Jim VandeHei shortened Axios's planning horizon from six months to seven weeks

Subtraction is the only work whose payback lands inside seven weeks. A systems migration or a hiring plan against next year's budget pays back in quarters, and a workstream you end pays back that week. A leader who says he cannot see past six or seven weeks in his own business ^[6] has, by that admission, narrowed the set of things he can honestly approve.

The number underneath the shortened horizon is a distribution number, and it comes from VandeHei himself ^[7]. Take the ends of the range he gave: 5 over 40 leaves an eighth of the old share, 10 over 30 leaves a third, so the search referral share fell by somewhere between 67 and 87.5 percent ^[9]. He said "that happened almost overnight" ^[8].

A drop like that measures a channel someone else owned and then repriced. The abridged transcript does not include Axios subscription or retention figures ^[15], so there is no way to tell from it how much of the lost traffic was worth anything past the visit. VandeHei described those visitors as people "who were customers or were coming to and tasting our content" ^[8].

The system has a name and a book behind it ^[2][4]. The practice described in the interview is narrower and more portable: refuse a planning horizon you cannot defend ^[6], and put deletion on a calendar invite by name ^[3]. He is also specific about why AI stalls inside a company that already exists. Introducing any technology is hard, he said, because "even if it's the coolest technology ever, you have to hook it up to all your clunky systems that you built over the last decade, and that's necessarily messy" ^[11]. Asked if it was harder than he expected, he said: "This is way harder than I thought, yeah" ^[12].

In the same conversation VandeHei said two incompatible things about the market. He said companies that use AI "persistently and aggressively live in a different world right now" than those that don't ^[13], and he also said most people tell him they don't really use it, or that it is "kind of a glorified search engine" ^[14].

For anyone applying this next week, two tests per active workstream will sort most of them: whether the payback lands inside a horizon you can actually defend, and whether the work runs through a channel you control. I would cut the long-payback items on borrowed channels first, and the cost of that rule sits in the same sentence: a seven-week ceiling also kills the nine-month project that would have built distribution you own. Axios lost at least two thirds of its search referral share ^[9] while its CEO declines to plan past seven weeks ^[6], and whatever replaces that traffic will take longer than seven weeks to build.