The Recipe R&D dashboard

The Recipe R&D inbox opens with a dashboard, not just a grid of cards. Four headline numbers across the top tell you whether the playground is healthy this week. A row of category chips lets you slice the view to “just pizza” or “just cocktails” with one click. And a Grid / Kanban switch lets you flip between the photo-led card layout (good for scrolling new inspirations) and a pipeline view (good for seeing every dish in flight in every status at once).

What the four numbers mean

The four tiles at the top, left to right:

Inbox — how many candidates are sitting in New status, waiting for someone on the team to look at them. The “did I open the playground today?” number. When this climbs over twenty it usually means the AI ingester is doing more work than the chef is reviewing — time for a quick triage pass.
This week ingested — how many candidates have landed since Monday. The hint line below shows the month-to-date total too, so you can see whether you’re keeping pace with previous weeks or running cold.
Promotion rate — the percentage of all candidates the team has promoted to a draft recipe. Healthy venues sit around 25-40%. Lower than 10% means the ingester is bringing in a lot of noise the chef rejects; higher than 60% means the chef is being too generous (or hasn’t built up an active discard habit yet).
Avg days to promote — how long a typical candidate sits from ingestion to becoming a real draft recipe. Shorter is sharper. Three to five days is the sweet spot — long enough that the chef has time to think but short enough that the inspiration doesn’t go stale.

Each tile carries a coloured left-border accent matching its theme: sky for the Inbox queue, indigo for ingestion volume, emerald for the conversion-to-draft rate, amber for time-to-decision.

The rule

Healthy R&D looks like a thin Inbox, steady weekly ingestion, a promotion rate over 25%, and an average time-to-promote under five days. If you’re outside one of those, the dashboard tells you which lever to pull before you even click into a single candidate.

Below the KPIs, your recipe categories appear as a row of clickable pills with the category emoji prefix — ”🍝 Pasta”, ”🍕 Pizza”, ”🍹 Cocktails”, etc. Click one to filter the candidates below to just that category. Click it again to clear. The “All categories” pill on the left is the default state.

The pill row scrolls horizontally on narrow screens, so even venues with twenty categories don’t break the layout. The link on the far right takes you to Recipe categories if you need to add, rename, or reorder them.

Grid vs Kanban — pick the view that matches what you’re doing

Two viewing modes, switchable from the toggle inline with the status tabs.

When the task is browsing fresh inspiration

Use Grid. The photo-led card layout is built for scrolling — three columns on desktop, every card a 4:3 hero photo plus title plus the AI’s vibe summary. The status tabs across the top filter to a single bucket: Inbox for what’s waiting, Reviewed for what you’ve triaged, Promoted for what already became recipes, Discarded for the ones you’ve passed on. This is the right mode when you’ve added five new candidates and want to flip through them like a Pinterest board.

When the task is managing the pipeline

Use Kanban. Four columns appear side by side — Inbox, Reviewed, Promoted, Discarded — each holding compact cards (thumbnail + title + meta line). At a glance you see how many candidates are in each phase, who’s stuck where, and what’s flowing. The status tabs grey out in this mode because the Kanban already shows all four buckets at once.

Marco runs Friday triage. He opens Recipe R&D on his laptop. The KPI strip says Inbox 14, this week ingested 23, promotion rate 31%, average 4.2 days to promote. The pasta chip is highlighted — he’s filtering to pasta only. He flips to Kanban. He sees nine candidates in the Inbox column, two in Reviewed, one in Promoted, two in Discarded. He clicks the topmost Inbox card, opens the detail page, marks it Reviewed, comes back, sees the column updated. Total time to triage twelve pasta candidates: nine minutes.

What’s behind each number

All four KPIs (and the tab counts on the status tabs) are computed in the backend without scanning every candidate. The platform keeps a small running counter per status that updates the instant a candidate is created, reviewed, promoted, or discarded. So even when your venue has thousands of candidates accumulated over a year, the dashboard loads in the same time as on day one.

The “Avg days to promote” specifically: the platform stamps the time-to-promote on each candidate when the chef clicks Promote in the detail page. The dashboard divides the sum of all those stamps by the count of promoted candidates. Older candidates that were promoted before this stamping landed get backfilled the first time the dashboard is opened — so historical averages are accurate from the get-go.

What’s not on the dashboard yet

Per-category counts on the chips. Today the chips are filter buttons only; they don’t show how many candidates each category contains. On the roadmap.
Drag-to-move on Kanban. Today the Kanban cards are click-to-open — you change a candidate’s status from inside the detail page. Drag-and-drop directly on the board is on the roadmap, behind a permission check (we don’t want a junior cook accidentally promoting a half-finished candidate by dragging the wrong card).
Time-of-day / day-of-week ingestion patterns. Useful for venues running multiple ingestion sources. On the roadmap if anyone asks.