Reading the portfolio for real risk

The review that was doing everything right

Attention is the scarcest thing anybody running a season of orders has. It goes, naturally enough, to whichever order is making the most noise. The orders that make none are the subject here. This lesson looks at a well-run weekly meeting whose agenda came from the problems everybody already knew about. Then it does something less comfortable: it decides in advance which orders will not be looked at.

Hollowell's critical path review for this season ran every Thursday morning from 18 June to 24 September. Fourteen meetings. Attendance 96%. Average length 52 minutes. The buyer came, the technologist came, sourcing came, and decisions were taken in the room rather than deferred.

The agenda was built from the flag list. Anyone could flag an order: a supplier emailing about a problem, a milestone going red in the tracker, or Bryony's own judgement. Across the fourteen meetings there were 131 order-appearances, covering 29 of the forty-one orders. Twelve orders were never discussed once.

At the end of the season four orders arrived in the DC after their window closed. Here is where they came from.

Read that carefully, because the obvious reading is wrong. The obvious reading is that the review missed things. The true reading is the opposite: the review worked. Orders that got attention shipped at 96.6%. Attention is effective, and that is exactly why the unattended group is where the damage was. Three of the four failures were in a group of twelve that a well-run, well-attended, genuinely useful meeting had been carefully not looking at for fourteen weeks.

A meeting whose agenda is the flag list can only ever review problems somebody already knew about. Its success rate is measured on the orders it selected, and it selected them because they were already visible. The fact card on this lesson is about a wartime version of the same mistake, and it is worth reading before you build your next agenda.

Why those twelve were quiet

None of them was hidden. Each had an ordinary reason to fall below the surface.

• Five were at Rupganj, which sends a beautifully formatted report every Monday, always on time, always green. It is generated from the order-entry dates rather than from the floor. Whether a system should be able to do that is a question for track 10, and for course 10.3 in particular. What matters here is that no eyes were behind the number.
• Four were small: under 6,000 units and under USD 40,000 each. They sat below everybody's private threshold of interest, including Bryony's. That threshold is worth examining, because a small order is not a small problem. It queues at the same mill, waits in the same approval line and books space on the same vessel as a large one, so it is exactly as likely to slip. What is not proportional is the rescue. A re-inspection, a partial air shipment or a second dye lot costs roughly what it costs whatever the quantity. So a small order that goes late is the most expensive kind there is to recover, measured per unit. Size belongs in the tier rule as one input among several, never as a reason to look away.
• Three were at a factory that had no problems last season. A supplier's history is a real input and it belongs on the scorecard, which is course 27.5, but a clean record is a reason to check less often, never a reason not to check.

Now the specific one. HL-8823, Perambur Knits, 7,200 interlock dresses, DC date 26 October. Fourteen weeks of weekly reports, every milestone on time, and, this is the signal, not one date changed in the entire season.

Its fabric was reported in-house on 21 August, and that was true: the rolls were in the store and the count matched the invoice. The four-point inspection was not done, because the quality man was covering another order that week and nobody chased the report, which does not exist to this day. When cutting opened the rolls on 9 September, 22% of the fabric was second quality. There was no overage to cover it, and the replacement dye lot took nineteen days.

The gap it walked through is the one named in the first lesson: in-house is not inspected. The board said fabric complete because a milestone with that name had a tick against it, and the milestone with the evidence behind it was a different milestone that nobody had asked for.

The silence is the signal

Bryony's tracker recorded every date movement of the season: 312 changes across forty-one orders, an average of 7.6 each.

Two of the four late orders came from the three that never moved a date. Three orders out of forty-one held two of the season's four failures.

This inverts the instinct. A supplier who keeps changing dates feels like a problem, and a supplier whose dates never move feels safe. In fact a factory that revises is a factory that is looking. The revision is the sound of somebody comparing a plan to a floor. Perfect stillness across a fourteen-week manufacturing programme is not stability. It is the absence of observation, and it is the highest-risk, lowest-attention state an order can be in.

