Habits & Starting · Mechanism

If-Then Plans: The Famous Number Just Got Cut in Half

The if-then plan works. But a 2025 meta-analysis of 642 tests puts it at d = .36, not the famous .65. Here's how to write one that still earns its keep.

by Alex July 29, 2026 19 min read

Save for later:  Instapaper · Readwise · Raindrop

You’ve probably met this number: d = .65. It came from a 2006 meta-analysis of 94 tests, and it turned the humble if-then plan into the most respected trick in habit science (Gollwitzer & Sheeran, NCI chapter).

In 2025, two of the same authors ran the numbers again. This time across 642 tests. The effect came in at d = .36. Studies finding nothing rarely get published, and a Bayesian model that corrects for it puts the effect at d = .15 (Sheeran, Listrom & Gollwitzer, 2025).

So the technique is real. It’s just a lot smaller than the internet told you. That changes how you should write one, and it changes what you should expect from it.

The Bottom Line

• The if-then plan survived a 642-test re-audit by the same authors, but at d = .36, and d = .15 under a Bayesian model that corrects for missing null results (Sheeran et al., 2025).
• The conditional wording is the active ingredient: “if X, then Y” beat plans written as a schedule, d = .43 versus d = .29.
• Name a time and a place. Time alone is weaker. Adding how and how long drops it back to where time-alone was.
• It converts motivation into action. It does not create motivation: people with weak intentions to exercise got roughly nothing.
• One plan beats three: three plans landed at d = .07, though the authors found no optimal number.

In this guide:

• What an if-then plan actually is
• Why the famous number shrank
• Why the wording matters
• Where it works and where it flops
• Do you still need motivation?
• How to write one that works
• Does the effect last?
• Is WOOP the same thing?
• FAQ

What is an if-then plan, exactly?

An if-then plan is a decision you make in advance about when, where and how you’ll act. Psychologist Peter Gollwitzer introduced the idea in 1993 and laid it out for a general audience in a 1999 American Psychologist review, drawing a line between two things people confuse (Gollwitzer, 1999).

A goal intention sounds like this: “I intend to reach X.” A run three times a week. A calmer evening. An emptier inbox.

An implementation intention sits underneath the goal and fills in the blanks: “When situation x arises, I will perform response y.” After I pour my coffee, I write one sentence. When my phone hits the nightstand at 10pm, I plug it in across the room.

Why bother? Because wanting things is not the bottleneck. In that same 1999 paper, Gollwitzer put it bluntly: intentions account for only 20% to 30% of the variance in behavior. In plain English, knowing what someone plans to do tells you surprisingly little about what they’ll actually do.

The companion chapter puts a face on that number. Across health-behaviour studies, people turned their good intentions into action about 53% of the time (Sheeran, 2002, reported in Gollwitzer & Sheeran). Roughly a coin flip.

The plan is supposed to close that gap. Gollwitzer’s argument was that pre-deciding hands the trigger to the situation instead of to you, so the action starts fast and cheap, and by his account it “may even occur without a conscious intent.” You’re not summoning willpower in the moment. You already spent it, once, in advance.

Did the effect really get cut in half?

Yes, and then some. The 2006 headline was d = .65 across 94 independent tests and more than 8,000 people, with the authors splitting it into d = .61 for getting started and d = .77 for not getting derailed (Gollwitzer & Sheeran). The 2025 update, from Sheeran and Gollwitzer themselves, pooled 642 tests from 294 reports and reported d = .36, with a confidence interval of .33 to .40 (Sheeran et al., 2025).

Quick translation, because effect sizes are jargon. Researchers score effects with a number called d. Roughly: .2 is small, .5 is medium, .8 is large. So the story went from “medium to large” to “small.”

Then it got smaller, depending on how you correct it. The 2025 authors tested whether studies finding nothing had quietly gone unpublished, and found that they had. Their statistical check for that hit the threshold they call “substantial” publication bias.

The authors ran two corrections. Trim and fill barely moved the number, to d = .35, with a range of .31 to .38. A Robust Bayesian Meta-Analysis, which models the missing studies directly, put it at d = .15, with a range of .08 to .22. That second analysis gave extreme evidence on two things at once: the effect is real, and the bias is real.

The authors’ own summary is the fair one: if-then plans have “a robust impact on outcomes though the effect size is small according to conventional guidelines and highly heterogenous.”

Here’s the part the productivity internet skipped. The most-quoted headline for this technique is that it doubles or triples your odds of success. The study usually cited for it is one 248-person trial over two weeks, with people reporting their own exercise (Milne, Orbell & Sheeran, 2002, British Journal of Health Psychology). A d is a standardised difference between two group averages. It is not a multiplier on your odds. Anyone converting one into the other is selling, not reporting.

