Time-Decay Weighting

The most common mathematical model for time-decay weighting is the exponential decay function. It applies a diminishing multiplier to a data point's original value based on its age.

• Formula: Weight = e^(-λ * t), where λ (lambda) is the decay constant and t is the age of the data point.
• Half-Life Concept: The half-life parameter defines the time it takes for a signal to lose 50% of its weight. A shorter half-life aggressively prioritizes recency.
• Smooth Degradation: Unlike step functions that abruptly cut off old data, exponential decay provides a smooth, continuous degradation curve that prevents ranking cliffs.

While related, recency boosting and decay weighting are distinct mechanisms in temporal scoring.

• Recency Boosting: A temporary, additive bonus applied to newly published or significantly updated documents to test their relevance against established content. It is a promotional signal.
• Decay Weighting: A continuous, multiplicative penalty that reduces the authority of aging signals. It is a suppressive mechanism.
• Interaction: A search engine might apply a recency boost for the first 48 hours, then transition the document to a standard decay curve where its ranking authority diminishes gradually unless updated.

The decay constant (λ) is not universal; it is dynamically adjusted based on the temporal intent of the user's query.

• QDF Queries: For queries exhibiting Query Deserves Freshness (QDF) signals—such as breaking news or trending topics—the decay rate is extremely aggressive, with half-lives measured in hours.
• Semi-Evergreen Queries: For topics requiring periodic updates (e.g., annual tax brackets), the decay function may use a stepped or slow linear degradation with a half-life of 12 months.
• Evergreen Queries: For timeless knowledge (e.g., fundamental physics concepts), the decay constant approaches zero, effectively disabling temporal weighting.

Modern freshness scoring does not apply a single decay function to the entire document. Instead, it performs granular, signal-level decay weighting.

• Content Body Decay: The publication or last-modified date of the core text is weighted against the query's temporal profile.
• Reference Decay: Citations and outbound links are individually scored. A document linking to a 404 error or an outdated source suffers a staleness penalty on that specific signal.
• Engagement Signal Atrophy: User interaction metrics like click-through rate (CTR) are weighted with their own decay function. A high CTR from three years ago carries negligible weight compared to recent engagement data.

In programmatic content infrastructure, time-decay weighting is operationalized through automated refresh triggers and delta detection engines.

• Threshold-Based Reindexing: A monitoring system continuously calculates the decayed score of a document. When the score drops below a predefined threshold, it triggers an automated content regeneration pipeline.
• Delta Detection: Before regenerating, a content diff algorithm compares the live page against a cached baseline. If only a single statistic has changed, the system updates only that component, preserving the historical authority of unchanged sections.
• Decay Velocity Monitoring: The system tracks the decay velocity—the speed of traffic or ranking decline—to predict future staleness and schedule preemptive updates before a critical ranking loss occurs.

Choosing between exponential, linear, or logarithmic decay models depends on the content's lifecycle stage and business objectives.

• Exponential Decay: Best for news and ephemeral content where value drops precipitously after the event. Provides aggressive suppression.
• Linear Decay: Suitable for semi-evergreen content like annual reports, where value diminishes at a constant, predictable rate until the next update cycle.
• Logarithmic Decay: Used for content that retains long-term residual value but experiences a rapid initial drop. Common for viral social posts that settle into a stable, low-level background authority.
• Gaussian (Bell Curve) Decay: Applied to seasonal relevance windows, where content weight increases toward a peak date (e.g., Black Friday) and symmetrically decays afterward.

A temporary algorithmic promotion applied to newly published or significantly updated pages. This boost allows fresh content to compete against established, high-authority pages for a probationary period. The boost magnitude and duration are typically query-dependent: breaking news queries may receive a 6-hour boost, while semi-evergreen topics might see a 14-day window. If engagement signals validate the content, the boost transitions into sustained ranking; if not, the page reverts to its organic position.

The measured speed at which specific content types lose organic traffic, backlinks, or engagement signals. Decay velocity is not uniform:

• News articles: High velocity, often losing 80% of traffic within 72 hours
• Technical documentation: Moderate velocity, decaying over 12-18 months as APIs evolve
• Evergreen reference: Near-zero velocity, maintaining relevance for years Monitoring decay velocity enables automated refresh triggers before traffic loss becomes critical.

An NLP model that analyzes a search query to determine the user's temporal needs. The classifier assigns queries to categories:

• QDF (Query Deserves Freshness): User wants the latest information (e.g., 'stock price today')
• Historical: User seeks a specific past snapshot (e.g., '2019 tax brackets')
• Evergreen: Time is not a factor (e.g., 'how to tie a tie') This classification determines whether time-decay weighting is applied at all, and with what aggressiveness.

A system that compares the current live version of a document against a cached baseline to identify only the modified sections. Rather than treating an entire page as 'fresh' after a minor edit, delta detection quantifies the semantic significance of changes:

• A typo fix yields negligible freshness gain
• Replacing outdated statistics with current data triggers a full recency reset
• Adding a new section on emerging trends generates partial freshness credit This prevents gaming through trivial updates while rewarding substantive revisions.

A graphical representation of how a page's click-through rate from search results diminishes over time. As content ages, the title and meta description may become less compelling relative to fresher competitors with more current language and dates. The curve typically shows:

• Initial peak: High CTR immediately after publication or update
• Plateau phase: Stable CTR during the content's prime relevance window
• Decline slope: Gradual CTR erosion as competitors publish newer content
• Obsolescence tail: Persistent but minimal CTR from long-tail, low-competition queries