New model
Start a model from a description, a workbook or a model file you already have, and find your way around the five-step wizard.
Where: New Model
New Model opens a five-step wizard: Describe, Clarify, Build, Load data and Solve. You say what the business does and what to optimize, answer RIA's questions, sign off the design, load your numbers, and get a first solved answer, the baseline.
Start a new model
- Click New Model in the rail. Step 1 opens on Tell us a bit about your business. Click Next.
- Under How do you want to author?, pick Basic to describe your business in words, or Advanced to bring a model you already have. Click Next.
- On Name your model, type a name and click Next.
Naming creates the model. It's in your Model Library from then on, with a Draft chip until it's built. "This name will be used throughout EODE to refer to this model. Try to make it unique." Renaming it on this screen later renames it in the library too.
Other ways to open the same wizard:
| From | Click |
|---|---|
| Command palette | Ctrl/⌘+K, type build, author or import, then New model… |
| The model menu in the header | New model… |
| Model Library | New model above the list |
| Home | Author a model, shown when nothing is in flight |
Opening New Model also starts a new, empty General conversation in the RIA console if the one open belongs to a model or already has your messages in it.
Choose how to start
| You have | Pick | Then |
|---|---|---|
| A description of the business | Basic | Describe your business |
| An Excel workbook with your numbers | Basic, and attach it under Upload or select relevant datasources | Start from a workbook |
A Python model spec, a set of CSV tables, or a model file (.corz) | Advanced | Bring a model you already have |
A model is either authored from your words or brought in from a file, never both; you can switch before the build.
Find your way around the wizard
The step rail on the left lists the five steps. Under the step you're on, it lists that step's screens. A step's number turns into a check when it's done.
| Part | What it does |
|---|---|
| Step rail | Done rows are green, the current row is highlighted, rows still ahead are grey. Click a done step, or a screen listed under the current one, to go there; Enter or Space works too. Steps ahead can't be opened. |
| Previous / Next | The bar at the top of the content. A dim Next still answers a click: it says what's missing beside the button. Next often carries the step's own action instead, like Finish answering or Solve for your baseline. |
| Header | STEP 2 OF 5 · CLARIFY and a pill: IN PROGRESS, WORKING, DONE, or STOPPED when a build stopped or a workbook can't be built yet. |
| ? beside a title | Shows or hides a line of help about that screen or row. |
Clicking a step opens its first screen; Previous across a step boundary opens the last screen of the step before. Each step opens on an intro the first time you reach it, under "In this step you are able to:":
| Step | Intro |
|---|---|
| 1 Describe | Tell us a bit about your business |
| 2 Clarify | A few questions before I build, with the scrutiny choice |
| 3 Build | Here is the shape I propose |
| 4 Load data | Now the numbers; its bullets change with what the build ran on (Load your data) |
| 5 Solve | Let us solve it |
On a model started from a workbook, steps 3 to 5 list workbook bullets instead, such as "Choose the first slice: one business unit of your own rows".
Wait while eode works
Anything that runs on the server — reading your words, building, binding, solving — shows a working screen: the job's name, a progress bar and an elapsed clock, with Previous, Next and the step rail locked. "You can leave this page; the work keeps running."
| You see | While |
|---|---|
| Reading what you told me | eode reads your description after Describe |
| Reading your data / Reading your data against the model | eode reads an attached workbook, or checks it against the design |
| Checking whether anything else needs clarifying | after Finish answering |
| Building your model / Recalculating the workbook for your pick | the build |
| Accepting the design / Rebuilding with your notes | after you sign off or send the design back |
| Binding your data to the model / Mapping your workbook into the model | your data goes into the model |
| Submitting the solve | the baseline solve starts |
After 90 seconds the line reads "This is taking longer than it should. Anything that started is still running, and nothing you did has been lost." Stop waiting and go back returns you to the page you were on, notes and answers intact; it stops the waiting, not the work. The mapping wait has no way out; it moves on by itself when the mapping lands.
Come back to a model
Close the tab or go elsewhere at any point. The work carries on, and the model remembers where you were.
| To reopen | Click |
|---|---|
| From the library | Edit model (pencil) on the model's row, or click any model that isn't ready yet: a Draft, one paused for you, one building or one that failed |
| From the model | Edit model (pencil) in the model's header |
| From Home | The pencil on a model in the list, or a row under MODELS AWAITING YOU |
The pencils and Author a model are hidden for the Operator and Executive personas. New Model in the rail isn't.
Where it opens:
- On the screen you left, in this browser.
- Otherwise on the first step the model hasn't finished. Describe is filled back in from the model: your description, hard limits, goal and datasets.
- On a job's progress page if one is running, then forward by itself when it lands. A build that stops after you've moved on brings you back to Build; a build that has more questions for you opens Clarify on them.
- On Your model at Solve if the model is already solved.
Clarify answers and notes save to the model as you go, on any browser, with the question you were on: the line above the questions reads "Answers save as you go." If a save fails it reads "Your latest edits are saved in this browser only.", with Retry saving.