Full disclosure: d = .65 appears in several older posts on this site, including the piece on systems versus motivation. It was the best available number when those were written. This article is the correction, and those posts are next in line for it.

Why does the if-then wording beat a to-do list?

Because the conditional grammar is doing real work. In the 2025 meta-analysis, plans written as “if X happens, then I will Y” produced d = .43 across 373 tests. Plans written instead as a schedule, “I will do Y at time T in place P,” produced d = .29 across 269 tests. The gap was statistically reliable.

That’s the difference between hanging the action on a trigger and hanging it on a clock. A clock can be ignored. A trigger you meet anyway, coffee, doorway, nightstand, doesn’t need remembering.

The cleanest real-world test of the prompt on its own points the same way. In a 2011 paper, researchers reported a randomised trial run in autumn 2009 at a large Midwestern utility firm, 3,272 employees, using administrative vaccination records rather than self-report (Milkman et al., 2011). Everyone got a mailing about free workplace flu clinics. Some mailings also asked people to write something down.

Read those two rows carefully. Writing down a date moved nothing measurable. Adding a time moved vaccination by about four points, and it held up in the adjusted analysis at +4.2 points.

Four points, from one line of writing, at essentially zero cost. That’s a genuinely good deal for an employer. It is also nothing like a transformation, and no honest reading of it produces “doubles your chances.”

Where do if-then plans actually work?

Not everywhere, and the pattern is worth knowing before you pick your target. In the 2025 data, behaviour change was the weakest of the four outcome families at d = .27, behind cognitive performance (.47), social judgement (.52) and affect regulation (.66). If-then plans do more for your attention and your mood than for your actions.

Zoom in and the spread gets wider still.

All rows from Sheeran et al. (2025), Table 1.

An independent 2023 meta-analysis on drinking found the same shape. Across 16 studies, if-then plans nudged weekly alcohol consumption down slightly, d = -0.14, but did nothing at all to heavy drinking sessions, d = -0.01 (Cooke, McEwan & Norman, 2023). Notably, that review found no sign of publication bias, so the small numbers are probably the honest ones.

So use the tool where it’s strong. Remembering, starting, redirecting attention, catching a feeling before it drives. For heavy behaviour change, a plan is a start and not a strategy: pair it with a habit sized for your worst day.

Do you still need motivation?

Yes, and this is the most misused part of the whole literature. An if-then plan converts a want into an act. It does not manufacture the want.

In the 2025 analysis motivation was a strong moderator: across the 8 tests that allowed the comparison, the high-versus-low motivation contrast came in at d = .79. The exercise data makes it concrete. Among people with strong intentions to be active, planning delivered d = .31. Among people with weak intentions, the effect sat at essentially zero, with a confidence interval running from -.09 to .09.

The authors’ read on why physical activity looks so poor overall: the modest effect “appears to accrue from recruitment of participants who had little intention to change.” Studies enrolled people who didn’t really want to exercise, then asked them to plan exercising. Nothing happened. Of course nothing happened.

There’s a second gate nobody mentions: whether you’d have planned anyway. In 2008, ahead of the Pennsylvania Democratic presidential primary, researchers ran a get-out-the-vote experiment with 287,228 registered Democrats, published in 2010 (Nickerson & Rogers, 2010). Callers asked three plan questions: what time you’ll vote, where you’ll be coming from, what you’ll be doing beforehand.

Among people reached who lived alone, turnout rose 9.1 points. Among people reached who lived with other eligible voters, it moved -1.5 points. Nothing. A standard get-out-the-vote call did nothing for anyone.

The authors’ explanation is the useful lesson: people who live with others already make voting plans out loud, at dinner, in the hallway. The call had nothing left to add. Planning helps precisely where you would not otherwise have planned. Where a plan already exists, formalising it is busywork.

How do you write an if-then plan that works?

Four rules, all from the 2025 moderator tables, and each one contradicts advice you’ve read somewhere.

One caveat first, and it matters. These are comparisons between studies inside a meta-analysis, not head-to-head experiments. Read them as the pattern across the literature, not as proof. The authors themselves call for direct trials on some of these.

Rule 1: name a time and a place, then stop

Specificity has a sweet spot and you can overshoot it. Plans that fixed a time only came in at d = .25. Plans that fixed a time and a place hit d = .46. Plans that also spelled out how and for how long fell back to d = .24.

The authors read it this way: deciding what to do belongs before the planning, not inside it. Work out the action first. Then the plan is only a trigger and a location, and there’s nothing left to negotiate.

