding news narrative, and measure changes in historical analyst-grade sentiment.
|
FMP data input |
Analytical role |
Required fields |
|
Define the candidate universe of companies scheduled to report within the selected earnings window |
Symbol, company name, earnings date |
|
|
Retrieve company-specific news used to calculate news volume and sentiment change across the July monitoring window |
Symbol, publication date, headline, article snippet where available, source, URL |
|
|
Compare consecutive historical analyst-grade distributions and measure whether analyst sentiment became more positive or negative |
Symbol, date, Strong Buy, Buy, Hold, Sell, Strong Sell counts |
The Earnings Calendar API defines the initial candidate universe. Companies are screened using the predetermined eligibility and ordering rules before their news or analyst-grade alignment is known. FMP's Earnings Calendar supports retrieving earnings announcements across a specified period.
The Search Stock News API supplies the underlying company-specific news records. FMP provides the source data, while Claude performs the relevance filtering, deduplication, and Positive, Neutral, or Negative classification defined in the methodology. Publication date, headline, source, and URL are preserved for every retained article so the sentiment judgments remain auditable.
The Historical Stock Grades API supplies the second evidence layer. FMP describes this dataset as a historical record of analyst grades that can be used to track how Buy, Hold, and Sell sentiment changes over time.
For this demonstration, Claude compares the Historical Stock Grades observations dated June 1 and July 1, 2026. The Strong Buy, Buy, Hold, Sell, and Strong Sell counts are converted into an Analyst Grade Balance for each observation, and the difference between the two balances determines whether analyst-grade sentiment became Positive, Negative, or remained Stable.
The two datasets intentionally operate at different cadences. The analyst-grade comparison captures the latest available monthly shift in analyst consensus entering July, while the news analysis measures changes in company-specific coverage throughout July.
Before the final four-company set is selected, each candidate must have sufficient directional news evidence and two comparable Historical Stock Grades observations with a non-zero Grade Sentiment Change. Candidates are evaluated in the predetermined earnings-date and ticker order, and screening stops when four companies qualify.
This eligibility process ensures that companies enter the demonstration because the required evidence is available, rather than because they produce a preferred aligned or divergent result.
Accessing FMP Data via Claude MCP
This analysis runs by connecting the FMP MCP server to Claude. You need an active FMP API key and the following remote MCP connection URL:
|
https://financialmodelingprep.com/mcp?apikey=YOUR_FMP_API_KEY |
Replace YOUR_FMP_API_KEY with your active key. In Claude, open Settings → Connectors → Add custom connector, enter a recognizable connector name, and paste the URL into the Remote MCP Server URL field. After the connector is added, Claude automatically discovers the available FMP tools.
Before running the complete analysis, validate the three required data inputs with a small test:
- Upcoming earnings-calendar records
- Company-specific historical news
- Historical Stock Grades observations
For the analyst-grade component, confirm that Claude can retrieve the dated June 1 and July 1, 2026 Historical Stock Grades observations, including Strong Buy, Buy, Hold, Sell, and Strong Sell counts. The analysis requires both observations before calculating a Grade Sentiment Change.
The full prompt then screens earnings candidates in a predetermined order, evaluates July news evidence, checks the two historical analyst-grade observations, and stops once four companies satisfy both directional evidence requirements.
FMP remains the source of the underlying earnings, news, and analyst-grade records. Claude performs the relevance filtering, deduplication, sentiment classification, grade-balance calculations, evidence screening, and rule-based classifications defined in the methodology.
For retained news records, the headline, publication date, source, and URL are preserved so the sentiment judgments can be reviewed alongside the resulting classifications.
Running the Pre-Earnings Divergence Analysis in Claude
After validating the FMP MCP connection, the next step is to run a structured prompt that applies the same screening and classification rules across the candidate universe.
For this demonstration, the analysis uses July 31, 2026 as the run date. It screens US-listed operating common stocks scheduled to report between August 7 and August 14, 2026.
The two evidence layers use windows that reflect their underlying data cadence:
- News monitoring window: July 1 through July 30, 2026
- Historical analyst-grade observations: June 1 and July 1, 2026
Claude first retrieves the earnings-calendar universe, applies the predefined security-type exclusions, and sorts eligible candidates by earnings date and ticker. Candidates are then evaluated in that fixed order rather than selected according to whether their eventual result is aligned or divergent.
The news layer is evaluated first. A candidate must have sufficient relevant coverage and an Improving or Deteriorating sentiment trend to proceed. Claude then retrieves the June 1 and July 1 Historical Stock Grades observations and calculates the change in Analyst Grade Balance.
