Skip to content

Added analyst signals indicators

Ticket #273: Added Analyst's signals for my portfolio decisions
Type: Feature / Decision Support / Business Analysis
Affected Component: code_source_simule/portfolio_metrics.py, code_source_simule/flask_app.py, templates/dashboard.html, templates/titre_detail.html, tests/test_portfolio_metrics.py, tests/test_dashboard.py, tests/test_titre_detail.py, specs/016-dashboard-analyst-signals/

1. Context

This report is a direct continuation of my previous work on adding portfolio management-rule indicators. Even though the indicators here are new, they serve exactly the same vision: making my dashboard a true decision-support tool, and preparing a clear information foundation for the future advisory AI (see my Project Vision).

The logic remains the same as in the previous ticket: translate my investment strategy into simple, readable, and reliable signals displayed in the right place. This time, I focused on what analysts anticipate for each stock.

2. Objective

Make it immediately visible which stocks deviate strongly from analyst expectations, and provide on each stock detail page a visual reading of where the current price sits within the target range.

Concretely, the goal was to:

  • highlight stocks clearly above and clearly below the average analyst target;
  • isolate counter-intuitive cases, meaning stocks that are rising sharply despite negative market sentiment;
  • add an "Analyst Targets" gauge on the stock detail page, where I review a position in depth.

3. What was delivered

  • Three new indicators on the dashboard:
    • Near analyst high target: stocks whose price reaches or exceeds the highest analyst price target, sorted from the most extreme to the least extreme case.
    • Near analyst low target: stocks whose price is at or below the lowest analyst price target, with the same priority sort.
    • Contrary to expectations: stocks with negative or very negative sentiment that still show a gain increase above 50%.
  • One "Analyst Targets" gauge on each stock detail page: low target on the left, high target on the right, and the current price positioned between them for immediate reading.
  • A readable and disciplined presentation:
    • a clear display format for each line (price compared with actual target bounds);
    • an explicit empty state when no stock matches, to avoid ambiguity;
    • an "unavailable" or "invalid data" state on the gauge when targets cannot be used.
  • A deliberately autonomous computation scope: signals rely strictly on current CSV data, with target ranges defined directly by the individual analyst-published lowest and highest price targets.

4. Business impact

  • More focused decisions: I can spot at a glance the stocks that deviate most from market expectations, in either direction.
  • Right-level reading: the gauge on the stock detail page places valuation information exactly where I analyze a position.
  • Watch signal: the "Contrary to expectations" indicator draws attention to divergences that deserve investigation.
  • Vision continuity: these indicators extend the previous ticket and enrich the context that the future advisory AI will be able to use.

5. Validation and status

  • Full test suite green: 43 passed.
  • Key behaviors covered by dedicated tests: direct price-to-target comparison for high target (price >= highest_price_target), direct price-to-target comparison for low target (price <= lowest_price_target), strict rule of negative sentiment combined with gain increase above 50%, sorting from most extreme to least extreme, and gauge states (ready, unavailable, invalid range).
  • Coverage recalculated after the changes: portfolio_metrics.py at 96%, flask_app.py at 94`.
  • Related traceability ticket: #273.

6. Lessons learned

  1. A signal has value only if it remains reliable when data is incomplete; explicitly handling empty and invalid cases protects trust in the tool.
  2. Token usage summary: for this addition of analyst-signal indicators, the recorded gross token cost was $9.43 for 935.8 Copilot AI credits. Compared with a recent similar implementation, #272, this implementation therefore cost about 42.95% less, which is a remarkable improvement.

PS: Token usage update: after commit, following a technical issue (resolved by Github Copilot) during deployment, the token cost increased to $11.18 for 1099.1 credits used, which is 32.12% less than the previous implementation. This defect therefore cost me less than $2 to resolve, but saved me at least an hour of investigation and troubleshooting. Considering that an independent AI consultant's hourly rate can exceed $100, this additional token cost is ultimately a significant benefit.