A top stack recommendation is the composed card at the top of Unfair's ranked feed when a multi-compound protocol, taken together, outranks any single-compound change. It names three to six compounds, a dose for each, a timing schedule across one or more windows, a composite score, a confidence read, overlap and interaction flags, and a single "Activate stack" action. It is the answer to "what is the best full configuration for this goal right now," scored against the user's logs and constraints rather than against a population. When the signal for any single ingredient is cleaner than the signal for the composed set, a top supplement recommendation takes the top slot instead.
Why a stack card exists at all
Some goals do not resolve with one ingredient. A sleep-quality protocol built around magnesium, glycine, and l-theanine is a different proposition from magnesium alone — and a focus protocol built around caffeine, l-theanine, and creatine is a different proposition from caffeine alone. A stack card scores the full set, checks compatibility across the members, allocates doses to windows that do not collide, and returns a single recommendation rather than asking the user to assemble it themselves. This is the "build your own stack" surface made concrete, aligned with the mapping in the complete guide to supplement stacks.
Top-stack versus top-supplement, field by field
Most card fields are shared. The ones that differ are the ones that matter — a stack card must reason about composition, and a supplement card does not.
| Field | Top supplement | Top stack |
|---|---|---|
| Compounds | One compound, one product | Three to six compounds, each with a product |
| Score | Single-compound composite | Composed composite with an interaction penalty |
| Confidence | Per-compound, tied to that compound's history | Per-stack, gated by the lowest-confidence member |
| Dose | One dose | Per-compound doses with stack-level ceilings |
| Timing | One dose window | Two to four windows across the day |
| Overlap flags | Duplication with existing products | Duplication + interaction graph across members |
| Risk flags | Per-compound cautions | Per-compound cautions + stack-level "one stimulant at a time" rule |
| Rationale | Why this single compound | Why this configuration and why in this order |
| Primary action | "Add to stack" | "Activate stack" (replaces or extends the current plan) |
| Failure mode | Weak single trial | Weak single trial or internal conflict between members |
The single most important structural difference is the interaction penalty. A stack with two stimulants, duplicate magnesium forms, or a serotonergic member alongside existing SSRI use does not rank no matter how strong the individual components look. That enforcement comes from the stack-level rules in supplement stack mistakes to avoid — overlap, interaction checks, risk checks, and stop conditions are not advisory here; they remove the card from ranking until the conflict resolves.
How the composite score is built
The stack score is not a sum of the member scores. It is a composite with explicit caps and penalties so that no one dimension can dominate.
- Member evidence. The weighted average of each member's evidence tier score for the target goal, capped at 35 points.
- Composition fit. A scoring bonus when members target adjacent mechanisms (e.g., magnesium + glycine for sleep) and a penalty when they target redundant ones, net-capped at 20 points.
- Personal history. Each member's prior tolerance and response for this user, averaged and capped at 25 points.
- Logistics and cost band. Total cost, product availability, and schedule complexity across windows, capped at 10 points.
- Risk and overlap penalty. Subtractive; a single hard interaction zeros the composite regardless of the other inputs.
That last rule is deliberate. A stack that triggers a serotonin-excess flag or double-stacks stimulants never ranks, even if every other input looks excellent. The ranking is a filter first and a score second.
Confidence and activation gating
A stack card's confidence is gated by the weakest-confidence member. A six-compound stack where five members have high confidence and one member has low confidence reads as Medium at best, because the unknown member is the one most likely to disrupt the protocol. This is why well-tested three-compound stacks often outrank novel six-compound ones in the feed.
| Confidence | What the stack contains | Suggested action |
|---|---|---|
| Low | One or more members untested in this user's history | Queue for next cycle, test members individually first |
| Medium | All members individually tolerated, stack untested | Activate with a 28-day review and strict single-change rule during the cycle |
| High | Stack (or a near analog) previously activated by this user with tolerable outcomes | Activate with a normal 28-day review cycle |
A numeric example
A user with a "deeper sleep" goal, 10 weeks of log history, a prior magnesium trial, and no glycine or l-theanine history sees a top stack of magnesium bisglycinate 200 mg, glycine 3 g, and l-theanine 200 mg, all in the 30–60 minute pre-bed timing window. Composite score 78 — member evidence 30, composition fit 18, personal history 18, logistics 8, no interaction penalty. Confidence is Medium because glycine and l-theanine are novel for this user. The rationale snippet reads: "All three target pre-sleep mechanisms with low overlap, your prior magnesium trial was tolerated, and the single bedtime window keeps the schedule simple." The card proposes a 28-day cycle with no mid-cycle additions.
How this appears in Unfair
Top stack cards appear in the ranked feed whenever the composed composite beats every single-compound card for the active goal. The "Activate stack" action replaces the current plan for the affected windows, writes each member into the dose schedule, and schedules the next review. When a stack is activated, the feed automatically demotes competing stack cards for the same goal and surfaces a single-compound card only if a late-breaking safety flag fires on one of the members.
Clinical safety note
A top stack card respects hard contraindications rather than optimizing through them. A stack that triggers an interaction with a user-listed prescription, a known allergy, a pregnancy status, or a clinician-flagged contraindication is removed from ranking, not down-weighted. Ranking is never a substitute for clinician input, and any adverse event after activation is a reason to stop the stack and seek care before opening the app for a re-rank.