🔗 Habit Formation · 11 min read · Subtopic 3 of 5

When if-then plans fail

A plan is a tool, and like every tool it has a working envelope. The same meta-analysis that found an overall effect of d = 0.65 also found that some plans do nothing at all — and a few quietly backfire. This page maps the boundary conditions the literature actually documented: weak commitment, avoidance wording, cues that cannot fire, and lives that reorganize. Knowing where the tool breaks is how you keep it from breaking on you.

🔎 Evidence Snapshot ★★★★☆ Good — the boundary conditions come from the same well-replicated meta-analytic literature; the honest limit is that most studies ran over short time frames

What the evidence supports

  • Implementation intentions amplify an existing goal; when commitment to the goal is weak, the plan adds little (Gollwitzer & Sheeran, 2006).
  • Plans framed as adding a behavior outperform plans framed as suppressing one — eating-healthy meta-analysis found d = 0.51 for adding foods versus d = 0.29 for cutting unhealthy snacks (Adriaanse et al., 2011).
  • Vague or rare cues undercut the effect — the cue has to be concrete and actually encountered for the automatic link to fire.
  • When a competing behavior is already strongly habitual in the same context, the old habit tends to win the cue (Wood & Neal, 2007).

What remains uncertain

  • Most studies measured follow-through over weeks; how boundary conditions play out over months of real life is less precisely mapped.
  • Whether an avoidance plan can be rescued by pairing it with a substitute behavior is supported by clinical intuition more than by a large dedicated evidence base.
  • Individual differences — personality, stress load, life chaos — moderate the effect, but the moderation pattern is not yet precise enough to predict who will fail in advance.

Evidence last reviewed: August 15, 2026. Conclusions may change as new research is published.

where it breaks

The Honest Starting Point

An average hides a wide spread: in the 94 studies pooled by Gollwitzer and Sheeran, most showed a positive effect but the sizes varied enormously — exactly what you would expect if the method works under some conditions and not others. So the useful question is not "do if-then plans work?" but "which plans fail, and why?" The literature answers with a short list of failure modes, and nearly all of them are visible in the wording of the plan itself.

When the Goal Is Not Really Yours

Implementation intentions are a bridge, and a bridge needs two shores. The first shore is a goal you actually hold; the plan only carries you across the gap between intending and doing. If the goal is borrowed — adopted because you feel you should — the bridge has nothing to carry. The meta-analysis found larger effects when goal commitment was high, and several studies showed no benefit for people low on commitment. The warning is blunt: do not write a plan for a goal you do not own.

The Avoidance Trap

The clearest measurable failure mode is the wording of the "then" half. "If I crave chips, then I will not eat chips" works less well than a plan written to add a behavior. The eating-behavior meta-analysis found a meaningful difference: adding healthy foods showed a medium effect (d = 0.51), while cutting unhealthy foods showed a smaller one (d = 0.29) (Adriaanse et al., 2011). Part of the explanation is that "not doing" leaves a vacuum: the cue fires, the behavior is blocked, and nothing else has been scheduled to happen. The craving sits there waiting for the next weak hour. The fix is to write a substitute action into the "then" — something specific to do instead.

Approach Beats Avoidance in the Meta-Analysis
Effect sizes on eating behavior from the systematic review and meta-analysis by Adriaanse et al. (2011): plans written to add healthy foods (d = 0.51) outperformed plans written to cut unhealthy ones (d = 0.29). The overall figure across all goals (d = 0.65, Gollwitzer & Sheeran, 2006) is included as reference.
Any goal (2006 overall) d = 0.65 Adding healthy foods (approach) d = 0.51 Cutting unhealthy snacks (avoidance) d = 0.29 0 0.2 0.4 0.6

The broader literature on thought and craving suppression adds a caution: actively trying not to think about a tempting stimulus can increase how often it intrudes afterward. That is not a reason to abandon restraint — it is a reason to build restraint as an action you perform ("I will go for a walk") rather than an absence you attempt ("I will not eat the snack").

