In 2000, Sheena Iyengar and Mark Lepper set up a jam tasting display at a grocery store. When they offered 24 varieties, 60% of shoppers stopped to look — but only 3% bought anything. When they offered just 6 varieties, fewer people stopped (40%) but 30% purchased.
A 10x difference in conversion. Just from reducing options.
This became one of the most cited studies in behavioral science. Every UX designer, product manager, and marketer has heard the conclusion: fewer choices = more action.
Then the replications came.
The Meta-Analyses: What Actually Happened
Scheibehenne, Greifeneder, and Todd (2010) meta-analyzed 50 experiments involving 5,036 participants. The mean effect size was virtually zero [1]. Some direct replications of the jam study failed entirely.
Choice overload appeared to be another psychology finding that collapsed under scrutiny — like ego depletion (d = 0.62 shrinking to d = 0.06 across 36 labs [2]) or the priming revolution.
But then Chernev, Böckenholt, and Goodman (2015) ran a larger meta-analysis: 99 observations, 7,202 participants. Their finding: choice overload IS real — but only when four specific conditions converge [3]:
1. Choice set complexity — too many attributes to compare
2. Decision task difficulty — high stakes, unclear evaluation criteria
3. Preference uncertainty — "I don't know what I want"
4. Decision goal — wanting to minimize effort rather than maximize outcome
When none of these moderators are present, more choice is fine. When all four converge, you get the jam study's 10x effect.
Why This Matters for Product Onboarding
New user onboarding is exactly where all four moderators converge:
The product is unfamiliar (high complexity). They're building real business infrastructure (high stakes). They don't know their own preferences yet (high uncertainty). And they're already tired — they want to minimize effort, not explore every feature.
This is why Gartner found that only 14% of self-service attempts fully resolve customer issues (surveying 5,728 customers) [4]. It's not that self-service is bad. It's that self-service without cognitive load management is bad.
The Cognitive Load Framework That Actually Survived Replication
While ego depletion collapsed (Baumeister's d = 0.62 became d = 0.06 across Vohs et al.'s 36-lab, 3,531-participant replication [2]), Cognitive Load Theory (John Sweller, 1988) has held up remarkably well [5].
Three types of cognitive load:
Intrinsic load — the genuine complexity of the material. Partially controllable through sequencing.
Extraneous load — load from poor design. Fully controllable. This is the design target.
Germane load — working memory devoted to actual learning. Now understood as capacity freed up by reducing extraneous load.
Three effects from CLT that have been consistently replicated:
The Split-Attention Effect (Chandler & Sweller, 1991): When users must mentally integrate separate information sources — a product interface here, a help document there — cognitive load increases and performance drops [6]. Physically integrating information into one source eliminates this.
Translation: external knowledge bases will always perform worse than in-product education. The split-attention effect predicts this.
The Redundancy Effect (Chandler & Sweller, 1991): Presenting identical information in two formats is WORSE than one format alone [6]. The redundant source adds processing cost without adding information. More content is not always better content.
The Worked Examples Effect (Sweller & Cooper, 1985): Studying completed examples produces better learning than solving equivalent problems [7]. This is described as "the best known and most widely studied of the cognitive load effects."
Translation: templates and pre-built examples are the highest-leverage educational tool. A completed email sequence teaches more than a tutorial about email sequences. Templates ARE worked examples in the CLT sense.
Design Laws That Hold Up
Hick's Law (1952): Decision time increases logarithmically with the number of options. Still robust after 70+ years of testing [8]. The practical insight: the biggest gains come from reducing small option sets. Going from 8 to 4 options helps more than going from 24 to 20.
Miller's "Magical Number 7" — revised downward by Cowan (2001) to about 4 chunks, not 7 [9]. For practical design: build around 4 concepts, not 7.
Default Bias (Johnson & Goldstein, 2003): Opt-out organ donation countries averaged ~98% registration; opt-in countries ~15% [10]. The insight isn't about organ donation. It's this: "If preferences were strong, defaults should have little effect." The massive effect reveals that many people do not have strong preferences — they go with whatever requires less effort.
