How Gardiv calculates a garden plan

What Gardiv is—and is not

Gardiv is a deterministic planning model. The same validated inputs and data version produce the same allocation priorities. It does not call a language model, live weather service, or product-price feed when you create a plan.

The output is a conservative starting point. It cannot know your cultivar, soil biology, irrigation consistency, pest pressure, microclimate, skill, or the weather ahead. Dates, yields, grocery values, and plant counts are therefore ranges or editable assumptions—not guarantees.

Source hierarchy

1. Government and university extension publications with clear production context
2. Reviewed trials and crop-planning tables
3. Reputable seed-provider guidance where public sources leave a practical gap
4. Visible Gardiv owner assumptions for preference, maintenance, loss, and editable price

Numeric crop outputs carry one or more source IDs. The source registryshows the title, organization, access date, and fields each source informs. A source can support the range without implying that every home garden recreates the source’s conditions.

Plant-count range

For each crop, household members become adult-equivalent consumers:

adultEquivalent = adults + children × childServingFactor

Seasonal demand then combines appetite and preservation:

targetServings =
  adultEquivalent
  × defaultServingsPerAdultWeek
  × freshEatingWeeks
  × preferenceFactor
  × preservationFactor

Expected servings per plant multiply yield by servings per unit and surviving share. Gardiv uses the higher of the crop’s normal loss assumption and the loss buffer you enter.

plantCount = ceil(targetServings ÷ estimatedServingsPerPlant)

Low, typical, and high counts reverse the crop’s yield range: a strong yield needs fewer plants; a weak yield needs more. “Must include” creates a hard attempt to fit at least one plant, not a promise that physics will cooperate.

Space and fit

Raised beds use reviewed square-foot density when it exists, plus a 15% breathing-room factor. In-ground plots instead multiply typical in-row and between-row spacing, then add 20% breathing room; this preserves conventional cultivation rows rather than treating a field like one large intensive raised bed. Vertical crops can receive a conservative footprint factor when support is available.

areaPerPlant = 1 ÷ squareFootPlants

or

areaPerPlant = inRowInches × betweenRowInches ÷ 144

The MVP uses a deterministic greedy allocation. It attempts must-grow crops first, rejects crops that cannot meet container or support constraints, scores remaining crop units for the selected objective, and adds units while physical space remains.

The engine reports unmet demand. It does not silently shrink spacing or invent vertical room to make an impossible request appear complete.

The allocator first aims for each crop’s typical household target. One of three optional-extra presets then sets a seasonal surplus allowance: Demand-led allows up to 50%, Balanced coverage up to 100%, and Full beds up to 200%. Full beds is the default. The selected percentage can apply to a very small target—for example, four optional plants beyond a two-zucchini target—but the allowance follows a square-root taper as counts grow, so one hundred carrots do not become three hundred merely to color the map.

seasonExtraBudget = floor(
  presetScale × min(seasonTypicalTarget, 2 × sqrt(seasonTypicalTarget))
)

perRoundExtra = floor(seasonExtraBudget ÷ plannedRounds)

Optional extras are labeled in the result and remain bounded by actual area, spacing, support, and container constraints. Succession rounds share the one seasonal extra budget instead of each receiving a fresh allowance. The allocator activates compatible bed interiors and distributes coherent crop groups before tightening an already-used area. It then gives remaining visual area to wider crop zones, row spacing, and working room instead of adding plants without limit. A raised-bed zone can widen to at most twice its compact assigned footprint, or four times for high-density roots and greens whose square-foot baseline is intentionally intensive. In Full beds plans with multiple raised beds, a final pass works from the emptiest bed upward when compact occupancy is below 50%. It uses only selected annual crops and can add at most two broad-crop or four small/dense-crop positions per crop beyond the normal tapered allowance. The result reports that exception explicitly. Extras from one crop never hide unmet demand for another. Optional fill keeps a small packing reserve—10% in raised beds, 15% in in-ground rows, and 8% in containers—and a final geometry check removes positions rather than displaying them closer than the stated spacing.

When Full beds is selected for a large path-divided in-ground plot, succession crops use no more than two larger frost-safe rounds. This preserves the same seasonal demand and surplus budget, but replaces many small repeat sowings with longer rows and fewer rounds of work. Raised beds and smaller plots retain the ordinary demand- and climate-limited succession schedule.

