Let's think through what happens in real life (from one perspective... that of my wife and I) and see what happens.
A delightful evening ensued (if you need more detail than this, you're on the wrong SE site).
And the next morning, we go to work.
This is important. While there may be a few individuals or couples who are so obsessed that they will start using pregnancy tests LONG before they can even detect conception, many (if not most) people simply go on with their lives and let nature take its course. But, just to cover the bases...
Even when using the very sensitive pregnancy tests on the market today, you still need to give your body time to start producing the pregnancy hormone. ... The best time to take a pregnancy test will always be after you have missed your period. Even the least sensitive pregnancy tests will be more accurate at this point. By the time your period is due, you will have given your body enough time to produce enough hCG to register as a positive result on the pregnancy test. (source)
That's 2-4 weeks. Let's assume the worst-case of 2 weeks. But, what really happened?
Here is where the OP needs to make a story decision about how the pregnancies are failing. Are they miscarriages? Are they actual unferilizable eggs? This is REALLY IMPORTANT because there are two possibilities...
(A) All those excited-to-see-what-happens-next women experience menstration.
(B) Or they don't.
Let's go with (A) first...
Our knee-jerk reaction is that sales of pregnancy tests would drop. But, as @Alexander points out, that's actually not true. In fact, the data identifying a problem will be hidden in the wash of real life.
First of all, all those babies concieved the day before "the event" are testing happily positive. The pregnancy test companies need to watch for a trend to develop, because it won't initially be a step function. But, in this case, it won't develop fast enough.
Because those enchanting evenings really haven't stopped (at least one hopes they haven't stopped...). Which means even if the OP's decision is (A), things just keep on rolling untill the next month. Honestly, how many women successfully conceive after a single encounter? According to this source the odds are only about 20% at best and drop with age very quickly.
So now we're up to 7-8 weeks before a real, detectable trend can be seen. Now the obstetricians mentioned by @RonJohn are starting to see an odd decrease in new-baby patients that can't be explained by a lack of black-outs, a lack of a successful World Series win, or the lack of any of the many other reasons that couples celebrate.
And during all that time women keep buying pregnancy tests in the hopes that this time it'll be positive.
From the perspective of (A), the doctors will identify a detectable trend before the pregnancy test people do because, well... hope springs eternal and the tests are cheaper than a visit with the doc.
At this point: doctors discover the issue first in 7-8 weeks after "the event."
But what happens if the OP chooses (B)?
Here's where @StephenG's comment becomes important. If the OP's answer is (B) then sales TANK because all those women think they're pregnant. They tank like New Coke. They tank in a way that makes Fox News and CNN agree on something. However, even this takes a little time. A week, maybe ten days past menstration.
So, the pregnancy test (PT) people in this scenario could see a detectable trend in 3-4 weeks (remember what the quote above tells you about the body needing time...). But, are the PT people really the first?
Maybe not... because the real problem here is when all those happy women visit their doctors and the doctors see (1) an explosion of new-baby visits and (2) those visits all have really bad news... no pregnancy.
This would be a very-fast-to-detect trend, a lot of new pregnancies that aren't really pregnancies (be they miscarriages or any other unannounced reason). The statistics would be blaring.
Which means while the PT people could be the first to know, it's still unlikely because they're behind the prenatal care curve. I still believe the doctors would know first.
But the data is still a bit hidden. Remember all those pregnancies that happened in the days before "the event." That will muddy the water for a while. Maybe 1-2 weeks before the doctors really start realizing that more women are coming in with false pregnancy problems than are with successful pregnancy issues.
You'd think it would be hard to dismiss the "You know, Nurse Chapel, that's the fifth woman this week with a false-positive. That's a bit weird, dontchathink?" reaction, but remember that 20% and less chance to become pregnant at all. Unless "the event" also causes women to conceive at 100%... but this answer is long enough already.
So, in this case, 4-5 weeks.
The main issue here is that people don't stop living their lives after an attempt to conceive, and nobody knows they're pregnant (or not) at the moment of theoretical conception.
Many if not most women won't even bother with testing before their next menstral period suggests a reason. But it doesn't matter anyway because...
Women don't conceive with every encounter. That's life. There's about a 20% chance. That means a guaranteed "there's no doubt anymore" realization is 5 months max. But, for the most part, the happy couple is momentarily disappointed and then renew their dedication to the cause.
Which means pregnancy tests are being bought long after "the event" that stopped all conception.
The data will be clouded by all the conceptions that occured during the days (even weeks) before "the event," leading to new-baby and ongoing-baby visits with doctors for quite a while.
Doctors would eventually realize either (a) the number of new-baby visits have dropped a long, long way (path (A) above) or (b) the number of new-baby-false-pregnancy visits has jumped (path (B) above). But, either way, thanks to the clouded data, this might not actually be detectable for a while.
If the OP chooses to simply stop fertilization (path (A) above), then it would likely take 7-8 weeks for doctors to realize something was wrong and actively start investigating it because not conceiving children is actually 5X more likely than conceiving them. AKA, it's normal and wouldn't be detected as quickly.
If the OP chooses to permit fertilization, but the pregnancies are somehow terminated (path (B) above), then it would likely take only 4-5 weeks to realize something was wrong because consistent false-positives are not normal and would be detected more quickly.
DISCLAIMER: This answer is from the perspective of a happy couple living in the United States. There is a massive amount of variability in fertility, health care, etc. around the world. Absolutely massive. I'm (somewhat arrogantly) assuming that the U.S. has the capacity to detect such a problem first due to the combination of high tech levels and large population bases. Addressing this question beyond this perspective might (and likely does) make the question too broad.