2b:null 2c:["$","title",null,{"children":"Understanding Optimal Batch Size | Troy Magennis | Observable"}] 31:T456,chart = { const svg = d3.create("svg") .attr("viewBox", [0, 0, width, height]); svg.append("g") .call(xAxis); svg.append("g") .call(yAxis); svg.append("path") .datum(data) .attr("fill", "none") .attr("stroke", "red") .attr("stroke-width", 3) .attr("d", lineTotal); svg.append("path") .datum(data) .attr("fill", "none") .attr("stroke", "steelblue") .attr("stroke-width", 3) .attr("d", lineTransactionCost); svg.append("path") .datum(data) .attr("fill", "none") .attr("stroke", "orange") .attr("stroke-width", 3) .attr("d", lineHoldingCost); /* svg.selectAll('circle') .data(data) .enter().append('circle') .attr('r', 5) .attr('fill', 'orange') .attr('cx', d => x(d.p)) .attr('cy', d => y(d.time)) */ svg.append("text") .attr("transform", "translate(" + (width/2) + " ," + (height + 0) + ")") .style("text-anchor", "middle") .text("Batch size"); return svg.node(); }32:T557,md`### Observation 1: Batch size of "One" only best if setup cost is zero In the ever increasing race for smaller batch sizes, one is only the lowest cost if it is free to setup and deliver each item individually. This is where automation is key. If deployment is fully automatic, then a batch size of one make sense. If it isn't, there will be a > 1 batch size that will be more economically viable. Your goal should be to find ways to reduce the initial setup cost of your proesses. One example is performance and load testing. It can often be expensive to deploy to a test environment and then saturate that environment with simulated load. Depending on how expensive, it might be more economic to do this in batches. Similarly, text translation. Often translation vendors have a per job fee, so doing one string at a time doesn't make sense. But, waiting too long to do either string translation or performance testing could cause problems in its own way be delaying learning or delivery of a valuable feature. On the counter side of this argument (and mentioned below) if one item is massively expensive in holding cost (a system outage for example) doing that item alone makes economic sense. The title of this observation should probably be **"Batch size of "One" only best if the Cost of Setup is less than the Holding Cost (cost of delay) of that item."**`33:T8a6,data = [ { count: 1, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 2, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 3, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 4, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 5, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 6, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 7, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 8, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 9, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} }, { count: 10, get holdingCost () { return HoldingCostEquation(this.count);} , get transactionCost () { return TransactionCostEquation(this.count);}, get total () { return this.holdingCost + this.transactionCost;} } ]34:[] 2f:{"id":"ea2d7e3c614c5f4f","slug":"understanding-optimal-batch-size","title":"Understanding Optimal Batch Size","thumbnail":"72aeecdd63db1bf55fcc975048935ee62756b8888a0db505aaed87ce074ea8af","access_level":"public","create_time":"$D2019-07-29T17:46:08.625Z","update_time":"$D2019-11-01T17:18:17.644Z","view_time":null,"first_public_time":"$D2019-07-30T19:06:06.721Z","fork":false,"stars":1,"starred":false,"owner":{"type":"team","id":"5bfcd1f236454eeb","login":"troymagennis","name":"Troy Magennis","avatar_url":"https://avatars-next.observableusercontent.com/bd84c523cb40d1d089b9d8d0a4ac5e025e3593ce8554df5d578d6da6190496c5","tier":"starter_2024"},"version":178,"publish_version":178,"role":null,"authored":false,"authors":[{"type":"individual","id":"f950cac864488237","login":"troymagennis","name":"Troy Magennis","avatar_url":"https://avatars-next.observableusercontent.com/bd84c523cb40d1d089b9d8d0a4ac5e025e3593ce8554df5d578d6da6190496c5","tier":null}],"members":[{"type":"individual","id":"f950cac864488237","login":"troymagennis","name":"Troy Magennis","avatar_url":"https://avatars-next.observableusercontent.com/bd84c523cb40d1d089b9d8d0a4ac5e025e3593ce8554df5d578d6da6190496c5","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"subscribed":null,"forks":7,"collection":{"id":"e353f61379db4140","slug":"agile-software-development","title":"Agile Software Development","owner":{"id":"5bfcd1f236454eeb","login":"troymagennis","name":"Troy Magennis"}},"type":"notebook","body":{"title":"Understanding Optimal Batch Size","cells":[{"id":0,"value":"md`# Understanding Optimal Batch Size\nFor any production and delivery process there is two cost components that combine for total cost of any batch size. The transaction cost for setting up to do the production of that work, and then the holding cost of lost revenue between shipping that work. To calculate the optimal batch size for delivery of work, we need to understand how these two costs interact.\n\nThis calculator will graph the total cost given the setup cost for doing a batch and the holding cost for each item delayed by doing a larger batch size.