Permanent crops are allocated before annual crops and remain together in one permanent bed when their requested seasonal positions fit. Permanent trellises have linear capacity; compatible cages and stakes are counted one support per plant. A selected support with no entered limit means the gardener is willing to obtain as many as the plan needs; an entered limit caps the plan at that inventory. When a compatible cage or stake is enabled, the planner assigns it before consuming permanent trellis capacity, leaving trellises available to crops with fewer support options. An explicit vine choice to climb a trellis still takes precedence. Optional supported plants remain subject to the selected tapered seasonal allowance, with stronger preferences winning when finite support is scarce. In-ground plots can also allow freestanding row trellises. Those supports satisfy climbing-crop requirements without pretending that the crop occupies an exterior cardinal edge.

Round pots use circular area (π × radius²) and exclude square corner cells from the map. Round pots under two feet in diameter use conservative one- or two-plant limits for ordinary crops. If a small pot is the only available space, a dense crop may share concentric zones with one central crop. Rectangular containers and beds continue to use length × width.

Suggested bed layouts

In-ground plots wider than four feet default to an optional access-path layout. Gardiv keeps each cultivation strip no more than about three feet wide and separates strips with two-foot walking lanes. The paths follow that space’s selected row direction; Mixed chooses the plot’s long axis. Rows may continue for the full plot length in that direction: access comes from the neighboring lanes instead of artificial cross-paths every eight feet. The reach convention is consistent with guidance from theUniversity of Nevada, Reno Extension; the University of Maryland Extension describes two-foot paths. These dimensions are transparent Gardiv defaults, not a site-specific accessibility standard. Paths reduce the reported plantable area, interrupt crop zones, and appear directly on the map. Users can turn them off when their cultivation method or existing path system makes them unnecessary. The resulting rhythm is also consistent with theUniversity of Minnesota Extension example of approximately three-foot production beds with two- to three-foot paths.

Within those strips, tall crops receive the poleward band, medium crops the middle, and low crops the equatorward band. A nearby unused strip is preferred over crowding an unrelated crop into an occupied strip, and a crop that needs a second strip continues into an adjacent one. This keeps crop rows grouped while distributing useful modules across the field. Any plantable area still outside a mapped crop zone is reported honestly as room for another selected crop, a cover crop, or mulch instead of being filled with arbitrary extra roots or greens.

Each physical bed or container receives its own map with north at the top. Entered length runs east–west, entered width runs north–south, and a space quantity greater than one becomes separate maps. The arrangement canvas stores each space’s offset from a shared northwest origin. In the Northern Hemisphere the preferred tall-crop and trellis edge is north; in the Southern Hemisphere it is south. East and west support remain available, while the opposite, equator-facing edge is omitted to reduce avoidable shade.

Trellised plants consume the selected edge’s linear capacity using typical in-row spacing and are placed in a continuous strip against that edge. Other crops form balanced rectangular sections sized to hold at least one complete crop footprint in both directions. Tall crops stay generally north of shorter crops in the Northern Hemisphere and south of them in the Southern Hemisphere, though adjacent height groups may share a row when that avoids an unusably long, narrow section. The engine groups repeated plants of the same crop, uses stable east–west ordering, and reverses that ordering in the second of two paired nearby beds. This varies the visual sequence; it does not claim that differently composed beds are exact mirror images. Permanent crops such as strawberries, chives, and thyme remain in separately marked perennial zones. Small round pots use concentric zones instead of rectangular lanes.

Pumpkin, winter squash, melon, and cucumber ask whether vines should cover the bed, spill over a chosen edge, or climb a selected trellis. Spill and trellis plants share the edge’s linear capacity. The reduced footprints are visible Gardiv planning assumptions, not cultivar-specific guarantees; heavy fruit on a trellis needs sturdy support and slings.

Ordinary rows default to east–west in each rectangular space. A gardener can instead select north–south or Mixed, which lets each crop use the better-fitting direction. A physical trellis or spill edge overrides that preference because the attached row must follow its edge. The resolved direction appears in the legend. The bed maps and plant-spacing calculator share one point-placement engine. Plant symbols and their softly shaded footprints are scaled from typical in-row and between-row spacing over a six-inch and one-foot grid. The point engine chooses a balanced row-and-column arrangement, offsets alternate rows by half the in-row spacing whenever every assigned footprint fits, and keeps the full shaded footprint inside each boundary. North–south rows use the same treatment as offset columns. A tightly bounded offset row may safely hold one fewer edge plant than a square grid; managed support lines and tightly packed sections remain straight. Broad bush crops reserve the smallest complete spacing grid that fits their plant count and allowed row direction; dense neighboring crops can fill the remaining rectangular strips. A few selected marigold, calendula, or nasturtium plants may use opposite corners around one large bushy or spreading plant when the reserved footprints allow it; this is space-sharing geometry, not a claim of a biological companion benefit. Planner sections render up to 5,000 markers and the calculator renders up to 10,000 before evenly sampling. The numbered legend remains the planning record for full plant counts, position, growth profile, trellis use, and succession rounds; the seed packet should overrule the map when a cultivar needs different spacing.

