with the skill
# Why step 3 is probably losing people
Step 3 is a cliff in effort. Steps 1 and 2 take seconds and need no thought. Step 3 asks for 5 decisions among 40 options before the person has committed anything. Several biases stack up here:
- **Paradox of choice / Hick's Law:** 40 options is a lot of cognitive load, and it's the first real decision in the flow.
- **Activation energy / BJ Fogg:** Motivation is still moderate (they're just exploring), but ability has dropped sharply. Low ability plus modest motivation means no action.
- **Regret aversion:** "What if I pick wrong and hate them?" They're choosing food for a service they haven't tried yet, so every pick feels risky.
- **Hard gate:** Being unable to continue until exactly 5 are picked turns a decision into a chore. It also removes the sense of control.
- **No investment yet (commitment & consistency):** The IKEA effect only helps once people feel invested. At this point they've only given a zip code and a plan size, so quitting costs them nothing.
- **Possible price surprise:** If they don't see per-serving or total price until step 4, step 3 is also where the "is this worth it?" doubt builds up with no answer.
# What to change
**1. Use defaults (Default Effect / Nudge Theory)**
Pre-select 5 meals, such as "most popular" or "chef's picks", and let people swap them. This changes the task from building a selection to editing one. It's likely your highest-impact test.
**2. Cut the choice set (Hick's Law)**
- Show 8–12 curated meals first, with "See all 40" as a secondary option.
- Add filters (vegetarian, under 30 min, family-friendly, etc.).
- Add tags like "Most popular" and "New", which also bring in social proof.
Show the rest (32 more lines)
**3. Add a quick preference step earlier**
Ask 2–3 questions (dietary needs, foods they dislike) after plan size. This is a small commitment that makes the later meal list feel personalized, and it filters the 40 options down to relevant ones.
**4. Remove the hard gate**
Let them continue with fewer than 5 picks and auto-fill the rest ("We'll choose for you, change anytime before [cutoff date]"). Or test moving payment before meal selection, so meals become a post-commitment customization. Exit rates tend to drop once people have paid.
**5. Reduce regret (Regret Aversion)**
Near the selection, say things like "Swap meals anytime before [date]", "Skip any week", and "Cancel anytime." Add ratings or photos of real dishes.
**6. Give a sense of progress (Goal-Gradient)**
Add a counter like "3 of 5 picked" with a progress bar, and keep the 4-step bar visible. People speed up as they approach a goal.
**7. Show price earlier (Anchoring / Mental Accounting)**
Show the per-serving price at plan selection and a running total on step 3, for example "$8.99/serving, first box 50% off". That removes the surprise before step 4.
# Check the basics before redesigning
Following Occam's Razor, rule out simple causes first:
- Does step 3 load slowly (40 images) or break on mobile?
- Is the 68% drop worse on mobile than desktop?
- Do people pick 3–4 meals and then leave? That points to the hard gate. If they leave with 0 picks, it points to overwhelm or price.
- Do session recordings show scrolling and hesitation or an instant exit?
- Are the meals shown actually available for their zip code and plan?
# Suggested test order
1. Pre-selected defaults with a swap option (smallest change, likely biggest lift).
2. Curated shortlist with filters.
3. Allow continuing with auto-fill, or move payment earlier.
4. Price visibility and regret-reducing copy.
If you can share device split and how many meals people have picked when they exit, I can narrow down which of these is most likely your real culprit.