One founder says you should raise your prices because serious customers will pay for value. Another says to remove every barrier until your product has traction. Both sound convincing. Neither has seen your activation data, support queue or cash balance.
A memorable recommendation still needs to fit your business. Once you have acted, a good month can make the advice look obvious. A bad month can make you forget why you believed it. The original reasoning can disappear beneath a persuasive story about the result.
A decision journal gives you something more useful than a verdict on who was right: a record of what you knew, expected and chose before the outcome arrived. You can use a plain document, a notebook or one row in a spreadsheet. Here is how to build an entry, run a small test and review it without turning every business choice into a research project.
Separate the advice from the decision
Start by saving the original claim and its source. A speaker saying a tactic worked for a particular customer group is making a narrower claim than a short clip telling every founder to try it. Keep the date and relevant context, not just the sentence you liked.
Public investing commentary offers a useful illustration. In public stock-call records, an exact quotation and a link to the relevant second of the source video let readers return to what was actually said. That is a useful habit for evaluating advice even when your own decision has nothing to do with buying shares.
Next, write your interpretation on a separate line. “Raise prices” is advice. “We will test whether a higher entry price improves subscription sales per eligible visitor without creating unacceptable refund or support problems” is a decision you can investigate.
Name the assumption connecting the two. Perhaps your product is underpriced. Perhaps the speaker sells to large companies while you serve first-time buyers. If you cannot explain why their experience applies to your customers, you have found a question to answer before changing anything.
Use a journal entry short enough to finish
A useful entry does not need an essay. Write the following before the experiment starts, while the outcome is still unknown:
Decision and alternatives: What will we change, and what reasonable option are we setting aside?
Our hypothesis: Why might the advice work for this audience and product?
Expected result: What measurable change do we expect, compared with what baseline, and over which window?
Decision rule: What result would justify a rollout, another planned test or stopping?
Limits: What budget, customer experience or operational cost must the test not exceed?
Review plan: Who will assess the result, when, and what will we do if the evidence is inconclusive?
Keep the prediction separate from the business threshold. “I expect a 15% improvement” records a belief. “The change must improve the chosen measure by at least 10% to be worth the operational cost” records a business requirement. Neither number, by itself, establishes that the test can reliably detect such a difference.
Also decide which conditions would stop the experiment early. A broken checkout is a reason to pause; a disappointing first afternoon is not automatically a reason to rewrite the measurement window. Record genuine safety or service problems when they happen.
A pricing experiment with two competing stories
Consider a hypothetical self-service software business. The following figures are invented to illustrate the journal, not results from They Said Buy or another company.
The team tests a $39 monthly entry price against its existing $29 price for new customers. Eligible visitors are randomly assigned to one version and consistently see it. Product features and acquisition sources remain comparable. Purchases are counted within the same seven-day window after each visitor first becomes eligible. The review takes place after the last enrolled visitor has completed that seven-day window.
Before launching, the founders choose initial subscription sales per eligible visitor as their main measure. They set a 10% improvement as the minimum economically useful change, subject to assessing uncertainty. They also agree to review refunds, support demand and later retention before considering a broad rollout. At the scheduled review, each version has reached 500 eligible visitors:
At $29, 20 visitors subscribe: a 4.0% conversion rate and $580 in initial subscription sales, or $1.16 per eligible visitor.
At $39, 16 visitors subscribe: a 3.2% conversion rate and $624 in initial subscription sales, or $1.248 per eligible visitor.
Someone focused on sign-ups can report that conversion fell by 20% relative to the original price, from 4.0% to 3.2%. Someone defending the price increase can report that initial subscription sales per visitor rose by about 7.6%. Both calculations describe the same observations.
The journal keeps the discussion attached to the question chosen beforehand. The team’s recorded decision rule required at least a 10% improvement per visitor. The observed 7.6% increase falls short of that threshold. The calculation is also a snapshot before refunds, fees, service costs and renewal behaviour; it is not profit or lifetime value.
Nor do 36 purchases and a favourable point estimate settle the question of repeatability. The team still needs to assess uncertainty and whether the test ran as planned. A predefined review can end with “inconclusive.” If further observation was part of the design, follow that plan. Otherwise, use what you learned to design the next test instead of extending this one until the numbers look attractive.
Review what happened without rewriting what you expected
At review time, add a new dated entry beneath the original. Leave the original prediction intact. If a factual error needs correcting, preserve what changed and why.
First ask whether you carried out the intended test. A traffic-source change, missing purchase events or a support outage may explain why the result cannot answer the original question. Log those details rather than quietly treating them as background noise.
Then distinguish the observation from its explanation. You observed fewer purchases at the higher price. Random assignment helps estimate the effect of the offered price. It does not explain why each individual declined, establish that either group contains better customers or show how the price will perform next quarter.
A dated record of Tom Nash’s Amazon calls illustrates why preserving the sequence matters: you can revisit the original statement and source instead of judging from a later summary. A subsequent share-price move alone does not reveal someone’s actual trades or establish forecasting skill. For a founder, the equivalent discipline is to keep the initial hypothesis, the action taken and the observed result as distinct entries.
The analogy has limits. A startup can change its product, pricing and customer experience; a stock-call record documents statements about a market it does not control. Borrow the practice of keeping a dated source trail, not an assumption that one business outcome can be scored like a share-price change.
Turn the review into one useful next step
End the entry with an action and the evidence behind it. For the pricing example, that might mean retaining the current price while investigating objections from people who did not buy. Record how those interviews will inform a new hypothesis; do not treat a handful of comments as a replacement for the measurement you already planned.
Over several decisions, look for patterns in your own reasoning. Do you repeatedly assume that a competitor’s audience matches yours? Do acquisition gains turn into support costs that you forgot to track? These are more useful questions than whether a particular adviser deserves permanent trust or permanent dismissal.
Keep the habit proportionate. Reserve a full entry for choices that commit meaningful money, affect customers or are difficult to reverse. A small, reversible interface change may need only a short note. Choose a recurring review slot and give one person responsibility for keeping the record usable.
Before your next pricing, hiring or marketing decision, open a blank page. Write the advice you are considering, the assumption that makes it relevant and the observation that would change your mind. That is enough to begin a journal you can actually learn from.
About the author
Mikhail Nikitin is the founder of They Said Buy, a project documenting public investing statements and linking readers to original video sources.
