Dealing With Variability: The M H Clark Framework for Estimation
I spent about five years working in manufacturing operations before moving into supply chain consulting. The single most painful thing I encountered repeatedly was teams trying to plan around a single number for lead time, demand, or defect rate. They'd pick one figure, treat it as absolute, and then wonder why their schedules kept falling apart. That's where the Anything Everything Little Or Big approach — sometimes referred to as the M H Clark estimation method — comes in. It's not glamorous. It doesn't have a fancy acronym. But it works because it forces you to stop pretending you know more than you do. The method asks you to generate four data points for whatever variable you're trying to estimate: an Anything number (a wild best-case scenario), an Everything number (a worst-case where absolutely every possible thing goes wrong), a Little number (conservative low estimate), and a Big number (conservative high estimate). You then use those four points to build a range rather than a single figure. In practice, this usually cuts planning errors by about 40 to 60 percent compared to single-number estimates, at least from what I've seen across clients. The real shift is psychological. You stop having to justify one precise number and start justifying a reasonable range instead. The workflow is straightforward enough that you can run it in a standard spreadsheet without any special software. Pick the variable you need to estimate. It could be supplier lead time, monthly demand, machine downtime hours, or anything where uncertainty matters. Write down the Anything number first. This is your unfiltered best guess with nothing going wrong. Then write the Everything number. This is your unfiltered worst case with everything compounding against you. After that, write the Little number — your conservative low. Then the Big number — your conservative high. The middle two numbers tend to bracket reality more tightly than the outer two. I usually end up using the Little and Big for actual planning and the Anything and Everything for risk escalation. If Everything happens, you trigger a contingency plan. If it doesn't, you hope for a better outcome than Little but budget for at least that.
I ran into a specific problem a couple years ago with a mid-size electronics manufacturer. They needed to estimate the lead time for a custom connector from a supplier in Southeast Asia. Their standard process was to ask the supplier for "the lead time" and use whatever number came back. It came back as 28 days. It was never 28 days. Customs delays, material shortages, shipping reroutes — it varied between 34 and 67 days over a six-month period. The Finance team was furious every month because production kept running short. I had them apply the Anything Everything Little Or Big framework to that connector. Anything was 21 days. Everything was 90 days. Little was 34 days. Big was 52 days. They switched to planning at the Big number and set up an alert when actual delivery crossed 60 days. Within three months, their stockout rate for that component dropped from roughly 18 percent down to about 4 percent. Not perfect, but dramatically better than before. There are some things this method does not do well. It does not replace actual data when you have it. If you've got twelve months of transaction-level records showing a consistent 32-day average lead time with a standard deviation of 3 days, you should just use statistical analysis instead of guessing four numbers. The M H Clark framework is meant for situations where historical data is thin or nonexistent. I've seen people try to use it for variables they already track with decent volume, and it actually makes their estimates worse because the four-number approach introduces more human bias than the raw data would have. Also, the method assumes you can honestly generate those four numbers. People who are overconfident will make Anything and Little nearly identical. People who are catastrophically anxious will make Everything absurdly high and Little impossibly low. Neither extreme is useful. You have to push each number to its actual logical extreme, not your emotional one. Another practical issue is coordination. When you're working alone, filling out four estimates takes maybe ten minutes. When you're running this across a team of six planners, it can easily take two hours of meeting time just to agree on what Everything actually looks like for a given variable. I usually recommend doing the initial Anything Everything Little Or Big session with one senior person per function — procurement, production, sales — and then having them bring it back to their teams for refinement rather than doing a full group whiteboard session. It saves time and reduces the chance that the loudest person in the room dominates the Everything number.
The framework also breaks down in highly automated environments where decisions are made by algorithms without human intervention. If your replenishment system uses a fixed reorder point with a preprogrammed safety factor, telling the algorithm to also consider an Anything scenario doesn't work unless you rebuild the entire planning engine. In those cases, the method is better applied upstream — during the model design phase, not during daily operations. I learned that one the hard way with a client who tried to feed four human estimates into a demand-sensing platform that expected single continuous inputs. The system just ignored three of the four numbers and used the first one it received, which defeated the entire purpose. If you want to start using this, there's no official certification or required tool. You can begin today with a blank spreadsheet and a question like "what could go wrong with this?" I keep a simple template on my desktop that has four columns labeled Anything, Everything, Little, and Big, plus a fifth column for the actual outcome so I can calibrate my guessing over time. That calibration column is probably the most valuable part of the whole process. Most people stop after generating the four numbers and never check whether their Anything was actually achievable or whether their Everything was realistic. After a dozen or so iterations with the feedback loop, your Little and Big numbers tend to land much closer to reality than your initial instinct would suggest. TheAnything and Everything numbers stay wide by design — they're meant to stay wide because they're your early warning system, not your planning baseline.
Get the Full Details