The same inversion applies to your inbox. Rupganj sent Bryony 214 emails across the season and finished, on average, six days ahead of its ex-factory dates. Abu Rawash sent 38 and slipped nine days. Nine suppliers is an anecdote and not a correlation, and it would be dishonest to present it as one. But it is enough to stop you doing the thing a busy week makes you do, which is to let inbox volume decide where your attention goes. The supplier who writes to you is the one whose problems you already know about.

Triage, and naming what you will not look at

Bryony's week is 37.5 contracted hours. Approvals and the chasing they generate take about 14 hours. Meetings take 9.5. Costing, purchase orders, amendments and sample administration take another 9.5. What is left for actually reading the portfolio is 4.5 hours: 270 minutes for forty-one orders, or 6.6 minutes each.

Spreading 6.6 minutes evenly is the worst available option. It is too little to understand anything, and it is spread across orders that do not need it. So tier the book by float and exposure, recompute it weekly, and let orders move between tiers as their float falls.

That comes to 242.5 minutes of the 270, leaving 27.5 minutes for whatever arrives on Tuesday. Tier A gets nearly two thirds of the time for under a quarter of the orders, which is correct: they hold the risk.

Act as a sceptical portfolio reviewer. I will give you my order book and the record of what has been discussed, and I want you to find what is being missed rather than to summarise what is known. Here is the book: [ORDER, FACTORY, PRODUCT, UNITS, FOB VALUE, DATE DUE AT THE DISTRIBUTION CENTRE, FLOAT THIS WEEK]. Here is every date change recorded against each order this season: [ORDER, NUMBER OF CHANGES, AND THE DATES IF I HAVE THEM]. Here is which orders appeared on a review agenda and how often: [ORDER, APPEARANCES]. Here is my inbox volume by factory: [FACTORY, MESSAGES]. Do the following. First, list every order that has never appeared on an agenda, with its float and its FOB value, and rank them by exposure. Second, list every order whose dates have not moved in three weeks or more, and treat that as a risk signal rather than a good one — explain to me in one line each why. Third, cross-reference the two lists and tell me which orders are in both, because those are the ones to look at first. Fourth, tell me where my attention has actually gone, by comparing inbox volume per factory against the float held at that factory, and say plainly whether the loudest supplier is the riskiest. Fifth, for each order in your top five, give me the single question to ask and name the dated document that would answer it. Sixth, propose the agenda for this week's review with a time budget per order that fits [MINUTES] in total, and list what the agenda deliberately excludes so I can put that exclusion in front of my buyer in writing. Do not reassure me about an order because nothing has been reported against it.

Two things make this honest rather than merely tidy.

First, the tier is recomputed from float, not decided in June. HL-8817 was a tier B order in June with eleven days of room, and a tier A order in July with two. A tiering set once at the start of a season is a filing system, not a triage.

Second, write down what you are choosing not to see. Tier C is looked at once a month, and once a month is not enough to catch a fabric failure in week two of the gap. Something in tier C will bite. The point of naming it in writing, in the review pack, in front of the buyer, is that when it does bite it is a known and accepted cost rather than a surprise. Nobody has to pretend afterwards that the system failed. It did what it was set up to do with the hours available. If the hours are wrong, that is a resourcing conversation, and the way to win it is to have written the trade-off down in June rather than to explain it in October.

What you leave with

A float board, a queue sorted by float, a tiered rota that fits your hours, and a written note of what you are not looking at.

Two of those are free and one of them is a conversation. None of them requires a system, a headcount or a supplier to behave differently. What happens when the float finally goes negative, the escalation, the expediting, the claim, and the question of whether a claim is worth more than the supplier, is course 27.6. How you judge the nine factories against each other once the season is over is course 27.5.

This course has been about the weeks before either of those becomes necessary. It has been about the two facts that a portfolio teaches and a single order cannot. You are on the critical path. And the order you are not looking at is the one to look at.