Reacting before reading
May 14, 2026 • 4 min read
Monday morning: conversion is down fourteen percent. By Tuesday the head of marketing has three calendar invites and the product team has been pulled in. By Wednesday a homepage redesign is on the table. By Thursday a small task force is meeting to talk about whether the funnel needs to be rebuilt. By Friday nobody is sleeping.
But was any of it justified?
A dip on a chart is not automatically a thing to act on. Most of what looks like a signal is one of a small number of explanations dressed up to look like a signal. Before any of the levers in the room get pulled, the explanations below are worth ruling out.
Seasonal. Most metrics worth tracking have a season. An ice cream shop comparing November to September is going to find a hole. Comparing November to last November is going to find nothing wrong. A lot of dips are the chart doing what it always does at this time of year.
Noise. Real numbers wobble. A metric that moves five percent in either direction every week is not telling a story when it moves seven percent this week. It is being itself. One week of wobble does not move a trend line.
Anomaly. One day went sideways. Tracking dropped for a few hours, a payment provider blinked, a single huge customer churned for reasons unrelated to anything anyone can change. The chart bounces back the next day and the dip stays in the data forever.
Market-wide. Your line dropped twenty percent. So did everyone else's. The story is not about the company, it is about the category, and the lever the room is about to pull would not have moved either.
Pulled forward. A promo, a sale, or a deadline made people buy now instead of later. The week after Black Friday looks weak because the customers who would have bought this week already bought last week. Reacting to it usually means cutting spend on a channel that is working fine.
Tracking changed. The number moved because the way it is measured moved. An ad platform shortened its attribution window overnight. The drop is in the measurement, not the customer, and the spend gets cut anyway.
Missing data. The right edge of the chart is incomplete. The last few days are still filling in. The conversions data on our Meta ads for the playbook always looks terrible for the first 48 hours because they attribute sales late.
Interpretation is the part of the job that is hardest to teach. The rest of analysis is mechanical and most of it can be picked up in a week. This part takes years, and you mostly learn it by being wrong a few times and remembering it. Pause on the dip before reaching for a lever.
