famichiki.com

Connecting

๐Ÿ— Famichikis sold in Japan today

๐Ÿ”ฅ 0.0 a second right now, on track for 0

๐Ÿ—พ Unfolding 16,202 storesโ€ฆ

๐Ÿ” Zoom in for ๐Ÿ—, tap a store

๐Ÿงพ Live receipt

JST

Reconnectingโ€ฆ

    ๐Ÿ“ˆ Today, hour by hour

    Expected sales per hour across all stores, using today's weather and calendar. Golden bars are already sold.

    ๐Ÿ† Prefectures

      Estimated Famichikis sold since midnight.

      ๐Ÿ— Flavor mix

        Plain outsells Red about 3.6 to 1, as in real receipt data.

        ๐Ÿง  Did you know?

        Famichiki went on sale on October 17, 2006, for ยฅ140. That's why October 17 is Famichiki Day.

        ๐Ÿงฎ How we count chickens

        FamilyMart doesn't publish store sales, so we rebuild them from public data.

        1. 1

          Start from the national total

          The series passed 2.0 billion sold in 2023 and 2.9 billion in 2026:

          Nห‰โ€…โ€Š=โ€…โ€Š(2.9โˆ’2.0)ร—1093ร—365โ€…โ€Šโ‰ˆโ€…โ€Š821,918โ€‰/โ€‰day\bar N \;=\; \frac{(2.9-2.0)\times10^{9}}{3\times 365}\;\approx\; 821{,}918\,/\,\text{day}

        2. 2

          Split it between 16,202 stores

          Each store's share depends on its type, local store density, opening hours and age:

          Diโ€…โ€Š=โ€…โ€ŠNห‰โ€‰aiโˆ‘jajai=ฯ„k(i)โ€‰g(ฯi)โ€‰ฮทiโ€‰ฮฝi\begin{gathered} D_i \;=\; \bar N\,\frac{a_i}{\sum_j a_j} \\[4pt] a_i=\tau_{k(i)}\,g(\rho_i)\,\eta_i\,\nu_i \end{gathered}

          Station ฯ„=1.5\tau=1.5, hospital ฯ„=0.8\tau=0.8; gg rises with density, with diminishing returns.

        3. 3

          Shape the day

          Each store type has a daily curve peaking at lunch, masked to the store's opening hours:

          si(m)โ€…โ€Š=โ€…โ€Šck(i)(m)โ€‰oi(m)โˆ‘mโ€ฒck(i)(mโ€ฒ)โ€‰oi(mโ€ฒ)s_i(m) \;=\; \frac{c_{k(i)}(m)\,o_i(m)}{\sum_{m'} c_{k(i)}(m')\,o_i(m')}

        4. 4

          Add the weather and the calendar

          Live weather from the nearest of 594 JMA stations, plus weekends, holidays and Famichiki Day:

          ฮปi(m)โ€…โ€Š=โ€…โ€ŠDiโ€‰uiโ€‰Okโ€‰Cโ€‰W(T,r)โ€‰si(m)WT(T)=minโก{1.25,โ€…โ€Š1+0.012โ€‰(20โˆ’T)+}\begin{gathered} \lambda_i(m) \;=\; D_i\,u_i\,O_{k}\,C\,W(T,r)\,s_i(m) \\[4pt] W_T(T) = \min\{1.25,\;1+0.012\,(20-T)_+\} \end{gathered}

          +1.2% per degree below 20 ยฐC, up to +25%; rain lowers it. uiu_i is the store's daily luck.

        5. 5

          Turn rates into real-feeling sales

          Each sale can trigger follow-ups at the same store, and national demand drifts:

          ฮปiโˆ—(t)โ€…โ€Š=โ€…โ€Š(1โˆ’ฮท)โ€‰ฮปi(t)โ€‰Mtโ€…โ€Š+โ€…โ€Šโˆ‘tj<tฮทโ€‰ฮฒโ€‰eโˆ’ฮฒ(tโˆ’tj)\lambda^{*}_i(t) \;=\; (1-\eta)\,\lambda_i(t)\,M_t \;+\; \sum_{t_j<t}\eta\,\beta\,e^{-\beta(t-t_j)}

          Since E[Mt]=1\mathbb{E}[M_t]=1, E[ฮปiโˆ—]=ฮปi\mathbb{E}[\lambda^{*}_i]=\lambda_i: same totals, bursty timing.

        6. ๐Ÿ“„ The full method

          Proofs, validation and all 25 sources, in 4 pages.

          Read the paper (PDF)