Short answer
Because something other than a purchase is being counted as one. The usual culprits are auto-created codeless conversion actions counting page views or button clicks, a duplicate purchase action, or enhanced conversions crediting Google for orders that came from other channels. Each leaves a different fingerprint, and each can be found in an afternoon.
The number that should never happen
Under-counting is the tracking problem everyone expects. Over-counting is the one that does more damage, because it makes the account look profitable. ROAS rises, the owner raises the budget, and the spend follows a number that was never real.
On one Shopify store doing about 40 orders a day, a single conversion action was booking between 11,300 and 32,887 conversions a day, with up to ₹31 lakh of daily value. Over its life it had recorded more than 6 lakh conversions. None of them were orders.
Cause 1: codeless conversion actions
Google Ads can create conversion actions on its own from page URLs and button clicks. They show up with names like a contact page URL or "Add to basket (3)". The one above was counting every visit to the store's contact page.
They are easy to miss because many are excluded from the main Conversions column, so bidding is not directly affected. But the All conversions column the owner reads is fiction, they stop trusting every number, and the real problem underneath stays hidden.
The tell: a daily count that exceeds plausible order volume by 100 times or more. No real purchase action does that.
Cause 2: a page view counted as a purchase
A purchase tag placed in the site-wide header instead of on the order confirmation fires on every page anyone opens. We found one written with an empty value field, which broke it so it fired nowhere. The danger was the fix: anyone correcting that syntax would have turned every page view into a purchase.
The tell: conversions that track sessions rather than orders, rising and falling with traffic.
Cause 3: enhanced conversions matching too loosely
Enhanced conversions match a hashed email or phone against any past Google ad click. So an order that came from a Meta ad, from organic search or a free Shopping listing gets credited to Google if that customer ever clicked a Google ad.
On one store, sending orders matched only by email pushed conversions to 1.22 times the orders that actually carried a Google click, within hours. Restricting to orders with a verified click id brought it back under 1.0. Treat email and phone as extra detail on an order that has a click id, never as the only match.
How to find which one you have
In Google Ads, open Goals, then Conversions, and look at every action, not only the primary ones. Add the All conversions column and a daily date range.
Anything counting far more per day than you have orders is junk. Anything counting roughly your sessions is a misplaced tag. If the purchase action looks sane but total conversions still beat the orders Google actually drove, look at how enhanced conversions are being matched.
Removing a conversion action cannot be undone, and it drops its history. Decide deliberately, and write down what you removed and why.
What "correct" looks like
Google should report close to the orders that came from Google clicks, and not above them. Not close to all orders: on most stores only around half come from a Google click at all. Read the number for a closed day, at least 24 hours after it ends, because Google restates recent days as attribution settles.