When Numbers Twitch
Churn jumps from 4% to 6%. Two managers see the same spike.
The first slams on the brakes. She freezes hiring, commissions a costly redesign, calls an emergency review. A month later churn slides back to normal - the spike was a client merger, a one-off blip. The fixes cost more than the problem.
The second pauses. She checks whether the data source is reliable, whether this is a pattern or a one-off, whether timing fits seasonal behaviour. With no firm trend, she nudges her concern down, runs a couple of retention plays, and stays on course. Six weeks later churn stabilises - and so does her plan.
Same data, different disciplines. One treated every wobble as signal. The other weighted the evidence whilst waiting for more.
This pattern repeats across organisations. A one-point NPS slip creates an emergency taskforce. A single bug complaint pushes the roadmap sideways. One weak sales month rewrites the quota plan. Each decision feels prudent. Collectively they create a company in permanent twitch.
The cost isn’t just wasted motion. Energy that should compound gets spent reacting to noise.
Good leaders weight evidence whilst they wait. They don’t ignore the wobble - they score it proportionally before deciding how much to update their belief.
The Bayesian mindset helps. Start with yesterday’s belief. When new evidence arrives, update in calibrated steps, not lurches. A single data point moves you a notch. Multiple consistent signals justify a bigger shift. Like a thermostat - small adjustments based on temperature, not slamming the dial because the room felt cold for five minutes.
Most teams lack the discipline because there’s no routine for it. Urgency wins by default.
After your weekly metric reviews, block ten minutes for evidence-scoring before any decisions. Keep a belief log tracking key assumptions and how confidence shifts. Say out loud what you believed last week and how much you’re updating. “I was seven out of ten confident, now I’m six” creates a culture of proportionate updates instead of dramatic pivots.
Reward accurate weighting, not heroic course corrections. The person who correctly identified noise and held steady did better work than the person who launched three initiatives in response to variance.
Look at your last month of decisions. How many were reactions to single data points? How many would you reverse now with more context? If the answer is more than one or two, you’re lurching.
Metrics will always wobble. The test isn’t spotting the wobble - it’s weighting it properly. Pause, score, nudge, move. That’s how you stop an organisation from twitching and help it compound instead.
Related: From Data to Information · Hidden Bottleneck · Reading Guide