From report to decision: five questions

The monthly report says the share of late orders has risen from 4 to 7 percent. At the meeting everyone shakes their head, says "yes, that is bad" and moves to the next topic. A month later the figure is back in the same report. A report shows the situation; a decision turns it into work that gets done after the meeting. To get there, put five questions to every important number.

The five questions

1. What changed? The number, the base of comparison and the usual range. For example, over the last six months it was 3.6-4.5 percent.
2. Why? Break it down by shift, day, category, customer and zone: the cause usually gathers in one or two places.
3. Does it matter? Put it in so'm: lost sales, extra work, cancelled orders.
4. Who will act? A specific job title or name, not "the warehouse team".
5. By when? The date of the decision and the date the result is checked.

One line in the report looks like this: "Late orders: 7 percent (was 4). Cause: evening shift. Decision: extend the shift by an hour. Owner: shift leader. Check: in two weeks".

An example: late orders

There are 2,500 orders a month. 4 percent is 100 late ones and 7 percent is 175: 75 more.

What changed? 3 points, 75 orders. The figure has left the usual 3.6-4.5 percent, so it is not chance. Why? Of the extra 75 orders, 52 fall in the evening shift: online-shop orders have started arriving mostly after 17:00 while the pickers leave at 18:00. Another 14 are down to a lack of stock, and for 9 no cause was found. So 52 ÷ 75 ≈ 69 percent sits in one place.

Does it matter? Suppose that last quarter one in ten late orders was cancelled. 75 × 0.1 = 7.5, about 8 orders. With an average basket of 420,000 so'm the monthly loss is 7.5 × 420,000 = 3.15 million so'm. Who will act? The evening shift leader and the warehouse manager. The decision: extend the picking shift by an hour, at an extra wage cost of 0.6 million so'm a month. The gain is 3.15 − 0.6 = 2.55 million so'm. By when? From tomorrow, with the result checked in two weeks. Write the success criterion down in advance: out of 1,250 orders in half a month, late ones should not exceed 4.5 percent, that is 56. If they do, the cause or the measure was wrong.

"Leave it alone" is also a decision

Another figure from the same report: stock accuracy slipped from 96.8 to 96.5 percent. Over the last six months it ranged between 96.2 and 97.4, and a drop of 0.3 points fits inside that range. Decision: leave it alone and come back if it falls again next month. Write that down. Otherwise a month later someone asks, "Why was this ignored?" and there is no answer.

Be careful: if such a drop repeats three months running, it is no longer chance and the decision to "leave it alone" must be reviewed.

Decisions come in three kinds: act, watch (with a date) and ignore (with a reason). "We'll see later" without a reason is not a decision.

The decision log

Keep one table: date, number, decision, owner, deadline, result. At the start of each month review last month's decisions: did late orders fall back from 7 percent to 4.3? If not, the cause was misjudged or the measure was too small. A meeting should not run longer than 20 minutes: 5-6 numbers, one decision for each.

Decisions usually get lost in these places:

• A guess in place of a cause: "staff are careless", without splitting the data by shift and category.
• "Everyone" as the owner: a task without a name does not get done.
• Watching without a deadline: if you write "we will monitor", say until when and until which figure.
• A decision whose result is never checked: if it is not reviewed next month, keeping a log is pointless.

Conclusion

A report gives a number and a decision turns it into action: what changed, why, does it matter, who will act, by when. The cheapest improvement is to add a "decision" column to the report and open every meeting with last month's decisions. CountLab gathers figures from stock counts and discrepancy analysis into a management report together with causes, owners and a suggested next step.