Why "yesterday's numbers" are always wrong, and keep climbing for weeks

Here's the part that trips almost everyone up the first time they watch a dashboard closely: when someone clicks your ad on Monday and buys on Thursday, Google and Meta both file that purchase under Monday, not Thursday. The conversion is stamped with the date of the click that started the journey, not the date the money actually changed hands.

That's not a bug or a reporting delay in the usual sense. It's how click-based attribution is designed to work, because the platform's job is to tell you which click deserves credit, and the click happened on Monday. But it means Monday's numbers can't be finalized on Monday, or even by Tuesday. They can only be finalized once every person who clicked on Monday has finished either converting or not converting, and depending on the platform and the account's settings, that window can run anywhere from same-day out to around 90 days.

So a day's reported conversion count isn't a snapshot, it's a running tally that keeps getting new entries added to it for weeks after the day itself has passed. Every time someone who clicked on that day finally converts, whether it's 2 days later or 60, that historical day's number ticks up. If you've seen the term "conversion lag" or "IBNR" used for this, that's the short name for the phenomenon, borrowed from insurance (Incurred But Not Reported): a claim happened, but hasn't been filed yet.

Picture a campaign that's performing identically, week over week, with no real change in results. Now look at it through a "last 7 days" report. The oldest days in that 7-day window (day 7, day 6) have had a full week or more to accumulate their delayed conversions. The newest day (day 1, yesterday) has had almost none. It hasn't had time.

That means a flat, unchanging campaign will still show a declining trend line inside any short recent window, purely as an artifact of how much time each day has had to "fill in." It isn't declining. It's just younger, and younger days are always undercounted relative to older ones.

This is what makes the classic "last 7 days vs. the 7 days before that" comparison structurally misleading for anyone new to reading ad data. That comparison is biased against the more recent period by construction, not by anything that actually happened in the account. Even a campaign that's genuinely improving can appear to be falling off a cliff, because the newer period simply hasn't finished counting yet.

The fix isn't a formula, it's a habit: don't judge a campaign's health from data younger than about a week old. A week gives most of the fast conversions (same-day through a few days out) time to land, so the picture stops moving as violently day to day.

Full maturity takes longer. Expect the most recent week's numbers to keep revising upward for roughly 3 to 4 weeks before they're close to their final, settled value. This doesn't mean you're blind for a month. It means: when you're looking at day 1 through day 7, treat those numbers as a floor, not a final answer, and reserve confident conclusions for windows that have had a few weeks to breathe. If you need a number today, look further back.

The single most common mistake this creates: someone glances at a keyword, ad set, or campaign, sees "0 conversions this week," and pauses or cuts the budget in a panic, when the true answer is simply that this week's conversions haven't finished being counted yet.

Imagine a keyword that reliably drives one or two purchases a week, spread out with a lag of anywhere from a same-day sale to three weeks later. Look at it on a Wednesday and you'll often see zero conversions for the current week, not because the keyword stopped working, but because none of this week's clicks have had time to convert yet. Pause it here and you're cutting something that was never actually broken.

The way to tell a real zero from a not-yet-counted zero is to check volume against maturity, not just the conversion count in isolation. Two questions: is there meaningful click or spend volume behind this zero (a true dead campaign usually shows weak traffic too, not just weak conversions), and how old is the window you're looking at. A zero on a week-old window with healthy traffic is very likely just immature data. A zero on a month-or-more-old window with healthy traffic, after conversions have had real time to land, is a genuine signal something is wrong.

Everything above describes click-attributed conversions, the standard model both Google Ads and Meta use by default, where credit flows back to the click that started the journey. Not every measurement setup behaves identically. Attribution windows are configurable (some accounts count conversions out to 30 days, others to 90), and how much lag you'll actually see depends on your product's typical buying cycle, your specific conversion actions, and how your account's attribution settings are configured. The direction of the effect (recent days always undercounted, older days always more complete) holds everywhere click attribution is used. The exact size of the effect is specific to your account, so it's worth knowing your own numbers rather than assuming a generic timeline.