Seven Ways Adaptation Can Go Wrong

An anti-pattern is a recurring approach that creates a problem instead of solving it. You do not need the terminology to use this page: look for the familiar failure, then consider the alternative.

Use this when reviewing a website or signed-in application. The suggested patterns are options to investigate, not automatic fixes.

1. An unexplained personal assumption

Reference name: Creepy Over-Personalization.

Description: Using too much inferred data too early or too aggressively, making users feel watched. Classic example: "We noticed you looked at shoes yesterday…" shown on what the user experiences as a first visit.

Why It's Harmful: The person may feel watched or misunderstood. The site has not explained what information it used or given them a useful way to correct it.

Consider instead: Privacy-First Context Layer Pattern + Cold-Start Graceful Pattern.

2. A visible change with no useful purpose

Reference name: Static Template Masquerading as Aware.

Description: Adding a few dynamic widgets or recommendation blocks on top of an otherwise completely rigid, fixed page template — and calling the result "personalized."

Why It's Harmful: A cosmetic change can be marketed as meaningful help without improving the visitor’s task. The failure is the unsupported promise—not the existence of a static template or a useful fallback. A small explicit-choice rule can be a good first implementation.

Consider instead: Dynamic Composition Pattern.

3. Recognizing a device without explaining it

Reference name: Silent Fingerprinting.

Description: A browser fingerprint combines characteristics of a device or browser to try to recognize it. Collecting that information without an understandable explanation and appropriate controls can create tracking the person did not expect.

Why It's Harmful: Unexplained fingerprinting can undermine trust and create privacy and legal risk. Applicable duties depend on the actual use and jurisdiction; obtain an appropriate review rather than assuming a design label establishes compliance.

Consider instead: Privacy-First Context Layer Pattern.

4. An old preference that will not let go

Reference name: Stale Memory Syndrome.

Description: Retaining and acting on outdated user data long after the context has changed — showing last year's promotions, surfacing interests the user has clearly moved past, or "remembering" preferences that are no longer relevant.

Why It's Harmful: Feels outdated and annoying, reduces perceived relevance, and undermines the trust that good memory is supposed to build. Stale memory is worse than no memory.

Consider instead: Context Retention Pattern + Cross-Session Memory Pattern, both implemented with automatic data expiry and easy user controls.

5. Help that interrupts the task

Reference name: Interruptive Proactive Nudges.

Description: A popup, suggestion or forced sequence blocks what the person was doing. For example, a sales invitation interrupts someone trying to find a support number. A short, dismissible introduction can help on arrival; repeated or inescapable prompts are a different experience.

Why It's Harmful: An interruption can increase frustration and obstruct the original task. Time or pointer behavior does not prove hesitation; offer help without forcing a detour.

Consider instead: Proactive Flow Adaptation Pattern + Intent Inference Pattern, implemented as optional assistance with a clear reason and always-visible dismiss controls.

6. A guess that keeps reinforcing itself

Reference name: Bias Amplification Loop.

Description: The site guesses an interest, shows mostly that topic, then treats clicks on the limited choices as proof that its guess was right. The person has fewer chances to discover alternatives or correct the assumption.

Why It's Harmful: Leads to unfair or exclusionary experiences, damages brand reputation, and can cause regulatory exposure. This is the anti-pattern where good intentions (relevance) produce actively harmful outcomes.

Consider instead: Ethical Guardrail Pattern + Real-Time Feedback Loop Pattern.

7. A category the person cannot escape

Reference name: One-Size-Fits-Most Personalization.

Description: Assigning people to broad inferred buckets and making those assumptions difficult to inspect or change. This is different from letting someone choose a useful route for their task.

Why It's Harmful: A broad category can hide relevant options or become a false claim about the person. Explicitly chosen routes are not inherently bad; keep them understandable, reversible and complete, and test whether they help.

Consider instead: Dynamic Composition Pattern + Intent Inference Pattern.

Closing: Recognize, Then Fix

Start with the effect on the person: can they understand what happened, continue their task and choose another route? A complicated architecture does not excuse a confusing experience.

We recommend auditing your projects against both this gallery and the Patterns Library. The patterns and anti-patterns are designed to be read together — each anti-pattern points directly to the pattern that corrects it.

Want to apply this to your website?

Compare familiar problems and choose a useful first step. This guide does not scan or score your site.