A company qualifies for the final demonstration only when both evidence layers produce a valid directional signal. Screening stops once four companies qualify.
The prompt performs four connected analytical tasks:
- Measures changes in company-specific news sentiment across the two July half-windows
- Measures the change in historical analyst-grade sentiment from June 1 to July 1
- Classifies the two signals as aligned positive, aligned negative, or divergent
- Assigns an author-defined pre-earnings review priority and evidence-confidence label
FMP supplies the underlying earnings-calendar, news, and Historical Stock Grades records. Claude performs the relevance filtering, deduplication, sentiment classification, grade-balance calculations, evidence screening, and rule-based classifications.
For every news record included in the sentiment calculation, the publication date, headline, source, and URL are retained in an audit trail. Mixed, ambiguous, routine, or otherwise non-directional stories are classified as Neutral unless there is a clearly dominant company-specific near-term implication.
The classification thresholds and confidence rules are author-defined operational rules applied consistently throughout the demonstration. They are not statistically validated thresholds, probabilities, confidence intervals, or measures of expected volatility.
The analysis does not attempt to predict an earnings beat or miss, subsequent stock-price direction, realised volatility, or a trading outcome. Its purpose is to identify where the combination of recent news sentiment and historical analyst-grade movement warrants greater analyst attention before earnings.
Claude Prompt
|
Use Financial Modeling Prep data through the connected FMP MCP server to run the pre-earnings review analysis below. This is a compact demonstration, not a market-wide screening exercise. A preliminary eligibility screen was completed using the fixed earnings-date/ticker ordering and the evidence requirements defined above. That screening identified the following four companies for the compact demonstration:
Do not screen additional companies or substitute different tickers. Scope Run date: July 31, 2026 Use FMP MCP only. Retrieve only:
Do not use EPS estimates, price targets, individual upgrade/downgrade events, stock-price performance, historical earnings surprises, transcripts, external web sources, or post-earnings data. The rules below are author-defined operational rules for this demonstration and are not statistically validated. 1. News sentiment For each company, retrieve company-specific news from July 1 through July 30. Remove exact duplicates, syndicated copies, broad market stories without a material company-specific development, and incidental mentions. Preserve for every retained article:
Split the news into:
Classify each retained article:
Do not use stock-price or broad-market movement when assigning sentiment. Calculate for each period: Sentiment Balance = (Positive − Negative) ÷ Total Relevant Articles Then: Sentiment Change = Recent Balance − Earlier Balance Classify:
News Activity is Accelerating only when the recent period has at least two more articles than the earlier period AND at least 1.5× as many articles. Otherwise use Not accelerating. 2. Historical analyst-grade sentiment Retrieve FMP Historical Stock Grades observations dated:
Use:
Calculate for each observation: Analyst Grade Balance = (Strong Buy + Buy − Sell − Strong Sell) ÷ (Strong Buy + Buy + Hold + Sell + Strong Sell) Then: Grade Sentiment Change = July 1 Grade Balance − June 1 Grade Balance Classify Analyst-Rating Direction:
Do not reconstruct or infer missing observations. 3. Alignment Classify:
4. Pre-earnings review priority Apply:
These are analyst-review priority labels, not volatility predictions. 5. Evidence confidence Use:
Confidence represents evidence depth only, not statistical confidence or probability. Required output Keep the response compact. First provide a one-paragraph Screening Note stating that CGC, CYD, UAA, and FERG are the first four qualifying companies from the predetermined screen and confirming whether each still satisfies the required evidence conditions. Do not search for replacement companies. Then produce exactly two tables. TABLE 1 — Pre-Earnings Evidence and Classification Columns:
Round sentiment balances/change to 2 decimals. Round grade balances/change to 3 decimals and calculate changes from unrounded values. The Analyst Follow-Up Action should be one concise company-specific research question based only on the news evidence and historical analyst-grade change. TABLE 2 — News Sentiment Audit Trail Include every retained news item used for the four companies. Columns:
Table 2 must reconcile exactly with Table 1. Use Unavailable for missing raw fields. Never invent evidence. After Table 2, stop. Do not provide additional analysis, predictions, recommendations, or conclusions. |
What the Prompt Produces
The prompt produces two compact tables.
The first table combines the evidence and final classifications for the four demonstration companies. It shows changes in July news sentiment, the June-to-July Historical Stock Grades balance, analyst-rating direction, alignment, pre-earnings review priority, and evidence confidence.
The second table preserves the news evidence used by Claude to calculate sentiment. It includes the publication date, source, headline, assigned sentiment, and URL for every retained article, making the relevance and sentiment judgments easier to audit.