A whole-garden overview repeats the saved bed positions and colors zones by crop height. For close north–south bed pairs it reports whether the equatorward bed contains a tall crop that may shade its poleward neighbor. This is a planning check rather than a sun-angle simulation; buildings, terrain, season, and latitude still matter.

Whole-garden coordinates make tall crops prefer poleward beds—north in the Northern Hemisphere, south in the Southern Hemisphere. Beds separated north-to-south by less than 1.5 feet share the same directional decision: tall crops avoid casting toward a poleward neighbor when another eligible bed exists, and a warning remains when the risk is unavoidable. This is a directional shade precaution, not a sun simulation. Gardiv does not know the position of buildings, trees, slopes, tunnel clearance, irrigation, or observed shade, and it does not make companion-planting claims. Radial placement indicates geometry, not a claim of beneficial companion planting. Adjust the suggested zones for access and site conditions before planting.

Frost-relative dates and succession

Indoor start, direct-sow, and transplant windows are stored as week offsets from the average last spring frost. Expected first harvest adds typical days to maturity to the relevant outdoor planting date. The harvest window is capped at the first fall frost for presentation.

Succession is optional. When enabled, succession-capable crops receive rounds separated by their typical interval. Climate-feasible rounds are reduced when dividing the household target would create impractically tiny batches (for example, only one or two radishes at a time). The map's plant count is the number of simultaneous planting positions used in each round. Seasonal household demand is divided across the planned rounds before those positions are allocated, so succession does not multiply an already-seasonal target. The interface also reports the planned season total across all rounds. Eligible harvested zones can hand their area to another selected, direct-sown crop. A handoff starts three days after expected last harvest and must fit the crop's latest planting date, typical maturity, freed area, and first-frost limit. Clear, sow, and harvest actions appear in the same timeline and exports. Repeated rounds of the primary crop are included in its seasonal yield range; yield from a different follow-on crop is not. Gardiv does not yet model a summer heat pause, so current conditions should overrule a suggested round.

Crop fit and discovery

The normal crop list and guided starter list use the same deterministic viability score. It combines frost-window maturity, a short-season warning for very tender crops, the brightest entered bed, container and support fit, maintenance, and beginner suitability. The six sunlight inputs are shade, about two hours, 3–4 hours, 5–6 hours, 6+ hours, and 8+ hours. These are coarse direct-sun bands, not a site survey. Stronger matches appear first; marginal crops remain searchable with a reason or warning.

Yield and grocery value

seasonPlantings = positionsPerRound × plannedRounds
yieldRange = seasonPlantings × yieldPerPlantRange
grossValue = yieldRange × editableLocalPrice
annualizedSetup = setupCost ÷ usefulLifeYears
netPlanningValue = grossValue - annualizedSetup - annualInputs

Catalog retail values are Gardiv owner assumptions, clearly marked low-confidence and editable in the value calculator. They are not scraped or live local prices. Labor is excluded by default because many people garden partly for recreation; add labor when financial return is the decision you are testing.

Confidence and correction

High confidence usually means a biological relationship such as one beet root per surviving thinned plant. Medium confidence indicates useful trial or row-derived planning context. Low confidence marks broad home-garden quantities that vary sharply.

To suggest a correction, email data@gardiv.com with the crop, field, proposed range, authoritative source, and why the current value is misleading. Gardiv records reviewed changes in the repository rather than silently changing facts from an automated scrape.

Known limitations at launch

• Manual frost dates; optional NOAA nearest-station lookup is not yet active
• No live weather, soil-temperature, or microclimate model
• No cultivar-specific maturity or plant-habit selection
• Approximate crop-zone geometry rather than a surveyed landscape or sun-angle engine
• Spill-edge clearance and vine reach are not surveyed against paths or neighboring beds
• No site-specific shade simulation, detailed light competition, or companion-planting claims
• Automatic paths use fixed two-foot lanes; they do not model mobility devices or equipment
• No multi-year crop-rotation model in the free plan

The product earns trust by exposing these limitations. A narrower honest tool is more useful than a precise-looking answer that cannot explain itself.