Run a what-if
Test a change to your business in its own conversation with RIA, run it when it reads right, and keep every what-if current when the model changes.
Where: Models→a model→Scenarios→New scenario
A what-if is a scenario: a copy of the model with one change you describe in a sentence, such as "What if flour costs jump 30%?" Each what-if has its own conversation with RIA, eode's assistant. You describe the change, RIA works out what it changes in the model, and nothing is solved until you tell RIA to run it. eode then compares the result with Base, the model's own solve.
Start a what-if
- On Scenarios, click New scenario in the list's header, or click the +
beside Scenarios in the rail. A new conversation opens across the page, headed New scenario.:
"Describe the what-if you want to test on
<model>: what changes, and by how much. Your first message starts the scenario. Nothing is solved until you say it looks right." - Describe the change and send it. The box's placeholder gives an example in your model's own terms. The suggestions under the box, Change a price or a cost, Change demand and Change a capacity, start a sentence for you to finish.
- Your first message makes the what-if, named from your words, and the what-if opens beside the conversation on its record. From then on, this conversation is the what-if's own: opening the what-if from the list or the rail opens it again.
Work out the change with RIA
RIA reads your change against the model and says in the conversation what it will set, and asks when something is unclear; a question with set answers shows them as buttons to click. Keep talking to adjust it, for example "make it 20% instead".
What you agree is kept on the what-if's record, beside the conversation. It has the same shape as a model's record (see Read a model's record): a top card with Since your last message, Where it stands and Needs you, the steps, and the what-if's history. A what-if that has never run opens on its record; one that has run opens on its screen, with Show the record to switch (see Open a scenario).
A what-if's conversation with nothing in it yet opens with Ask about scenario <name>, "A what-if on
<model>. Ask what it changes, what it does to the plan, or what to try next.", and four questions under the
message box: How does it differ from Base?, What is binding here?, Explain the result and What
should I try next?. Clicking one sends it as your message.
A what-if's conversation is about that what-if only. To test another change, start a new scenario, or ask in the model's conversation (Ask for a what-if in another conversation).
Run it
When the change reads right, tell RIA to go ahead, in any words: "Looks right", "Run it", "Yes". RIA runs the what-if; it never runs one before you say so.
- While it runs, the conversation says what stage the run is at and how long it has run, and the record's top card says "RIA is running this what-if." Its row in the list reads Running. You can leave; Home lists what's running.
- When it ends, RIA gives the result in the conversation. The record's Where it stands shows the result against Base and up to three of the changes it made, or "Same plan as the base." when the plan didn't move.
- The what-if's screen holds its numbers, KPIs and the diagram over its plan. Reading them is on Read results. A what-if with no plan says so, with what was found; see Read a what-if with no plan.
Stop a run
While RIA is answering or the what-if is solving, Stop sits under the send arrow beside the message box. Click it to stop RIA's turn and the run with it. Beside it eode says Stopped., or Nothing was running. when there was nothing left to stop. To start again, tell RIA to run it.
Run it again
Ask RIA to run the what-if again, for example after you change it. If you change a what-if after it ran, its record says so until it runs again: "You changed this what-if to “…”. The result below is still from “…”: run it again to update." (RIA changed… when RIA made the change), and Needs you lists "Run this what-if again, so its result is current."
In the list, Run again on a row that's out of date, or that stopped at its time limit, runs it from there.
Start from a suggestion
Starter what-ifs, on a scenario's screen in the pane, lists changes RIA thinks are worth testing on this model, numbered, each with its reason. It's open on Base and folded elsewhere; eode remembers which way you left it on each scenario. Starter what-ifs belong to the model, so every scenario shows the same list.
Click a suggestion. eode starts a new scenario conversation and sends the suggestion as your first message, so the what-if is made, named after the suggestion, and opens beside it. RIA then works it out with you as above; nothing runs until you say so. While it starts, the suggestion shows a spinner and the others are off. If the new conversation can't open, the suggestion's words wait in the message box for you to send.
Turn an outside factor into a what-if
Outside context, on a scenario's screen, lists market factors RIA is watching for this model: n market factor(s) RIA is watching. It shows once a scenario has been solved, and not on BASE DATA. Click the line to open it.
Each factor has a direction (Positive, Negative or Uncertain), a title and summary, what it Affects, its source with the date (per the site, or no source given), and Use as what-if.
Use as what-if starts a new scenario conversation the same way a suggestion does, with "What if
<factor>? Adjust <what it affects> accordingly." as your first message.
- The card heads the list with the date it was researched (RIA's reading of the news, then the date) and how many signals it found. The factors change only when you click Refresh, which searches again.
- With nothing to show, the opened line says why, with Refresh: for example "No market factors turned up for this model. Last searched …", "No search for market factors has run for this model yet. Refresh to search now." or "No industry signal was picked up automatically. Refresh to search anyway."
The same factors show on the diagram, where they can start a what-if too; see Topology.
Ask for a what-if in another conversation
In the model's conversation, or the console on the Scenarios list, describe a change you want to test and RIA proposes it in a card: Would you like me to open a separate conversation for this what-if?, with the what-if's name, RIA's reason and WHAT CHANGES. Yes, open it opens the what-if's own conversation with your request as its first message; A solve starts after Looks right. No makes nothing. See RIA's proposals.
Make a what-if a model of its own
On a solved what-if, ask RIA to make it a model of its own, in the what-if's conversation or naming it in the
model's. RIA asks in a card: Make "<name>" a model of its own? "A new model of yours, already solved
with this what-if's figures. <model> stays as it is."
- Make the model (Making the model…) makes the new model and opens its page. It's yours, whoever owns the model you started from.
- Not now, or the X (Hide this proposal), drops the offer.
- If the new model can't be made, the card says why, for example "Only a solved what-if can be made a model of its own. Run it first, then elevate it." Base can't be made a model of its own: copy the model instead (Models).
To keep the change in this model instead, make the what-if permanent; see Change a model.
Bring Base up to date
Base is the model's own solve. Opening it shows its screen beside the model's conversation, not a what-if conversation of its own. Base changes when the model is built again, in the model's conversation (see Follow a build). When Base is Out of date in the Scenarios list, Run again on its row solves it again against the current model.
Keep results current
When the model changes (rebuilt, or edited), scenarios solved before the change show Out of date in the Scenarios list: "The model changed since this ran." A scenario's screen counts them in a banner: n scenarios solved against an earlier model revision. A what-if's record says "This what-if ran before a later change, so its result is out of date until it runs again."
- Regenerate all (n) in the banner solves every out-of-date scenario again against the current model.
- Run again on an out-of-date row in the list solves just that one.
- If a change no longer fits the rebuilt model, the banner says how many may break and names them.
Until it's solved again, an out-of-date scenario still shows its value, but no change against Base in the list, the scoreboard or comparisons. A failed scenario and Base itself show no change either.
Limits
- A what-if named from your first message: up to 60 characters. Named from a suggestion, an outside factor or RIA's card: up to 80. A first message that would name it Base or BASE DATA names it What-if: … instead.
Related
- Scenarios: the list, the pane a what-if opens in, its row menu and bulk actions
- Read results
- Verify a result
- Read a model's record: the record a what-if shares its shape with
- RIA's proposals
- The RIA console: the message box, attachments and RIA's replies
Scenarios
Keep every what-if for a model in one list: search, sort and group them, open one beside its conversation, and compare, move, duplicate or delete several at once.
Read results
Read what a solved scenario says: the scoreboard, each KPI and its change against Base, how sure each number is, and where it came from.