Your place, the Describe draft, the scrutiny choice and your notes on Load data are kept in this browser, per model. A model you started but didn't name yet comes back the next time you click New Model. Files picked under Advanced aren't kept.
Go back and change something
Previous and the rail never rebuild anything: go back and change nothing, and Next moves on without rebuilding. Changing a step that has work after it asks first. A build that's still running can't be reopened; wait for it to land or stop.
| You change | You're asked | Confirm with | Redone |
|---|---|---|---|
| Describe, then Next | Change the description? | Discard and read it again | everything after it is discarded: your Clarify answers and notes, the model design, the build, the data you loaded and the solves; the description is read again from the start |
| A Clarify answer, then Finish answering | Change your answers? | Change my answers | the build and the baseline; the data you loaded carries over wherever it still fits |
| The data choice on a solved model | Change the data? | Change the data | the baseline |
| What's used from a workbook you agreed | Change what is used from your data? | Go back to the reading | see Start from a workbook |
Change your answers? and Change the data? add "Nothing you entered is lost." Keep the description as it was backs out of a description change and puts the description back as the model has it. The check compares with the description the model holds, so a change made on another browser asks too.
Stop or cancel
Only a solve can be stopped, with Stop on its progress screen (Build and baseline). A build, a reading of your data, a mapping and the check after Finish answering can't be cancelled.
Bring a model you already have
Advanced replaces the description screens with Use advanced imports: "Pick one. The model you bring is the model: nothing is authored on top of it." Open a row with Add, fill its screens, then Finish to come back to this list. Next here runs the import.
| Row | Next reads | Brings in |
|---|---|---|
| Python model specs | Build from my spec | A spec.json your own script writes, or a .zip with it |
| CSV data packs | Scaffold on my data | Your tables of sites, products and customers, with a one-line brief |
| The third row, for a model file | Import my model | A .corz model file |
Opening a second row asks Switch imports? — "One import per model. Switching drops" the first "and opens" the second — with Switch and Cancel. The files you pick live only on the page; after a reload, Next asks "Pick the file again: a reload keeps its name, not the file."
Python model specs
- Get the starter: Starter script downloads the script; JSON schemas opens the schemas in a
new tab. Your script writes a
spec.json. - Upload your spec: "The spec.json your script emitted, or the .zip with data/*.csv inside it." Click Upload spec (.json / .zip).
- Finish, then Build from my spec.
Review structure before building is ticked by default. It holds a spec with nodes and edges for
you to review before it builds. Any other spec builds and solves straight away, ticked or not. A .zip
only works for a spec with nodes and edges; other specs carry their data inside the JSON.
CSV data packs
- Your brief and objective: one line in Model brief ("One line: what should this model decide? (e.g. weekly production mix across our plants)"), at least ten characters, and Maximize profit or Minimize cost.
- On the second screen, pick your tables:
.csvor.xlsx, one per table. A.mdfile is read as a data dictionary. Starter workbook (.xlsx) and Starter tables (.csv zip) give you the layout. - Finish, then Scaffold on my data.
The tables are read by name: products with a product column for what you sell, works with a
works column for sites, and customers with a customer_account column for accounts. Capitals,
spaces and punctuation don't matter; the words do. The tables become one dataset, and the model goes on
as if you had attached a workbook (Start from a workbook).
A model file
The third row takes a .corz file you already have; attach spreadsheets under Basic instead. Click
Import .corz, then Finish and Import my model. Tick Run a base solve after import (off by
default) to solve it and create the Base scenario in the same step. The model is imported as it is, with no
questions and no data step, and solves on its own numbers. A stopped import offers Upload the file again
on Build, which starts a new model under the same name; the stopped one stays in the library.
If an import is refused
| Refused with | Fix |
|---|---|
| "file must be a .json spec or a .zip bundle" | Upload a .json or .zip |
| "not a readable .zip bundle" / "bundle has no spec.json at its root" | Zip again with spec.json at the top |
| "spec failed validation" and a count | Correct each field the error names against the schemas |
| "only .csv tables, .xlsx workbooks and .md dictionaries are accepted" | Drop the other file |
| "your tables could not be read as a dataset, so no model was started" | Check each table has its column names in the first row, then upload again |
| A model file's reason, in the RIA console | Fix what it names and upload again |
Ask RIA instead
With New Model open, the RIA console takes a description too: "Describe your business: products, prices, capacities, demand." Its chips are Plan a 10-product bakery, Describe my own decision and Import a workbook.
Limits
- A wait shows Stop waiting and go back after 90 seconds.
- A model file can be up to 512 MB. Other uploads go through a 100 MB limit.
- A CSV data pack's brief needs at least 10 characters.
- One import per model, and one dataset per model.
- Your place in the wizard is kept per browser, not per account.
Related
- Describe your business: the description, the goal and RIA's questions
- Review the design: sign off the design before the build
- Load your data and Build and baseline: steps 4 and 5
- Start from a workbook: build on your own spreadsheet
- From description to model: why eode asks what it asks