Business owners often face two opposing risks.
Decide too early, and the business may make an expensive mistake with insufficient evidence.
Wait for complete confidence, and a competitor may move first, costs may increase or the opportunity may disappear.
The objective of Data-driven Decision Making is therefore not to eliminate all Uncertainty. In most business situations, that is unrealistic.
The objective is to reduce the Uncertainty that matters to a level appropriate for the Risk of the Decision, and then act while the decision still has value.
Value of Information, or VOI, provides a useful way to think about this trade-off. Additional information is valuable when it can improve a decision, while the cost of collecting that information and the Cost of Delay also need to be considered.

“Enough data” does not mean knowing everything. It means knowing enough to choose a reasonable next action.
More information can reduce Uncertainty, but waiting for it also has a Cost.
Before gathering more data, ask:
1. How costly would a wrong decision be?
2. How easily can the decision be reversed or adjusted?
3. Which remaining Uncertainty could actually change the choice?
4. Is new information likely to change what we do?
5. Is the value of waiting for that information greater than the Research Cost and Cost of Delay?
The principle behind Value of Information is that additional information has value when it can improve the decision—not simply because it increases the amount of data available.

There is no universal percentage of information required before a decision

Some decisions can reasonably be made using a few critical data points.
Others require Market Research, Financial Analysis or an Experiment.
The difference is not simply the amount of data.
It is the nature of the Decision.
Compare:
Decision A
Change the copy on a Landing Page, with the ability to switch back tomorrow.


Decision B
Sign a five-year lease and invest several million baht in a new location.
These decisions should not require the same Evidence Standard.
Decision A is easy to Test, Learn from and Reverse.
Decision B involves substantial Commitment and is harder to correct later.
The useful question is therefore not: “Do we have enough data?”
It is: “Is the evidence we have appropriate for the risk of this decision?”

Start with the Decision, not the Data

Before opening another Dashboard or commissioning more Research, define the Decision.
Too broad: “We want to understand the market.”
More useful: “We need to decide this month whether to open Location A.”
Once the Decision is clear, ask:

• What are the Options?
• What needs to be true for this Option to work?
• Which uncertainties matter?
• Which evidence would help distinguish between the Options?
• Would different results change the Decision?

This follows a core principle of Decision Analysis: first determine what should be done based on the best evidence currently available, then ask whether the Remaining Uncertainty is valuable enough to justify gathering more evidence.

Do not reduce every Uncertainty. Find the ones that can change the Decision.

Imagine a restaurant is considering a new location.
The team is uncertain about:

• Foot Traffic
• Rent
• Competitor Count
• Average Spend
• Conversion Rate
• Repeat Purchase
• Parking
• Delivery Demand
• Interior Design Preferences

All of these may be interesting.
They are not equally important to the Decision.
If the Financial Model shows that the decision changes from GO to STOP when Rent exceeds one level or Conversion Rate falls below a certain threshold, those variables are Critical Uncertainties.
Interior Design Preference may not yet matter to the Go / No-go Decision.
Value of Information approaches prioritize information that reduces uncertainty capable of affecting the preferred Option. Research that is unlikely to change the decision has less Decision Value.

Additional information is valuable when it can change what you do

Before commissioning more Research, ask:
“If the result is A, what will we do?”
“If the result is B, what will we do?”
Suppose the answer is: “We will do exactly the same thing either way.”
The information may not be necessary for this Decision.
For example, imagine a team has already committed to launching a Product regardless of whether Brand Awareness is measured at 20% or 40%.
If the Study will not change the Timing, Budget, Target, Positioning or Launch Decision, it may have low Decision Value at that moment.
The National Academies describes the VOI principle in similar terms: research that is unlikely to change a decision has little ex-ante value for that decision, and if no possible result could alter the choice, the Expected Value of Information is zero within the framework.

Waiting has a price: Cost of Delay

Businesses often calculate Research Cost but overlook the Cost of Waiting.
Suppose additional Research takes eight weeks.
During that period:

• A competitor may launch first
• Peak Season may pass
• Media Costs may rise
• Supplier Capacity may disappear
• Customer Needs may change
• The team may lose Momentum
• Revenue may be delayed

Value of Information research shows that additional evidence should not be evaluated only by how much Uncertainty it reduces. The Cost of Data Collection and the consequences of Delayed Decision Making also matter.
Therefore: More Information ≠ Always a Better Decision
More accurate information arriving after the opportunity has disappeared may have much less Business Value.