`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":94,"value":"viewof costSetup = slider({\n min: 0, \n max: 10000, \n step: 100, \n value: 1000, \n format: \"$\",\n title: \"Batch Setup Cost - the cost it takes to do the first item (the subsequent items are essentially free)\", \n description: \"How much does it cost to setup to perform or deliver any size batch.\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":99,"value":"viewof costHolding = slider({\n min: 0, \n max: 1000, \n step: 10, \n value: 100, \n format: \"$\",\n title: \"Holding Cost Per Item - cost of not delivering one item to receive value\", \n description: \"How much does it cost to delay delivery of one item. For example, how much value isn't delivered for each item on average. (Cost of Delay)\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":113,"value":"html `

Optimal Batch Size - Look for the lowest total cost

Legend: Total cost (setup + holding), Setup cost, Holding cost`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":86,"value":"$31","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":162,"value":"md `\n### How does this calculation work?\nThis calculation is a very simplistic model originally based upon manufacturing and delivery of goods. \n\nThe batch setup cost is computed simply by amortizing that cost over the batch size. Batch size of one means total cost for that one item. Batch size of ten, means that setup cost is 1/10 per item (ten times less). This causes the decaying pattern as batch size gets larger. \n\nThe holding cost curve is the addition of holding cost for the batch size. This causes a linear increasing holding cost, but be careful because in the software context, not every item is the same holding cost, so your milage may vary. If you have high value items being held back to finish a batch, consider going to a smaller batch size.\n\nTotal cost is just the sum of these two costs.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":148,"value":"$32","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":152,"value":"md`### Observation 2: Long U Shape (don't sweat about the upper end)\nThere is significant cost at the very low extreme, falling to th lowest cost, before a long ascent to high cost again. This means that whilst it is essential to not go too low, after a certain point the broad U shape means \"near enough is good enough.\" Batch size is a blunt instrument at the higher end of batch sizes.\n\nAdvice: Round up a few batch items after the lowest cost to make certain you avoid the high cost of going too low.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":156,"value":"md`### Observation 3: Beware - Holding Cost Isn't Always Linear\nIn a manufacturing context, the holding cost of every item is equal. This is unlikely in the software world, where a few items are likely orders of magnitude more important. Consider this when setting batch size. If an item is urgent and has massive delay cost, it WILL BE lowest cost to deploy that item alone. Use this calculator as a guide, but always consider the balance of setup and holding cost in YOUR context.\n\nAdvice: If you have high cost of delay items in a batch, go lower in batch size equal to the top 80% of holding cost in the batch value.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":169,"value":"md `#### Reference\n\n[The Principles of Product Development Flow: Second Generation Lean Product Development](https://amzn.to/2LPJItU) by by Donald G. Reinertsen \n\n[Factory Physics](https://amzn.to/2LMXrlr) by by Wallace J. Hopp, Mark L. Spearman\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":135,"value":"md `

\n#### Training and consulting - contact us...\nFocused Objective offers training and consulting on metrics and forecasting applied to software development. If you are interested in on-site or public workshops to teach how to plan and manage software development using practical and proven techniques please email me: [troy.magennis@focusedobjective.com](mailto:troy.magennis@focusedobjective.com). If you are interested in following more of my work on Twitter connect at [@t_magennis](http://twitter.com/t_magennis). Always happy to spend an hour on the phone to see if I can help in any way.\n

`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":133,"value":"md `## Inner workings of this notebook. Ignore unless interested`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":60,"value":"height = 500","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":62,"value":"margin = ({top: 20, right: 30, bottom: 40, left: 40})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":68,"value":"y = d3.scaleLinear()\n .domain([0, d3.max(data, d => d.total)])\n .range([height - margin.bottom, margin.top])","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":72,"value":"yAxis = g => g\n .attr('transform', `translate(${margin.left}, 0)`)\n .call(d3.axisLeft(y)\n .tickFormat(d3.format(\"$\"))\n )","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":70,"value":"x = d3.scaleLinear()\n .domain(d3.extent(data, d => d.count))\n .range([margin.left, width - margin.right])","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":74,"value":"xAxis = g => g\n .attr('transform', `translate(0, ${height - margin.bottom})`)\n .call(d3.axisBottom(x))","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":80,"value":"lineTotal = d3.line()\n .defined(d => !isNaN(d.total))\n .x(d => x(d.count))\n .y(d => y(d.total))","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":81,"value":"lineHoldingCost = d3.line()\n .defined(d => !isNaN(d.holdingCost))\n .x(d => x(d.count))\n .y(d => y(d.holdingCost))","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":83,"value":"lineTransactionCost = d3.line()\n .defined(d => !isNaN(d.transactionCost))\n .x(d => x(d.count))\n .y(d => y(d.transactionCost))","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":4,"value":"$33","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":7,"value":"function HoldingCostEquation(x) {\n return x * costHolding;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":10,"value":"function TransactionCostEquation(x) {\n return costSetup / x;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":17,"value":"d3 = require(\"d3@5\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":24,"value":"import {slider} from \"@jashkenas/inputs\"","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"}],"files":"$W34","stdlib":"1","titling":"auto"}} 30:"$2f:body"