Short answer
Usually because Meta's checkout count is not what it sounds like. It credits any checkout started by someone who clicked the ad within the attribution window, on any product, and counts checkout page loads where no details were entered. Check Shopify's abandoned checkouts and orders for the advertised product before you touch the checkout.
The case that looked like a broken checkout
A pro audio store ran a Meta campaign for one product. Over six days Meta reported 13 initiated checkouts and 0 purchases. Every sign pointed at a checkout fault, and pausing the campaign looked sensible.
Shopify said otherwise. There were 16 abandoned checkouts in that window, each with a real customer name, and not one contained the advertised product. The product did sell during the same week: a ₹62,990 order, paid normally, which came from a free Google Shopping listing.
What Meta's checkout number actually counts
With 7-day click attribution, Meta credits a checkout to the ad if the person clicked the ad at any point in the previous week, whatever they ended up buying. An ad clicker who later started a checkout for something else counts.
It also counts checkout page loads. Shopify only records an abandoned checkout once the shopper has entered contact details, so someone who opened checkout and left immediately exists for Meta and not for Shopify.
Ad credit is not site health
The ad-level purchase and checkout fields report only events attributed to those ads. An empty value means the ads got no credit. It says nothing about whether the pixel fires or the site converts.
On another store the ad fields were empty across 228 landing page views and it looked like a dead pixel. The pixel's own event log for the same week showed about 1,000 page views, 219 product views, 22 add to carts, 15 checkouts and a purchase. Nothing was broken. The ads were bringing traffic that did not shop while other channels converted, and that is a targeting and creative question, not a tracking one.
The check that settles it
Pull Shopify's abandoned checkouts and orders for the campaign dates and filter by the advertised product. If the product never appears in a checkout, the checkout is not the problem.
Then look at the pixel's event totals in Events Manager, not the ads report. Healthy event volume with no ad credit means the pixel works and the audience does not buy.
Only if shoppers reach checkout with the product and then fail should you test the checkout itself.
When the numbers genuinely disagree
There are real faults too. The most common is purchases counted twice because the pixel and the Conversions API send the same order without a shared event id, so Meta cannot deduplicate them. On stores with a third-party checkout the opposite happens: the pixel never sees the purchase page at all, and only a server-side purchase event fixes it.