Reversible decisions should have a different Evidence Standard from hard-to-reverse decisions

Reversibility is a practical way for SMEs to decide how much evidence is enough.

Reversible Decision

Examples:

• Change an Ad Creative
• Test a Promotion
• Change Landing Page Copy
• Introduce a temporary Product Bundle
• Pilot a Weekend Service

If a mistake can be corrected quickly and cheaply, a Small Test may be more valuable than extended Research.
The pattern becomes: Decide → Test → Measure → Adjust

Hard-to-Reverse Decision

Examples:

• Build a factory
• Sign a long-term Lease
• Invest in a major system
• Purchase substantial Inventory
• Conduct a major Brand Repositioning
• Enter a market with high Fixed Costs

These decisions deserve stronger evidence because the Cost of Error and Cost of Modification are higher.
The National Academies notes that when decisions are difficult to modify, decision-makers should consider the value of additional information alongside Research Costs, Delay Costs and the cost of changing the decision later.

Let Decision Risk determine how much you need to know

Consider two simple dimensions.
Impact if Wrong
How much money, time, customer trust or reputation is at risk?
Reversibility
How easily can the decision be changed?
A practical framework is: Low Impact + Easy to Reverse
Decide quickly and run a Small Test. High Impact + Easy to Reverse
Run a limited Pilot and monitor closely. Low Impact + Hard to Reverse
Validate the key assumptions before committing. High Impact + Hard to Reverse
Require stronger evidence through Research, Scenario Analysis, Financial Modeling or Experiments where appropriate.
This is not a statistical formula.
It is a Decision Framework for matching the strength of evidence to the consequences of being wrong.

Do not ask only “How confident are we?” Ask “What happens if we are wrong?”

Imagine two decisions both have an estimated 70% chance of being right.
Decision A: A wrong decision costs 10,000 baht and can be corrected next week.
Decision B: A wrong decision costs 5 million baht and creates a three-year commitment.
The Confidence Level may be the same.
The Evidence Requirement should not be.
Decision Quality therefore depends not only on the Probability of Error but also on its Consequence.
This is why management heuristics such as: “Make the decision when you have 70% of the information”
may be useful in some settings but should not become universal rules.

Use Sensitivity Analysis to find assumptions that can flip the Decision

An SME does not always need a sophisticated Decision Model.
A simple Spreadsheet can help.
Imagine a new location with:
Expected Revenue = 600,000 baht per month
Gross Margin = 55%
Fixed Cost = 280,000 baht
Revenue remains uncertain.
Test different assumptions:
500,000 baht → the investment may not work
600,000 baht → close to the threshold
750,000 baht → more attractive
If the Decision changes significantly when Revenue changes but barely changes with a small variation in Marketing Cost, Revenue Potential deserves more Research attention.
Sensitivity Analysis helps answer: “Which Uncertainty should receive the next research baht?”
VOI and Bayesian Decision Analysis use a related principle: identify the parameters or assumptions contributing most to Decision Uncertainty and prioritize additional information accordingly.

Do not collect more data simply because an estimate is still imprecise

Suppose the current Market Size estimate is 80–100 million baht.
The business is considering a 1-million-baht investment.
If the decision remains GO whether the market is 80 million or 100 million baht, spending another month narrowing the estimate to 88–92 million may not change the decision.
Now consider:
80 million baht → STOP
100 million baht → GO
Reducing the uncertainty suddenly has much greater Decision Value.
Precision matters when it helps distinguish between Actions.
Not simply because it makes the Report look more precise.

Before requesting more Research, ask how the next piece of information could change the Decision

Use four questions:

1. What do we still not know?
2. Can this Uncertainty change the Decision?
3. What is the fastest and least costly way to reduce it?
4. What would we do differently after receiving the answer?

If question four has no clear answer, collecting that information may not yet be necessary.
This helps distinguish information that is interesting from information that is decision-relevant.

