แบบจำลองอคติมองโลกในแง่ดีเกินไป

Separate Hope From Data

You are describing the classic inside view trap. Your forecast rests on a viral loop that has not yet demonstrated a single organic referral. That is not a projection; it is a wish wearing a spreadsheet.

Apply reference class forecasting. Do not ask "what will our loop do?" Ask instead: "What did the bottom quartile of SaaS companies with unproven virality achieve in their first 12 months?" Pull 20–30 comparable products, chart their growth rates, and place your number inside that distribution. If 10k sits at the 95th percentile, you have your answer.

Stress-test the loop itself. What is the actual K-factor today? If it is 0.1, you need 100 users to invite 10, and 1 to convert. A loop that has not produced one self-sustaining cycle cannot justify a 50x jump.

Force a devil's advocate session. Have your co-founder argue the bear case: what does the data show if the loop fails entirely? Write down the number you would commit to in front of investors—then cut it by half. That is your realistic floor.

Optimism is fuel, but only when the engine is measured. Trust the distribution, not the story.

Stop trusting the "we learned so much" story.

You are describing the planning fallacy in its purest form: the inside view. Your brain is telling you a narrative about this project's specifics — the new skills, the smoother process — while ignoring the base rate of your own organization's performance.

Build a reference class forecast.

1. List your last 3–5 comparable migrations. Not just software — anything similar in size, team, and complexity.
2. Record the actual schedule and cost overrun for each. Write down the number, not the excuse.
3. Calculate the average overrun. If your team historically runs 40% over, that is your starting forecast — not your optimistic one.

Then apply the correction.

Take your "optimistic" estimate and multiply it by the historical overrun factor. If you think it takes 10 weeks, and history says you run 40% over, forecast 14 weeks. Present that number to stakeholders before the work starts.

Kill the narrative loop.

When you catch yourself saying "we learned so much," translate it into a question: "Did last time's lessons change our actual throughput, or just our confidence?" If you cannot point to a measurable process change — a new tool, a checklist, a different staffing model — assume the lessons are noise.

The past is not a suggestion. It is your most honest estimator.

You’re already ahead by noticing the gap between your forecast and your experience. The fix isn’t pessimism—it’s reference class forecasting. Start by listing the last five ward openings in your hospital, including the ones you didn’t manage. For each, record actual sick-leave days, turnover rate, and time to full staffing. That’s your “outside view.” Now place your plan inside that distribution: where does your smooth ramp-up fall? If it’s at the 10th percentile, you’re in overoptimism territory.

Next, force a pre-mortem. Imagine it’s six months from now and the launch went badly. Write three concrete failure stories—e.g., three nurses out on maternity leave simultaneously, a wave of flu in week two, two resignations after orientation. For each, ask: what buffer would absorb this without chaos? That turns vague worry into specific triggers.

Finally, build buffers as contingency capacity, not pessimism: a float pool agreement, cross-training two staff per unit, a hiring pipeline with pre-vetted candidates. You’re not predicting failure—you’re buying insurance. If the smooth path happens, you lose nothing; if it doesn’t, you’ve already decided what you’ll do. That’s the difference between hope and a plan.