Samsung Food: Tailored Plan

Context

The casual planner had shipped: a list view with a check per meal, and tailored meals reachable from the add flow. Adoption of the recommendations was still gated on intent. A user had to open an empty planner, go to the tailored plan, add meals to their own plan, then add from their plan to a shopping list. Four manual steps, each a drop-off point, from an empty and uninspiring entry point. From memory, roughly one in eight active users planned, and planners were the stickiest cohort, which is what made the entry point worth the investment.

The hypothesis on the board from a 1:1 in November: follow the prepopulated pattern (Mealime, MealPrepPro, Innit). Generate the plan, let the user tweak it, and derive the shopping list from it. Opt-out rather than opt-in, with loss aversion doing some of the work. Two how-might-we questions framed the workstream: move people through planning faster to reach the next phase, and reach the moment of a populated plan and shopping list in a fraction of the time and effort.

From an empty planner to a generated one

Iterations 0 to 3 settled the container. Iteration 0 explored a prepopulated list and a prepopulated calendar in parallel, each reviewed with the PO. Iteration 1 added tailored plan variants and plan collections. Iteration 2 carried the current versus future diagram and two ways to connect an opt-out plan to the shopping list: a short confirm-and-generate flow that verifies the pantry and removes unwanted items, or surfacing the plan’s contents inside the list to confirm before adding. Iteration 3 fixed the two candidates as test variant A (list) and test variant B (calendar) and reviewed them with Growth.

01Setup

Setup flow

Iterations 4 to 6 built the setup. First access from the empty planner or a home placeholder: which meals to plan for (weekday dinners recommended as the start), meal preferences prepopulated from onboarding and Samsung Health data, a preview of what the plan would look like, ingredients at hand to seed the shopping list, and final toggles for a weekly plan, weekly shopping and reminders. Cancelling returns to the empty plan. Six questions were reduced to five after the previz split one step in two and the design team agreed it worked as one.

A jam with the second designer added the date picker variants, the flavour profile (diet, avoidances, favourite cuisines, dislikes), and the setup question that shaped the plan most: how many times do you want to cook per week, since fewer cooks means fewer distinct meals and more leftovers planned. Iteration 6 added plan states, plan review options and the generating state, an animation on the Recipe AI pattern, with a copy note that it may take up to twenty seconds.

Design reviews and research

Iterations 7 and 8 went to the wider team as design reviews, each page carrying its questions beside the screens. Two forms of setup were put side by side: the stepped flow, and a conversational setup generated with ChatGPT, a chat that asks the same questions in turn and ends in the same plan. Two forms of review were put beside each other too: stepped review, as in setup, or a single page review. The freemium section appeared here for the first time, with three entry points explored. Iteration 9 added the shopping integration and the review-day model: Sunday is review day, the next plan is prepared as the current one ends, and the user is told in app and by email with time to adjust before it starts.

Research ran alongside the reviews, on the same cadence as the casual planner: moderated sessions on the list and calendar variants to inform the container decision, then on the setup flow and the plan review to validate them once direction was agreed. The research files sit outside the design files and are not reproduced here; the decisions below carry their outcomes.

Choosing the setup model

The design reviews and the research sessions settled three things. Stepped setup over conversational: the chat read well as a demonstration and cost more to build, review and localise for the same inputs. Single page review over stepped review, since a returning user is changing one setting, not answering five questions. And a freemium gate, with three variants taken to handover so the PO could pick one against the subscription data.

02Around the plan

Shopping, leftovers, nutrition

Iterations 10 and 11 hung the rest of the planner off the generated plan. Shopping list integration in two variants: add missing ingredients at generation, or a review missing ingredients banner on the plan; both resolve to the same week-of picker in Lists. Leftovers in made it, with portions auto-scheduled to tomorrow’s lunch and a manual override. Meal history in two variants. Nutrition in the prepopulated plan: calories per day and the macro split in the plan header and on each day, from the same data the health dashboard reads. Notes, daily view, date variants and header variants filled out the rest of iteration 11.

03Handoff

Two handovers

Handover one, iteration 10, carried setup and the three freemium variants: three free plans with a counter, limited meals per plan, and straight to the paywall. Handover two, iteration 12, carried the shopping list integration, the food list integration and swap. Each handover page is the journeys only, with the review pages left behind it, so engineering reads one page per release.

What was cut, and what shipped

The left column is every option that asked the user for more than the plan needed, or asked engineering for more than the first release needed. The right column is what a generated plan needs to be accepted, kept and shopped for.

One field to add anything

Iterations 12 and 13 took on the add flow itself, starting from data. The PO had captured what users typed when they created a new recipe from inside the planner, which made little sense mid-planning, and I ran the set through ChatGPT for themes: 41% were referencing dish names, 22% adding food items or ingredients, 18% quick-saving online recipes, 12% managing leftovers and meal prep, 7% creating a full recipe. The same analysis on user notes, run through Claude, found the same pattern independently: 42% meal names and basic descriptions, 20% leftover and snack planning, 14% placeholder entries.

The rule that followed replaced the add menu with a text field. If the query contains a meal with an exact match in the database, return it; if it contains a meal with no match, generate one with an exact match; beneath the match, surface leftovers, saved, collections, communities and creators the user follows. If the user has pasted a URL, import the recipe. Otherwise it is a note. High confidence adds without a prompt; low confidence asks. Auto notes came from the same rule: a note that names meals offers to add them, and asks whether to keep the text.

Outcome

Handover one and handover two went to engineering in the first quarter of 2025, and the Tailored Plan setup and generation were in delivery when I left Samsung Food. The workstream closed the loop the health dashboard and the casual planner had opened: onboarding and Samsung Health data drive a generated plan, the plan drives the shopping list and the leftovers, and the plan’s nutrition reads back into the dashboard. The add field and auto notes stayed in design as the next brief. Samsung’s IFA 2024 announcement carried the first public version, “Tailored for You” plans with three days free and a full week on Food+, and the help centre now describes the For You section as a redesigned version of the Tailored Plan. From memory, after release almost every active user held a tailored plan and regular planning rose well beyond the earlier one in eight; the figures are not public.