Supply Chain Action Points
Read this first — the conclusion, and the moves to make:
- Before any platform work, inventory master data: count every active SKU, lane and supplier missing a lead time or a unit cost, and set the go-live gate at 100 percent of active SKUs carrying both, with a named owner per data domain.
- Write the decision rights down before go-live: who may override the plan, above what value threshold, and within how many hours; a tower without an override rule only relocates the argument.
- Run one full seasonal window in parallel, old process and platform side by side, for 8 to 12 weeks, and compare forecast error and replan latency before you cut over.
- Tie the second budget tranche to a measured metric, for example cutting the weekly replan cycle from 5 days to 1 day, and hold it until the number actually moves.
- Set the exit condition now: if two seasons pass and forecast error has not improved by an agreed margin, the constraint was the data, not the platform, so renegotiate scope rather than renew.
Supply Chain Dive reports Scotts Miracle-Gro is expanding its use of Kinaxis technology to unify planning, decision-making and execution on a single platform. The lawn-and-garden brand's move reflects broader CPG adoption of AI-driven control towers to manage demand volatility and tariff exposure.
The Analysis
A lawn-and-garden brand just widened a single-platform planning rollout. Supply Chain Dive reports Scotts Miracle-Gro is expanding its use of Kinaxis technology to pull planning, decision-making and execution onto one system, and frames it as one more consumer-goods name joining the AI control-tower crowd. The two drivers named in the report are demand volatility and tariff exposure.
I want to pull those two apart, because a control tower is a queue with a dashboard, and before anyone copies the announcement into a roadmap it is worth asking which problem is actually a systems problem and which one is a data problem wearing a systems costume. If you cannot say where the queue backs up, you are buying a screen.
So I take the claim apart the way I take any throughput claim apart. What enters, what leaves, where it piles up, and what the number on the wall is really measuring.
Start with what unify planning, decision-making and execution actually means at the wiring level. Those are three different clocks. Planning runs on a weekly or monthly cycle, the sales-and-operations rhythm, where you set the total volume and the inventory target. Decision-making runs daily to hourly, where you allocate stock, release reorders and pick which order ships from which node. Execution runs in real time, at the dock and the warehouse, where the truck is or is not there. Unifying them means one demand signal and one set of writes feeding all three clocks, instead of three teams maintaining three versions of the same truth. The value is not the screen. The value is that nobody has to reconcile a number by hand every Monday. Take the reconciliation away and you give back hours, and hours are the only unit that compounds.
Why now, in this category specifically. Lawn-and-garden is weather-driven, which means forecast error is structurally large: demand is lumpy, seasonal, and it lumps into a handful of weeks when the weather turns. Tariff exposure is the second driver, and it works differently. Volatility makes you want to replan faster. Tariff exposure makes the landed cost itself a moving target, so a plan that was correct in March is wrong in June without a single unit of demand changing. Both of these are exogenous. You cannot forecast your way out of either one. What you can do is shorten the loop between a change and a response, and that is the honest case for a single platform. Not smarter forecasts. Faster replans. Those are two different purchases, and only one of them is on the slide.
Who feels this first. A large shipper that already has a demand-planning function, clean master data and someone whose full-time job is the process gets most of the upside, because the tower has something to optimise. A mid-size or small seller usually does not have the master data to feed it, meaning lead times per lane, unit costs per SKU, supplier constraints, so the same platform hands them a live dashboard and very few live decisions. The tool is not the differentiator there. The data underneath it is. Within a category, the weather-exposed seasonal lines feel the volatility driver more than the staple lines, and any line that crosses a tariff code feels the cost driver more. Sort your own portfolio by those two axes before you buy anything.
The timing is the part people get wrong. A platform change does not shorten your lead time this season. Integration and master-data cleanup run in quarters, not weeks. So the first thing you harvest is visibility, and the second is service, and they arrive months apart. If you are on a seasonal calendar, that gap is the whole ballgame: you will see better data before you can act on it, and if the season closes in between, you paid for the year and got the picture.
Here is the point the announcement does not make. The bottleneck in most control towers is not the algorithm. It is the data model and the decision rights. If two systems disagree about what a SKU is, the tower picks one and is confidently wrong at scale. If nobody has written down who is allowed to override the plan, and within what window, then the platform does not settle the argument, it just moves it into a meeting with better charts. And a single platform is, by construction, a single point of failure and a lock-in: one outage stops the loop, and the switching cost climbs every quarter you stay. None of that is a reason to avoid it. It is a reason to make the boring parts explicit first.
Quantify before you optimise, which is the only order that works. Put your own numbers on the two drivers. Assume a seasonal line where 30 percent of annual volume lands in a 12-week window, forecast error around plus or minus 20 percent, and a carry-and-markdown cost of roughly 25 percent of unit cost. Assume a unit cost of 10 dollars and one million units a year. The window carries 300,000 units, a 20 percent miss is 60,000 units, and 25 percent of 10 dollars is 2.50 a unit, so one bad season is about 150,000 dollars of mis-forecast inventory. That is the number the faster replan is competing against, and it is the number that should set your scope, not the demo.
The counter-case is simple. If demand volatility falls, or a tariff question gets settled by a ruling, the urgency drains out of the purchase. A stable, predictable flow of goods is exactly the world where a single platform is an expensive way to look at things you already knew. The tower is priced against chaos. Remove the chaos and you are paying for the picture.
↑ Back to the key points