Rule 2: write one plan, not five

One plan produced d = .41 across 478 tests. Two plans, .30. Three plans dropped to d = .07, barely distinguishable from nothing. Four and five weren’t any better. Oddly, seven or eight plans scored as high as one (.42 and .41), and the authors’ own read is that they found no optimal number. Treat this as “don’t build a mid-sized portfolio”, not as a law. The pattern says a single loaded trigger beats a handful of them, which is exactly what happens when you try to redesign your whole week on a Sunday night. Pick the one that unblocks the most.

Rule 3: say it once, then leave it alone

This one is genuinely awkward for anyone who sells planners. Plans that participants rehearsed once landed at d = .50, versus .33 with no rehearsal. Rehearsing more than once added nothing.

Recording the plan looked worse, not better: d = .31 when it was written down and kept, .44 when it wasn’t. The authors’ interpretation is that people who could revisit their plans went back and deliberated, and deliberation eats the automatic quality that makes the plan work in the first place.

The practical version: write it once if writing helps you remember it, say it once, then stop editing it. A plan you keep reopening has become a decision again.

Rule 4: borrow a concrete cue, don’t invent a vague one

When participants were left to choose their own cue, that was the weakest cue condition in the entire analysis, d = .16 across 20 tests. Not an argument for someone else running your life. It’s an argument for holding your own cue to a stricter standard: a specific moment you actually hit, not “in the morning” or “when I get a chance.” The easiest way to pass that test is to hook it to something already automatic, which is what habit stacking does.

Does the effect last?

Better than you’d expect, with a caveat. Motivational boosts famously fade. In the 2025 data, the planning effect didn’t follow that curve: under a week it was d = .48, at one to four weeks .30, at one to six months .19, and at six months or more, .51.

That last number looks great and deserves suspicion. It rests on only 12 tests, with a confidence interval from .22 to .80 and high variation between studies. The authors’ own conclusion is careful: they observed no decline over time, and the key finding is that if-then plans can influence outcomes measured six or more months later. “Can,” not “reliably do.”

One more reassuring detail. The effect got bigger, not smaller, when a device or a record did the measuring instead of the participant: d = .43 for objective outcomes versus .28 for self-report. It shrank in the field compared to the lab (.27 versus .49) and when nobody was there to check the plan was written properly (.31 online versus .53 in person). Real, then, but sensitive to how carefully it’s set up.

The strongest single long-run trial is a 2019 cluster-randomised study across 48 randomised schools, 45 of which finished, and 6,155 adolescents who had never smoked, followed for four years (Conner et al., 2019). Ever-smoking came out at a risk ratio of 0.83, and smoking in the past 30 days at 0.77. Both meaningful.

Read the method before you quote it, though. Those students formed plans eight times over four years, bundled with anti-smoking messaging. That supports repeated planning as part of a programme. It does not support one sticky note lasting four years.

Is WOOP the same thing as an if-then plan?

No, and the two get merged constantly. WOOP, also called mental contrasting with implementation intentions, adds a step in front: you picture the outcome you want, then face the obstacle in your way, and only then write the if-then move. The 2025 meta-analysis explicitly excluded those tests. Different literature, different numbers.

Its own evidence is decent. A 2021 meta-analysis pooled 21 articles and 15,907 people and found g = 0.336 overall (Wang, Wang & Gai, 2021). Close to the 2025 if-then figure, and in the same “small but real” territory.

The moderator is the interesting bit, and it’s rarely quoted. When a researcher walked people through the exercise, g = 0.465. When people worked through it alone on paper, g = 0.277. Roughly 40% weaker.

That’s an uncomfortable finding for every worksheet ever sold, including the fill-in restart sheet on this site. It doesn’t make the format worthless. It does mean a self-guided sheet is the weaker delivery method, and the honest thing is to say so rather than quote the coached number and let you assume it’s yours.

FAQ

Do if-then plans double your chances of success?

No, and the claim is a category error. A Cohen’s d is a standardised gap between two group averages, not a multiplier on odds. Best current estimates are d = .36 raw and d = .15 under a Bayesian correction for publication bias, across 642 tests (Sheeran et al., 2025). Small, cheap, real.

What happens if I don’t follow my plan?

You keep going, and that’s measurable. In studies where the planned route to a goal was blocked, if-then planners made more subsequent attempts to get through than people with a plain goal (Gollwitzer & Sheeran). A broken plan doesn’t handicap you. The bigger risk is the story you tell yourself after breaking it.

Should I write my plan down or not?

