Bayesian updating of a decision: a worked example
The essentials in 30 seconds
- Bayesian updating of a decision multiplies the prior odds by one likelihood ratio per fact.
- In the illustrative Vietnam distributor case, a prior of 0.25 rises to 0.57 after a sales plan and 0.79 after a pilot order.
- A resignation then lowers it to 0.58, below a commit line of 0.65.
- Priors of 0.15 and 0.35 give final values of 0.43 and 0.69.
- The EQOS Asia engine scores 9 structural dimensions with a deterministic core (EQOS Asia method page, as of October 2026).
Bayesian updating of a decision means revising the chance of success each time a new fact arrives. This worked example follows an illustrative distributor deal in Vietnam through 3 pieces of evidence, using counts first and odds second. It extends the guide on how to update a decision with a different case, and it ends with a commit line that holds the budget back. All figures are illustrative, as of October 2026.
On this page
Compare the EQOS Asia offers and request a 30-minute pre-diagnostic.
How do natural frequencies make a Bayesian update of a decision clear?
Picture 1,000 comparable distributor deals in Southeast Asia. Of these, 250 succeed and 750 fail, so the prior chance of success is 0.25 (illustrative figures). Counting cases is easier than juggling percentages, because each step answers a plain question: of the deals that look like this one, how many ended well? Anyone in the room can follow the count on a whiteboard.
The same counts feed every update. Each new piece of evidence splits the surviving deals into those that show it and those that lack it, and the next prior is the share of successes among the deals that show it. Writing the counts beside each step makes the arithmetic visible to a reviewer who prefers cases to formulas.
How does a first piece of evidence update the Bayesian probability of a deal?
The distributor shares a 12-month sales plan. This occurs in 80 percent of deals that succeed and in 20 percent of deals that fail. Of the 250 successes, 200 share a plan. Of the 750 failures, 150 do. The deals that show the plan number 350, and 200 of them succeed, so the posterior is 200 divided by 350, or 0.57.
The evidence moved the probability from 0.25 to 0.57 because the two likelihoods differ by a factor of 4. A plan is a cheap signal for a committed distributor and a rarer one for a distributor who is only exploring. The ratio between the two likelihoods, here 4, is the strength of the evidence and the number worth recording.
How does the odds form speed up Bayesian updating of a decision?
Odds compare successes with failures. A probability of 0.25 equals odds of 1 to 3, or 0.333. The update rule becomes a single multiplication: posterior odds equal prior odds times the likelihood ratio. For the sales plan, 0.333 × 4 gives 1.333, and odds of 1.333 convert back to a probability of 1.333 divided by 2.333, or 0.57, the same answer as the count.
The odds form suits a spreadsheet, because each piece of evidence becomes one column holding its ratio. A ratio above 1 raises the odds, a ratio of exactly 1 leaves them unchanged, and a ratio below 1 lowers them. Teams that log the ratio for every fact see at once which facts carry weight and which only feel important.
How does a pilot order change the probability that a decision succeeds?
A pilot order of 500 units arrives within 60 days. It occurs in 70 percent of successful deals and in 25 percent of failed ones, a ratio of 2.8. Odds rise from 1.333 to 3.733, which converts to a probability of 3.733 divided by 4.733, or 0.79. The count agrees: 140 of the 177.5 expected deals that show both facts succeed.
Commit lines turn the number into an action. Suppose the team commits the second-warehouse budget only when the probability reaches 0.65. After the sales plan, 0.57 sits below the line, so the team waits. After the pilot order, 0.79 clears it, and the budget can be released with the evidence on file for anyone who asks why.
How does bad news lower the probability in a Bayesian update?
Then the distributor’s local sales manager resigns. This occurs in 15 percent of successful deals and in 40 percent of failed ones, a ratio of 0.375. Odds fall from 3.733 to 1.4, and the probability falls from 0.79 to 1.4 divided by 2.4, or 0.58. One fact erased most of the gain from the pilot order.
The new value sits below the 0.65 line, so the team pauses the second tranche and asks for a successor plan. The combined ratio is 4 × 2.8 × 0.375, or 4.2, and the order of the facts leaves it unchanged. The same 3 facts in any sequence return 0.58, which makes the method easy to audit.
How much does the starting prior change an updated decision probability?
Run the same 3 facts from 2 other priors. A prior of 0.15 gives odds of 0.176, which the combined ratio of 4.2 lifts to 0.741, or a probability of 0.43. A prior of 0.35 gives 0.538 and then 2.262, or 0.69. The range from 0.43 to 0.69 straddles the 0.65 line, so the prior matters here.
When the range crosses the commit line, the team needs a better prior before it needs more evidence. Pull the outcomes of the last 20 comparable deals from the sales log and count the successes. A base rate from 20 cases narrows the gap between 0.15 and 0.35 far more cheaply than a fourth round of calls with the distributor.
How does the EQOS Asia decision engine relate to Bayesian updating?
Bayesian updating revises one probability from sequential facts. The EQOS Asia engine starts elsewhere: the Audit pillar measures a system on 9 structural dimensions and returns a φ(S) score between 0 and 1, computed by a deterministic core (EQOS Asia method page, as of October 2026). Both habits turn a feeling into a number that a colleague can recompute.
The prediction pillar projects 3 trajectory scenarios and reports deterministic probabilities, which the method page separates from AI estimates. A reader can try a first score in the simulator on a market entry in ASEAN, and the 30-minute pre-diagnostic, held in Singapore time (SGT), applies the equation to a real decision.
See also in this guide
Frequently asked questions
What is Bayesian updating in simple terms?
Bayesian updating revises the probability of success each time a new fact arrives. Start with a base rate, multiply the odds by the likelihood ratio of the fact, and convert back to a probability. In the illustrative case above, 3 facts move a prior of 0.25 to a final 0.58, and every step stays visible on one sheet.
What is a likelihood ratio?
A likelihood ratio divides how often a fact appears in successful cases by how often it appears in failed cases. A sales plan seen in 80 percent of successes and 20 percent of failures has a ratio of 4. Ratios above 1 raise the odds, and ratios below 1 lower them.
How do I choose a commit line for a decision?
Set the line before the evidence arrives and tie it to the cost of a wrong commitment. A costly step that is hard to reverse deserves a higher line, such as 0.65 in the example, and a cheap, reversible step can accept a lower one. Writing the line down keeps each update honest.
Where can I test a decision with the EQOS Asia engine?
The simulator on the EQOS Asia site gives a first score on 9 dimensions. For a real case, request the free 30-minute pre-diagnostic, scheduled in Singapore time (SGT) and held under NDA (EQOS Asia contact page, as of October 2026). The call points to the dimensions that carry the most risk.
Test your knowledge
1. What is the prior chance of success in the worked example?
- 0.25
- 0.15
- 0.35
Show the answer
0.25.
2. What is the likelihood ratio of the sales plan?
- 2.8
- 4
- 0.375
Show the answer
4.
3. Which probability follows the pilot order?
- 0.57
- 0.79
- 0.58
Show the answer
0.79.
4. Which final probability results from a prior of 0.15?
- 0.58
- 0.69
- 0.43
Show the answer
0.43.
5. What is the commit line in the example?
- 0.50
- 0.80
- 0.65
Show the answer
0.65.
About the author
Stéphane Jambu
President, EQOS Forecast Asia
Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile
Sources: EQOS Asia method page (eqos-asia.com/method.html), simulator page (eqos-asia.com/simulator.html), and contact page (eqos-asia.com/contact.html), consulted on 22 October 2026. Page updated on 22 October 2026.