More information does not always require a large Survey

Once a Critical Uncertainty is identified, choose the smallest useful method that can address it.
If customers may not understand the Product: Run Customer Interviews or a small Concept Test.
If the Landing Page may not convert: Run an A/B Test or Traffic Test.
If Willingness to Pay is uncertain: Use an appropriate Pricing Test, Pre-order Test or Behavioral Test.
If local Demand is uncertain: Start with Secondary Data and add Local Observation or a Small Market Test.
If Unit Economics are uncertain: Build a Financial Model and run Sensitivity Analysis first.
The principle is: Uncertainty → Minimum Useful Evidence
not: Uncertainty → Full Research Project every time

Small Experiments can turn “we must know before acting” into “we can learn while acting”

Imagine an SME is uncertain whether a new Subscription Package will sell.
Option A: Research for three months before launch.
Option B: Pilot the package with 100 customers for four weeks using predefined Success Metrics and a Stop Rule.
If the Pilot has low Cost and is easy to stop, Option B may provide evidence closer to Actual Behavior than Stated Intent while reducing Cost of Delay.
This is the value of a Reversible Experiment.
Not every Uncertainty must be resolved before Action.
Some can be resolved through Action designed to generate learning.

Define the Decision Rule before collecting more data

One reason Research never seems finished is the absence of a Stop Rule.
Every new answer creates another question. Before starting Research, define:
What result means GO?
What result means REVISE?
What result means STOP?
Example:
GO Pilot Conversion Rate ≥ 8% and Gross Margin clears the required threshold
REVISE Conversion is 5–7.9%, but Customer Feedback identifies a fixable Barrier
STOP Conversion < 5% with no Segment showing Strong Demand
These figures are illustrations, not universal benchmarks.
The value of a Decision Rule is that the team defines in advance what evidence is “enough,” reducing the temptation to move the goalposts after seeing results.

Watch for Analysis Paralysis disguised as Data-driven Decision Making

Requesting more data does not automatically improve Decision Quality.
Warning signs include:

• More Dashboards without a new Decision
• Multiple Research rounds without Thresholds
• Every answer creates another question without a Stop Rule
• Larger Samples are requested even though new results are unlikely to change the Action
• The team waits for more Precision while the Opportunity declines
• “We need more data” is repeated without identifying which information is missing

These can indicate Analysis Paralysis rather than Evidence Discipline.
The answer is not to stop using data.
It is to return to the question: “Which Uncertainty is actually preventing the Decision?”

Example: Open the new location now or collect more evidence?

Imagine an SME considering a new branch.
Current evidence:

• Location Traffic looks promising
• Rent is high but still within the Financial Threshold
• Three competitors operate nearby
• Customer Interviews are reasonably positive
• Demand Estimate remains wide
• The Lease requires a five-year commitment

Because the Lease is hard to reverse and Fixed Costs are high, Demand Uncertainty has high Decision Value.
The business might gather more targeted evidence:

• Foot Traffic by Daypart
• Competitor Traffic
• Small Pop-up Test
• Delivery Demand in the area
• Customer Origin from nearby branches

It does not need to research everything.
If Interior Preference cannot change the Go / No-go Decision, that question can wait.
Research Budget is concentrated on the Uncertainty that matters most.

Another example: Change the website Promotion or wait for more Research?

Suppose the current Promotion has weak Conversion.
The team has an Alternative Offer.
The change can be rolled back within hours, and there is enough Traffic to Test it.
This Decision has:
Low Reversal Cost
Low Implementation Cost
High Learning Potential
A Small Experiment may be more useful than spending a month interviewing customers before acting.
The Decision itself can generate evidence.

Seven questions before asking for more data

1. What exactly are we deciding, and by when?
2. How costly would a wrong decision be?
3. Is the Decision easy to reverse or adjust?
4. Which Uncertainty could change the Decision?
5. What direction does the current evidence support?
6. Which additional information is most likely to change the Action?
7. Is Research Cost + Cost of Delay lower than the expected benefit of the new information?

Answering these questions makes “Do we have enough data?” much easier to resolve.

The takeaway: Stop collecting when the next piece of information is unlikely to change the Decision enough to justify waiting

Data-driven Decision Making does not mean waiting until information is complete.
It means matching Evidence to Risk, Uncertainty and Time.
For Reversible, low-cost decisions: Decide → Test → Learn → Adjust
For High-stakes, hard-to-reverse decisions: Research → Challenge Assumptions → Reduce Critical Uncertainty → Decide
Value of Information captures the underlying principle: Additional Information is worthwhile when the Expected Benefit from making a better decision exceeds the cost of obtaining that information, including the Cost of Delay.
So the next time a team says: “Let’s get a little more data first.”
Do not ask only: “What else do we need to know?”
Also ask: “If we had that information, could it actually change our decision?”
If the answer is “probably not,” the evidence you already have may be enough for the next Action.