The final run produced valid directional classifications for China Yuchai International and Ferguson Enterprises, while Canopy Growth and Under Armour were marked Review required because their July news evidence did not support a valid directional sentiment signal under the defined rules.
This distinction is intentional. The workflow does not force every company into an aligned or divergent category when the underlying evidence is insufficient or non-directional. Instead, Review required acts as an evidence-control outcome that prevents unsupported classifications.
Interpreting the Pre-Earnings Review Output
The final Claude MCP run evaluated Canopy Growth Corp. (CGC), China Yuchai International (CYD), Under Armour (UAA), and Ferguson Enterprises (FERG). The analysis used July 31, 2026 as the run date, July 1-30 for company-specific news, and the June 1 and July 1 Historical Stock Grades observations for the analyst-grade comparison.
The results show why the evidence requirements matter. Two companies produced valid directional classifications, while two remained Review required because their July news evidence did not support a directional sentiment signal under the author-defined operational rules.
Pre-Earnings Evidence and Classification
|
Company |
Ticker |
Earnings date |
Earlier news |
Recent news |
30-day volume |
News activity |
Earlier sentiment |
Recent sentiment |
Sentiment change |
Sentiment trend |
June 1 grade balance |
July 1 grade balance |
Grade change |
Analyst-rating direction |
Alignment |
Review priority |
Evidence confidence |
|
Canopy Growth Corp. |
CGC |
Aug. 7, 2026 |
6 |
7 |
13 |
Not accelerating |
-0.17 |
0.00 |
+0.17 |
Stable |
0.200 |
0.111 |
-0.089 |
Negative |
Review required |
Review required |
High |
|
China Yuchai International |
CYD |
Aug. 7, 2026 |
4 |
3 |
7 |
Not accelerating |
0.00 |
0.33 |
+0.33 |
Improving |
0.750 |
0.800 |
+0.050 |
Positive |
Aligned positive |
Low |
Medium |
|
Under Armour, Inc. |
UAA |
Aug. 7, 2026 |
0 |
2 |
2 |
Accelerating |
Unavailable |
0.00 |
Unavailable |
Insufficient evidence |
0.038 |
0.074 |
+0.036 |
Positive |
Review required |
Review required |
Low |
|
Ferguson Enterprises Inc. |
FERG |
Aug. 10, 2026 |
3 |
9 |
12 |
Accelerating |
0.67 |
0.00 |
-0.67 |
Deteriorating |
0.696 |
0.667 |
-0.029 |
Negative |
Aligned negative |
Elevated |
Medium |
China Yuchai produced the clearest aligned-positive result. Its news sentiment improved from 0.00 to 0.33, while its analyst-grade balance increased from 0.750 to 0.800. Because news activity was not accelerating, the author-defined rules assigned a Low review priority.
Ferguson showed the opposite pattern. News sentiment deteriorated from 0.67 to 0.00 while its analyst-grade balance also declined, producing an Aligned negative classification. Recent news activity was accelerating, resulting in an Elevated review priority. The appropriate follow-up is to investigate whether developments such as the FloWorks acquisition or other company-specific changes help explain the weakening evidence entering the announcement.
Canopy Growth demonstrates why a change in analyst-grade sentiment does not automatically create divergence. Its analyst-grade balance declined from 0.200 to 0.111, but its July Sentiment Change was only +0.17, below the +0.20 Improving threshold. The news signal therefore remained Stable, so an alignment classification was withheld.
Under Armour was constrained by evidence coverage. Only two qualifying company-specific articles were retained, both in the recent half-window. That fails the minimum news-evidence requirement, so its positive June-to-July analyst-grade movement cannot be combined with a reliable July sentiment trend.
News Sentiment Audit Trail
Claude retained 34 company-specific news records across the four companies. The complete output preserves the publication date, source, headline, URL, and assigned sentiment for every article used in the calculations.