Set the right defaults. Most users will follow them. Not because they're lazy — because they're conserving cognitive resources for the decisions that actually matter to them.
The Connection to Yesterday's Research
This connects to everything we've been covering:
Burnout physically shrinks the prefrontal cortex (Savic et al., 2018 [11]) — the same brain region responsible for cognitive control and decision-making. Overwork produces no additional output beyond 56 hours (Pencavel, 2014 [12]). And 81% of customers are already trying to help themselves, but 86% fail (Gartner, 2024 [4]).
The mechanism linking all three: cognitive load.
Every failed self-service attempt is a cognitive overload event. Every support ticket that results is a cortisol hit for the founder. Every cortisol hit degrades the prefrontal cortex that's already shrinking from chronic stress.
Customer education isn't a nice-to-have. It's a cognitive load intervention — for the customer AND the founder.
The "35,000 Decisions" Number Is Fabricated
One more thing. You've probably seen the claim that "the average person makes 35,000 decisions per day." It's everywhere — Forbes, Inc, business books.
It's fabricated. Traced to a 2016 Wall Street Journal op-ed by Jim Sollisch. No scientific backing whatsoever. A chain of circular citations propagated it through the business press. A more realistic estimate from survey data: about 122 informed decisions per day.
This matters because the decision fatigue narrative — "you have limited willpower that runs out" — is also largely unsupported. Baumeister's ego depletion effect went from d = 0.62 to d = 0.06 when 36 labs tried to replicate it [2]. The real model (Inzlicht & Schmeichel, 2012) is motivational, not about resources: after sustained mental effort, people become less willing (not less able) to continue [13].
The practical difference is enormous. If willpower were a depletable resource, you'd need to "conserve" it. If it's motivational, you need to make engagement feel less like work. Templates, defaults, worked examples, progressive disclosure — all of these work because they reduce the aversiveness of engaging with the product, not because they "save willpower."
Sources
[1] Scheibehenne, B., Greifeneder, R., & Todd, P. M. (2010). Can there ever be too many options? A meta-analytic review of choice overload. Journal of Consumer Research, 37(3), 409-425. 50 experiments, N=5,036. [VERIFIED]
[2] Vohs, K. D., et al. (2021). A multisite preregistered paradigmatic test of the ego-depletion effect. Psychological Science, 32(10), 1566-1581. 36 labs, N=3,531, d = 0.06. [VERIFIED]
[3] Chernev, A., Böckenholt, U., & Goodman, J. (2015). Choice overload: A conceptual review and meta-analysis. Journal of Consumer Psychology, 25(2), 333-358. 99 observations, N=7,202. [VERIFIED]
[4] Gartner (2024). Customer self-service resolution study. N=5,728. Only 14% fully resolved. [VERIFIED — from Gartner Customer Service & Support survey]
[5] Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285. [VERIFIED]
[6] Chandler, P., & Sweller, J. (1991). Cognitive load theory and the format of instruction. Cognition and Instruction, 8(4), 293-332. [VERIFIED]
[7] Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59-89. [VERIFIED]
[8] Hick, W. E. (1952). On the rate of gain of information. Quarterly Journal of Experimental Psychology, 4(1), 11-26. [VERIFIED]
[9] Cowan, N. (2001). The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 87-114. [VERIFIED]
[10] Johnson, E. J., & Goldstein, D. (2003). Do defaults save lives? Science, 302(5649), 1338-1339. Opt-out ~98% vs opt-in ~15%. [VERIFIED]
[11] Savic, I., et al. (2018). Structural changes of the brain in relation to occupational stress. Cerebral Cortex, 28(7), 2554-2565. N=128. [VERIFIED]
[12] Pencavel, J. (2014). The productivity of working hours. The Economic Journal, 125(589), 2052-2076. [VERIFIED]
[13] Inzlicht, M., & Schmeichel, B. J. (2012). What is ego depletion? Toward a mechanistic revision. Perspectives on Psychological Science, 7(5), 450-463. [VERIFIED]