Write it once, then leave it. In the 2025 data, plans that were recorded and kept performed worse (d = .31) than plans that weren’t (d = .44), while rehearsing once beat not rehearsing at all. Use writing as a memory aid, not as a document you revisit and rewrite every Sunday.

How many plans should I run at once?

One. Across the literature, a single plan came in at d = .41 while three plans landed at d = .07. That’s a between-study comparison rather than a direct trial, and the authors state they found no optimal number, so hold it loosely. It still lines up with everything else: start with one anchored habit and add the next only once the first runs itself.

Is habit stacking just an if-then plan?

Effectively yes. “After I brush my teeth, I’ll floss one tooth” is an if-then plan with a behaviour as the trigger instead of a clock. The 2025 numbers apply to it directly, which is both the good news and the correction: real effect, small size, worth doing properly.

The Bottom Line

If-then planning came out of a 642-test audit smaller than its reputation and still standing. That’s more than the willpower literature managed. It works best when you already want the thing, when you’d otherwise not have planned at all, and when the plan names one trigger, one place, one action. It works worst as a five-plan Sunday ritual you keep rewriting.

Expect a nudge, not a transformation. Four percentage points on a flu shot. A few extra portions of veg. A morning you didn’t renegotiate. Stack enough of those and it adds up, but the moment someone promises you double the odds from one sentence on a sticky note, check the study behind it. Usually there’s one, from 2002, with 248 people and a two-week follow-up.

One small action today: write a single if-then line for tomorrow, naming a time and a place and nothing else. “When I sit down at my desk at 9am, I’ll open the doc before the inbox.” Say it out loud once. Then don’t touch it again.

Alex is the voice of Self Lab: practical psychology for people who are done with motivational fluff.

Sources

• European Review of Social Psychology (open access via KOPS, University of Konstanz), Sheeran, Listrom & Gollwitzer, “The when and how of planning: meta-analysis of the scope and components of implementation intentions in 642 tests”, 2025, retrieved 2026-07-29, https://kops.uni-konstanz.de/server/api/core/bitstreams/d703c468-46e9-47fc-8900-d32d7d19c8d9/content (publisher record: https://www.tandfonline.com/doi/abs/10.1080/10463283.2024.2334563)
• National Cancer Institute (hosted chapter PDF), Gollwitzer & Sheeran, “Implementation Intentions” (summarising Gollwitzer & Sheeran, 2006, Advances in Experimental Social Psychology, 38, 69-119, and Sheeran, 2002), retrieved 2026-07-29, https://cancercontrol.cancer.gov/sites/default/files/2020-06/goal_intent_attain.pdf
• American Psychologist, Gollwitzer, “Implementation intentions: Strong effects of simple plans”, 1999, retrieved 2026-07-29, https://www.prospectivepsych.org/sites/default/files/pictures/Gollwitzer_Implementation-intentions-1999.pdf
• PNAS (full text via NBER working paper 17183), Milkman, Beshears, Choi, Laibson & Madrian, “Using implementation intentions prompts to enhance influenza vaccination rates”, 2011, retrieved 2026-07-29, https://www.nber.org/system/files/working_papers/w17183/w17183.pdf (journal record: https://www.pnas.org/doi/10.1073/pnas.1103170108)
• Psychological Science (author copy, Wharton), Nickerson & Rogers, “Do you have a voting plan? Implementation intentions, voter turnout, and organic plan making”, 2010, retrieved 2026-07-29, https://marketing.wharton.upenn.edu/wp-content/uploads/2016/10/nickerson20and20rogers_FINAL.pdf
• Journal of Consulting and Clinical Psychology (via PubMed Central), Conner et al., “Effectiveness and cost-effectiveness of repeated implementation intention formation on adolescent smoking initiation: a cluster randomized controlled trial”, 2019, retrieved 2026-07-29, https://pmc.ncbi.nlm.nih.gov/articles/PMC6474716/
• Drug and Alcohol Review (via PubMed Central), Cooke, McEwan & Norman, “The effect of forming implementation intentions on alcohol consumption: a systematic review and meta-analysis”, 2023, retrieved 2026-07-29, https://pmc.ncbi.nlm.nih.gov/articles/PMC10087331/
• Frontiers in Psychology, Wang, Wang & Gai, “A meta-analysis of the effects of mental contrasting with implementation intentions on goal attainment”, 2021, retrieved 2026-07-29, https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2021.565202/full
• British Journal of Health Psychology (paywalled abstract), Milne, Orbell & Sheeran, “Combining motivational and volitional interventions to promote exercise participation: protection motivation theory and implementation intentions”, 2002, retrieved 2026-07-29, https://bpspsychub.onlinelibrary.wiley.com/doi/abs/10.1348/135910702169420