Real Time Is Table Stakes

Published 2026-07-16 · Updated 2026-06-23 · By Alan Wizemann

Topics: AI & Machine Learning, Data & Analytics, Enterprise, Product Strategy & Development

For most of my career, the holy grail in data was real time, and everyone in the field chased it. Stop waiting on the end-of-quarter report, and stop reconciling a number that's already three weeks old by the time anyone actually looks at it. See the business as it actually is, right now, this minute, and act on the truth of it instead of on a memory of it from a month ago. Every transformation I ever worked on had some version of that ambition baked right into it from the start, and we largely got there in the end. And then something quietly shifted in how I think about the whole thing, which is that real time turned out to be the floor, not the ceiling, and that reframe changed what I went looking for next.

When we normalized the data across our markets at Southern Glazer's, the data was running about thirty minutes behind reality. For reporting, that's nothing at all – for most decisions, thirty minutes old is functionally live (close enough to now that the gap doesn't change what you'd do), indistinguishable from the present. But the moment you can actually see the business almost as it happens, you start to notice an uncomfortable thing, which is that seeing it isn't even close to the same as being ahead of it. You're still reacting to events. You're just reacting to them faster than you used to, and faster reaction is genuinely valuable, but it is not the same animal as anticipation, and I came to want the second thing far more than the first.

Here's the analogy I keep coming back to, because it makes the difference concrete. When you put a Tesla into full self-driving mode, the genuinely impressive part isn't that it sees the road right now – cameras have always seen the road right now, that's the easy part. What it's actually doing, the part that matters, is simulating two seconds ahead of itself, constantly. Is that dog about to step off the curb, is that car going to turn across my lane in a moment, is that pedestrian about to cross against the light? It's running the next few seconds over and over again so that it can act before the thing actually happens, rather than scrambling to react to it after it already has. Seeing the road is table stakes, the absolute baseline. Anticipating the road is the entire point of the exercise, and business data is moving in precisely the same direction. Real time tells you there's a stockout right now, which is useful, and certainly better than first learning about it in the quarterly report long after the fact. But the version that actually changes the business is the one that tells you a stockout is coming, in this specific market, in this specific product, given these patterns, while you can still do something about it. The value isn't sitting in the present anymore. It's in a credible, honest view of the near future.

I saw what this actually looks like in practice through the supply-chain side of our operation, and it stuck with me. Once you have clean, normalized data sitting underneath everything, you can start asking forward-looking questions that used to be flatly impossible to even pose. What if we manufactured this product over here instead of over there? I know the cost of the bottles, the cost of the packing, the cost of the transportation, all of it. Would moving it actually save real money, or does it merely look cheaper on the surface until you run the whole thing through? And the interesting answers, the useful ones, are almost never the obvious ones you'd have guessed sitting in a conference room. A hub that's physically closer turns out, on the numbers, to be more expensive. A facility two states away turns out to save money. Those aren't things you would ever guess from intuition – they simply fall out of the data once you let the model build the entire scenario for you instead of building a rough sketch of it in your own head.

That's the real shift, when I lay it out plainly: from dashboards that describe the present to models that simulate options for the future, from "here is exactly what happened" to "here's what's likely to happen next, and here's specifically what changes if you pull this particular lever." An executive making a network decision shouldn't be sitting and staring at a static report at all. They should be able to drag things around on a map, in effect, and watch the projected cost and the projected risk move in real time right along with them, all of it grounded in actual operational data rather than gut feel and instinct.

A few things have to be genuinely true for any of this to work, though, and they happen to be the unglamorous parts that nobody puts on a slide. The first is the data foundation, which is the boring work that I will quite literally never stop talking about to anyone who'll listen. You cannot simulate the future on top of a mess, full stop. If your bottle costs and your transportation costs and your inventory positions all live in systems that don't actually agree with one another, then your forecast is just a confident-looking guess wearing a nice chart. The entire forward-looking magic is downstream of getting the present clean and consistent first – skip that step, and what you've built is a very expensive crystal ball that lies to you while sounding sure of itself.

The second thing is plain honesty about what a simulation actually is, underneath. It's an educated guess, grounded in real-world data, and that's a strength rather than a weakness, as long as everyone in the room genuinely understands that's what it is. The real danger arrives when a forecast quietly gets treated as a hard fact, simply because it happened to come out of a sophisticated and expensive-looking system. The discipline here is the exact same one we used for our reporting – where the model is uncertain, it has to say so, out loud, and a flagged "I'm honestly not confident about this one" is worth far more than a precise-looking number that nobody should actually trust. The third thing is knowing where simulation runs straight into walls that aren't technical at all. In our industry the obvious one is regulation. A scenario might show, cleanly and correctly, that moving product across a particular state line saves real money, and the data backing it can be entirely right, and the whole thing can still be completely irrelevant, because the law simply won't let you do it. A forward-looking model built without those constraints baked in is a foundation built on sand. The hard part of anticipation, more often than not, isn't the math at all. It's encoding the actual rules of the real world the business operates inside of.

None of this means that chasing real-time data was ever a wasted goal, to be clear – it's a prerequisite, not a detour. You genuinely need the present to be clean and current before the future is even worth modeling at all. But I've stopped treating "now" as the destination, the finish line, the thing we were all running toward. Now is simply the expectation, the price of being in the game. The question I'm actually interested in these days is the one that only shows up after you've already won the race to the present: what can you see coming, and what are you going to do about it while there's still time left to act? The companies that win the next stretch of this won't be the ones with the freshest, shiniest dashboards, because everyone is going to have those soon enough. They'll be the ones that managed to turn a clean view of the present into a credible view of what comes next, and then had the discipline to actually do something about it, well before the report would ever have told them to.