Why We Grade Every Marketing Decision Confirmed, High Confidence, or Hypothesis

Most marketing tools present every recommendation with the same tone of certainty. "Increase budget on this adset." "Pause this campaign." No indication of how sure the system actually is, because admitting uncertainty doesn't demo well.

We think that's backwards, and it's a bigger problem than it sounds. A recommendation backed by four weeks of stable, converging data is not the same kind of claim as a recommendation based on three days of noisy signal - but if both show up in a dashboard as a clean, confident action item, a founder or marketer has no way to tell which one deserves scrutiny before they act on it.

So every decision Niti surfaces carries one of three labels: Confirmed, High Confidence, or Hypothesis.

Confirmed means the evidence is stable, converging, and has held up across enough time and enough data that we're comfortable calling it a fact, not an inference. Act on it the way you'd act on something you measured directly.

High Confidence means the direction is very likely right, the evidence is solid, but there's still a reasonable chance a confounding factor is in play - seasonality, a concurrent promotion, a data lag. Worth acting on, worth a second look if the outcome surprises you.

Hypothesis means the pattern is worth testing, the signal is real enough to flag, but it hasn't been confirmed and shouldn't be treated as settled. Treat it as a lead, not a conclusion.

This isn't hedging. It's the difference between a system you can actually trust and one you learn to ignore.

We think about this the same way good judgment already works, and said plainly: trusting your gut and trusting the data were never supposed to be opposites. A founder's instinct is often a compressed version of pattern recognition built from years of decisions the data hasn't caught up to yet. The job of a decision system isn't to override that instinct with a more confident-sounding number. It's to tell you honestly which of your instincts the evidence backs strongly, which it backs weakly, and which it simply hasn't tested yet.

We have the track record to back the label, not just the framing. Looking back at 63 real decisions made for Pokonut, 46% turned out favorable, 38% neutral, 16% unfavorable. That spread is exactly why grading matters - if we'd presented all 63 with the same false certainty, the 16% that didn't work out would have quietly eroded trust in the other 84%. Grading confidence up front means a Hypothesis-labeled call that doesn't pan out doesn't cost you trust in a Confirmed one that did.

The uncomfortable but correct position is that not every marketing decision deserves the same certainty, and pretending otherwise is how teams end up making a confidently wrong call at exactly the moment it costs the most.