The table below shows a small sample of that audit trail for readability:
|
Company |
Ticker |
Period |
Publication date |
Source |
Headline |
Sentiment |
|
Canopy Growth Corp. |
CGC |
Earlier |
Jul. 7, 2026 |
The Motley Fool |
Canopy Growth: Is Another Reverse Stock Split Inevitable? |
Negative |
|
Canopy Growth Corp. |
CGC |
Recent |
Jul. 21, 2026 |
Business Wire |
Canopy Growth Announces Participation at Upcoming Canaccord Genuity Growth Conference |
Neutral |
|
China Yuchai International |
CYD |
Recent |
Jul. 30, 2026 |
Zacks Investment Research |
China Yuchai (CYD) Upgraded to Strong Buy: Here's What You Should Know |
Positive |
|
Under Armour, Inc. |
UAA |
Recent |
Jul. 17, 2026 |
PRNewswire |
Under Armour Announces Date for First Quarter Fiscal 2027 Earnings Conference Call |
Neutral |
|
Ferguson Enterprises Inc. |
FERG |
Earlier |
Jul. 13, 2026 |
Business Wire |
Ferguson to Acquire FloWorks for $1.6 Billion, Increasing its Total Addressable Market to $400 Billion and Expanding its Non-Residential Value-Added Capabilities |
Positive |
The full Claude output contains all 34 retained records and their URLs. Preserving this evidence is important because FMP supplies the underlying news records, while relevance filtering and Positive, Neutral, or Negative classification are judgments applied by Claude. Analysts can therefore inspect the source evidence before accepting the resulting sentiment calculation.
The final result demonstrates both outcomes the process is designed to produce. When sufficient directional evidence exists, it can classify the signals as aligned and assign a review priority. When evidence is sparse or non-directional, Review required prevents an unsupported alignment or divergence conclusion.
None of these classifications predicts whether a company will beat or miss earnings, how its share price will react, or how volatile the post-announcement move will be. They indicate only where the available pre-earnings evidence warrants more or less analyst attention.
Where Pre-Earnings Signals Need Analyst Review
The analysis can organize news and historical analyst-grade evidence consistently, but the resulting classifications still require analyst judgment. The most important review cases arise when news coverage is sparse, sentiment is non-directional, analyst-grade changes are small, or the two evidence layers appear to move differently for reasons that require company-specific interpretation.
Insufficient or ambiguous news evidence
A directional sentiment trend should not be assigned when the available coverage is too limited. Under Armour illustrates this constraint. Only two qualifying articles were retained, both in the recent half-window, so the minimum evidence requirement was not met even though its analyst-grade balance increased from June to July.
Mixed or ambiguous stories should also remain Neutral unless there is a clearly dominant near-term implication. This prevents routine corporate announcements, institutional holding updates, or broadly balanced commentary from being forced into Positive or Negative classifications.
Claude performs these relevance and sentiment judgments rather than FMP. The retained headline, publication date, source, and URL therefore remain important audit evidence.
Stable signals should not be forced into divergence
A measurable change in one evidence layer does not automatically create a valid alignment or divergence classification.
Canopy Growth's analyst-grade balance declined from 0.200 to 0.111, but its July Sentiment Change was only +0.17. Because that remained below the author-defined +0.20 Improving threshold, the news trend was classified as Stable and the overall result remained Review required.
This is intentional. The evidence rules should prevent the analysis from manufacturing a directional conclusion when one side of the comparison remains non-directional.
Historical Stock Grades reflect a monthly consensus snapshot
Historical Stock Grades operate at a different cadence from the July news analysis. In this demonstration, the analyst-grade signal compares the June 1 and July 1 distributions of Strong Buy, Buy, Hold, Sell, and Strong Sell ratings.
The resulting Grade Sentiment Change shows whether the aggregate analyst-grade balance became more positive or negative between those two monthly observations. It does not identify the specific analyst action or company event responsible for the change.
Analysts should therefore investigate the underlying business context before interpreting a rising or falling grade balance. The grade signal is a structured direction-of-consensus indicator, not an explanation of why consensus changed.
Alignment still requires company-specific interpretation
Even a valid aligned classification should not be treated as an investment conclusion.
China Yuchai showed Improving news sentiment alongside a higher analyst-grade balance, producing an Aligned positive result. Ferguson showed Deteriorating news sentiment alongside a lower grade balance and Accelerating news activity, producing an Aligned negative result with Elevated review priority.
In both cases, the next step is to identify the underlying business driver. Analysts should determine whether the news and grade movement reflect demand, margins, acquisitions, regulation, litigation, management developments, or another company-specific factor relevant to the upcoming reporting period.
The review-priority label indicates where deeper investigation is warranted. It does not establish whether the consensus view is correct or predict how the company will report.
Revalidate the evidence before the earnings event
Earnings dates, news coverage, and analyst opinion can continue to change after the initial run. Research teams should therefore confirm that the earnings date remains valid and review any material developments that appear after the monitoring window before relying on the output for coverage planning.
The value of the process is not that every company receives a directional classification. Its value is that the same evidence rules are applied consistently: valid signals can be classified as aligned or divergent, while sparse or non-directional evidence remains Review required.
Used this way, the analysis helps research teams allocate attention before earnings without presenting news sentiment, historical analyst grades, or review-priority labels as predictions of earnings outcomes, stock-price direction, realised volatility, or trading performance.
