Financial modeling — what-if scenarios
Lesson 4 · 2 minutes 51 seconds · Free to learn · Words printed on screen · Watch above, then try the exercise below.
What you'll learn
- How to build one month three ways: best, likely, and worst
- The one-lever rule: change a single number at a time
- How to follow the cushion month by month, down the worst road too
- How to spot a case that can't be right — and what to send back
Key points
Take the model from Lesson 3 and move one lever: sales, up or down by a quarter. Ingredients follow sales — they were $3,600 on $12,000, thirty cents of every sales dollar — while help, rent, and the rest stay put at $6,900. Three versions of the same month appear: best, $15,000 in and $3,600 left; likely, the month you already know; worst, $9,000 in and $600 under. Read the worst case first — it sizes the hole. Then carry the cushion forward: from $3,000 saved, three likely months grow it to $7,500, and even three worst months in a row leave $1,200. The bad road is survivable, and you know it before you live it. Last, the spot-it test: if a draft ever shows the worst month leaving more than the best month, the case is broken however tidy it looks — send it back and ask for each case explained in plain words. Camp rule: change one number at a time — and never trust a case you can't explain.
Try it
- Open your model from Lesson 3 and ask for the same month three ways, with sales up and down by a quarter.
- Read the worst case first. Then ask the model to carry your cushion through three likely months — and three worst ones.
- Run the spot-it test on every case: does worse always leave less? If anything looks impossible, ask for the case in plain words before you trust a single number.
The words in this video
Prefer to read, or want to follow along while you watch? Here is everything the narrator says in this lesson, word for word.
Welcome back to AI Camp, business track — the last lesson of the series. Last time, you built a model and checked its math. Today we play with it: what if? Every owner asks the same three questions at night. What if business booms? What if it sags? What if it just carries on? A model lets you answer all three — on paper first. So we build the same month three ways. The best month you can honestly imagine. The likely one — the month you already know. And the worst one you should be ready for.
The ask is one line longer than last time. Same bakery month, shown three ways — with sales moving up or down by a quarter. AI rebuilds the page three times — the costs behaving the same way, only the sales line moving. A minute later, you hold three versions of the same month. Here they are. Best month: fifteen thousand in, three thousand six hundred left. Likely — the month you already know. Worst: nine thousand in, and the left-over line goes six hundred under.
Read the worst case first. It is the most useful page of the three — it tells you exactly how big the hole is if the slow month comes: six hundred, covered by the cushion. Now the discipline that makes what-ifs honest: change one number at a time. Move two, and when the answer changes, you cannot tell which one did it. In our cases, only sales moves. Ingredients follow sales — more loaves, more flour. Rent and help stay put. That is the whole trick: one lever, pulled gently.
A single month is a snapshot. What you really steer by is the cushion, month by month — the savings the good months pile into, and the bad months draw out of. So ask the model to carry the cushion forward. Start where the bakery stands today, then walk it through the likely months — and down the worst road, too. Start with three thousand saved. Three likely months grow it to seven thousand five hundred. Even three worst months in a row leave twelve hundred. The bad road is survivable — and now you know it before you live it.
But what-if pages can be wrong in a sneakier way than bad adding. A whole case can be impossible — and still look neat and tidy. Spot this one. The draft says the worst month leaves more than the best month. Worse sales, more left over? That cannot be. The case is broken, however tidy it looks. When a case smells wrong, send it back with one instruction: explain each case in plain words. If the words do not make sense, the numbers will not either. Rebuild it.
Today's camp rule: change one number at a time — and never trust a case you can't explain. Try it on your own model from last time. Pull one lever — sales down by a quarter — and walk your cushion through three worst months. Knowing the bad road is most of the journey. That is the business track: write once, plan the week, build the model, play the what-ifs — and check every word and every number yourself. See you at camp.