Cues That Cannot Fire

The "if" half of the plan fails when the cue is not concrete, not encountered, or not noticed. The meta-analytic process findings showed the plan works by making the cue easy to spot; a cue that was never specific in the first place has nothing to become easy to spot. The fix is to reword the trigger until it names an event you will actually meet.

⚠️ Cue that failsWhy it failsReword it as
⏳ "When I have time" Not an event — an abstract condition that never announces itself "At 7:00 PM, after I finish dinner"
😣 "When I feel stressed" An internal feeling that is hard to notice until it has already passed "The moment I close my laptop for the day"
🎲 "When I get a chance" Depends on luck you cannot schedule around "After I park the car"
💭 "When I think of it" Assumes your memory will volunteer the cue — the assumption the plan was meant to remove "After I brush my teeth"
📅 "Next week sometime" Too rare and too distant to build the cue-response link through repetition A daily anchor from the stacking page

🚧 Stress does not break good plans — it breaks vague ones

One of the more useful findings in the literature is that if-then plans often help most under cognitive load: when your attention is stretched, the automatic cue-response link carries the behavior that a deliberative decision could not. But that same load is lethal to a vague plan, because nothing automatic has been built to fire. The stress test is the final proof of the wording — if the plan works on the day your brain is full, the wording was specific enough. If it collapses the first week you are busy, rewrite the cue and the action, and the review-systems protocol page has the weekly check that catches this before it becomes a month of quiet failure.

When Life Reorganizes

Implementation intentions are context-bound: the cue lives in a specific place and time, and when the context disappears, the plan dies with it. Travel, illness, a new job, a moved kitchen — any reorganization of the day removes the anchor the plan was wired to. The literature rarely follows plans through such disruptions, but the habit research is clear that context change is one of the few reliable ways to break even a strong automatic behavior. The defense is to treat the plan as a renewable contract rather than a one-time decision:

When a Stronger Habit Owns the Cue

The most quietly consequential boundary condition is the competing habit. Habits are triggered by context, and a context can be wired to more than one response. If you have spent years reaching for your phone when you sit on the couch, a new plan that says "if I sit on the couch, then I will read" is asking the couch to trigger a young behavior against a veteran one. The research on automatic behavior suggests the older, more frequently repeated response will usually win the cue. The honest workaround is not to out-will the old habit in its own territory but to move the new behavior to a context the old habit does not own — a different chair, a different time, a different room — because context is the one variable that shifts the odds for the underdog.

0.29
effect size for avoidance-framed plans — the weakest measured form of the tool
d = 0.51
effect size for approach-framed plans — write the substitute action, not the omission
0.65
the overall average — real, but an average that hides the failures above

Reading the Effect Honestly

Taken together, a pattern emerges: the plan fails when it is disconnected from reality — an unowned goal, a suppressed behavior with no replacement action, a cue that does not exist, a context that changed, a competitor that outmatches it. Every failure mode is fixable at the writing stage except one: if the goal is not yours, no wording fixes it. For everything else the fix is a rewording rule — own the goal, approach the behavior, name a real cue, schedule the re-entry, and give the new behavior a context the old habit does not occupy. The worksheet page builds all five rules into one page of paper.

The Bottom Line

  1. Plans amplify goals they cannot create — a plan for a borrowed goal is the most common silent failure, and no wording can save it.
  2. Approach beats avoidance — write the action you will take instead of the thing you will not do; the meta-analysis shows the difference directly (d = 0.51 versus d = 0.29).
  3. The cue has to actually fire — concrete, daily, and observable beats vague, occasional, and internal; stress proves which one you wrote.
  4. Context changes kill plans, and stronger habits win cues — plan the travel variant, the re-entry, and the fresh context before the plan needs them.

Related Topics

Sources & further reading