Forecasting Readout

Purpose

Turn a metric projection into a decision-ready forecast with explicit trend basis, assumptions, and uncertainty bounds - not just a point estimate.

When to use

Use this skill when:

• asked "what will revenue/users/orders be next quarter/month/year?"
• projecting a metric for planning, budgeting, or target-setting
• presenting a forecast to a stakeholder who will use it to make a decision
• evaluating whether a current trend leads to hitting or missing a target

When not to use

Do not use this skill when:

• there is insufficient history to support any projection (fewer than 3-4 comparable periods)
• the metric is driven primarily by an upcoming event or decision with no historical analog
• the question is why a metric changed, not where it is going (use root-cause-analysis)

Required thinking discipline

• Never produce a point estimate alone. A single number without a range implies false precision and misleads decision-makers.
• State the basis for the projection explicitly - what pattern does the forecast extrapolate?
• Distinguish extrapolation from causally grounded projection. Trend extrapolation assumes "what has been true will continue." That assumption needs to be named.
• Separate trend from seasonality. Projecting November revenue in July requires handling the seasonal pattern explicitly.
• A forecast is a decision input, not a commitment. State what would need to change for the forecast to be wrong.
• Evidence constraint: Every conclusion must cite specific data — a number, a rate, a segment, or a timeframe. Do not speculate without evidential basis. If data is insufficient, state what is missing rather than asserting an unsupported inference.

Workflow

1. Define the metric, entity, and forecast horizon (e.g., net revenue, all markets, next 90 days).
2. Characterize the historical trend basis:
   • Flat (no directional trend)
   • Linear growth or decline
   • Accelerating or decelerating growth
   • Mean-reverting or cyclical
   • Event-driven (spikes around campaigns, holidays, product launches)
3. Identify and separate seasonality: does the metric have known weekly, monthly, or annual periodic patterns? State how seasonality is handled (carried forward, averaged, ignored).
4. List key assumptions the forecast depends on:
   • No major product, pricing, or acquisition strategy changes
   • External environment remains consistent
   • Seasonality pattern from prior years applies
   • Any specific operational assumptions (new market launch, campaign planned)
5. Produce the point estimate plus a scenario range:
   • Base case: continuation of recent trend with normal seasonality
   • Upside case: trend continues at the favorable end of recent variance
   • Downside case: trend continues at the unfavorable end, or a known risk materializes
6. State conditions that would invalidate the forecast - what event or change would require revisiting the projection?

Output format

• Metric and horizon
• Trend basis (with time period used as foundation)
• Seasonality handling
• Key assumptions
• Point estimate (base case)
• Scenario range (upside / base / downside)
• Invalidation conditions
• Recommended review trigger (e.g., "revisit if weekly actuals deviate more than 10% from base case for two consecutive weeks")

Good example

Metric: weekly new paid subscriptions. Horizon: next 12 weeks.

Trend basis: linear growth of approximately +2.8% week-over-week over the past 16 weeks, excluding a one-week spike from a November promotion. Seasonality: end-of-year slowdown observed in prior two years (weeks 51-52 run approximately 20% below trend). Applied to base case. Key assumptions: no pricing change, current acquisition spend maintained, no major product change. Base case: 4,200 new subscriptions in week 12. Upside: 4,900 (if acquisition efficiency improves 15% as recently observed in test markets). Downside: 3,400 (if the year-end slowdown is stronger than historical average or if acquisition efficiency reverts). Invalidation conditions: a pricing change, a shift in paid acquisition budget greater than 20%, or a product change affecting the conversion funnel. Review trigger: revisit if actuals fall outside the upside/downside range for two consecutive weeks.

Bad example

Revenue will be $2.4M next quarter.

Why this is bad:

• no range or uncertainty - implies false precision
• no trend basis stated - where does $2.4M come from?
• no assumptions - what has to be true for this to hold?
• no invalidation conditions - under what circumstances is it wrong?
• a stakeholder will hold this number as a commitment rather than an estimate

Practical notes

• The further out the horizon, the wider the uncertainty range should be. A 90-day forecast should have a wider range than a 30-day forecast - if it does not, the uncertainty is being hidden.
• If the metric has high week-to-week variance, communicate that variance explicitly - a "flat" trend with +/-30% weekly swings is very different from a flat trend with +/-5% swings.
• When a forecast is used for planning, recommend what trigger would cause the plan to need revision - this converts a static number into a living decision tool.
• Do not project through a known structural break (pricing change, market entry, acquisition) without stating that the break makes historical extrapolation unreliable.

Optional variants

• Operational forecasting (capacity planning): focus on volume and throughput rather than revenue; add resource requirement translation (headcount, infrastructure).
• Financial forecasting (budgeting): be explicit about revenue recognition timing vs. cash timing; add margin assumptions if projecting profit.
• Growth forecasting: apply seasonality adjustment explicitly; distinguish organic growth from paid acquisition-driven growth in the trend basis.