The Big Ride We Are On

When I was seventeen I was offered an Air Force ROTC scholarship. Not once. They came back for me more than once. It was good at any university with an Air Force ROTC program, and I could have pointed it at any of them.

I pointed it at Purdue, and I pointed it at computer science, because Purdue is where the astronauts came from. Believe it or not, the Air Force wanted mathematicians more than computer scientists. Mathematics was the option above computer science, and I chose it later, though I kept a foot in computer science. I had been programming since I was fourteen, writing a DOS program on a TRS-80 to keep baseball standings, and then on an Apple II in the high school computer lab. I was also in a small computer class that met off campus. I took Advanced Placement classes too, so I arrived at Purdue with some college credit already. I came away from Purdue with a B.S. in mathematics and an A.S. in computer science.

There has always been a computer and math thread running through it: the TRS-80, the high school class, Purdue, a roommate-conflict study I co-authored with Bret Fuller from the career department at the University of Missouri-Columbia, published in the Journal of College Student Development, the statistics on the first job sites at Isaacson, Miller, and every year since at ExecSearches, where the work has always come down to statistics and search.

Bret's name came first on that paper, and for good reason: he was the smarter of the two of us, and the better writer. The study compared the Myers-Briggs Type Indicator (MBTI) with a roommate compatibility questionnaire. We found interesting results, but I don't find the MBTI helpful for recruiting. Bret was on another level, like many of the people I worked with at Isaacson, Miller: John Isaacson especially, and Arnie Miller, Lisa Savereid, Laura Gassner Otting, Karen Alphonse, Alan Wichlei, David Haley and Jack Gorman. They are some of the most intelligent people I have ever met. I know when I am outclassed.

I wanted to fly. That was the plan, in full, and I want to be clear that it was a plan and not a daydream. At seventeen, I knew what I wanted. I knew exactly what the path was, I knew what it required, and I went and got the part of it I could get.

I did not get the ride. I got the mathematics.

It took me about forty years to work out that I was going to end up on the rocket anyway, just not the one I was looking at. I am on it right now. This is it. Something we built from nothing is accelerating underneath us, and the reason I can tell that it is accelerating, rather than merely getting bigger, is the degree I took instead of the flight.

So here is what the curve is doing, and here is how you read it.

So far in September

September is likely on pace to pass the last three months combined in clicks.

I am using clicks here, not impressions. These are the clicks Google Search Console reports for our search results, not a count of unique visitors. Clicks are the metric I am following in this comparison.

Why a job board grows this way

Job boards grow differently from most websites, and the field we serve is still young. Being early and being thorough counts for a lot in a new category.

What I am not claiming

I am not claiming this rate continues. The curve may flatten eventually. We do not know when. Not yet, anyway: we are still accelerating. I am not claiming a cause with any confidence either, because the site changed in several ways at once over this period and separating them after the fact is math.

What I am claiming is narrow and checkable: September is likely on pace to pass the last three months combined in clicks.

A note on where we stop counting

Our published historical figures stop at the last settled reporting date. We do not report the most recent two or three days, because those days are still filling in and a number that is still moving is not a number yet. Cutting the series where the data actually ends is the fair way to publish, and it is the more accurate one.

That is a rule about what we publish. It is not a rule about what we look at.

Internally we do the other thing as well, which is to take the series we do have and work out what it implies. Divide each week by the week before it and you get a growth rate. Take the log of that and you can see whether the rate itself is moving. Do that across enough weeks and you can tell the difference between something that is growing steadily and something that is getting faster, and you can put rough bounds on where it lands if it holds. We use those estimates internally to decide what to build next. When we discuss an outlook publicly, including the September comparison, we identify it as a projection. We never present a projected result as a measured one.

So there are two sets of numbers here and they do different jobs. The published ones are conservative and stop early on purpose. The working ones run ahead of the data and are treated as what they are. That is fine, as long as you are honest about which one you are showing somebody.

How the extrapolation actually works

Here is the method. You have seven points. You divide each one by the one before it. Now you have six growth rates. You look at whether those rates are drifting up, drifting down, or sitting still. If they are sitting still you are compounding at a fixed rate and you can say where you land in eight weeks with a multiplication. If they are drifting up, the curve is steepening and your eight-week number is higher than the simple multiplication says. If they are drifting down, it is lower.

That is it. That is the entire technique. Seven points, six ratios, look at the trend in the ratios.

The ratios describe what has happened across comparable time intervals. Projecting eight weeks ahead requires an additional assumption about what those ratios do next. That makes the projection a conditional estimate, not a measurement or a promise.

It is math. For some it comes naturally, others see "data sets visually," and for others it takes a professor and a book, or a prayer. And that's OK. We all have our talents.

I said at the top that I went to Purdue for the flight and came away with the mathematics. This is where that trade pays. Reading a curve correctly is not a gift and it is not intuition. It is a technique, and like any technique it can be learned.

Where this goes

The next thing worth measuring is not whether the number keeps climbing. It is whether the people arriving are the right people. A click from someone who wanted a compliance job in Chicago is worth a great deal more than a click from someone who wanted a definition. That is a harder thing to count, and it is what I am working on next. Enter generative AI stats, and what the answer engine tells us.

None of that comes from watching our visitors. We don't track the people who use this board. The site sets one cookie, a Cloudflare security cookie it needs to run, and nothing else. What we know about who arrives comes from the searches Google reports, never from following anyone around.