In "Purchase planning", the weakest link is access to data. Every source has a keeper, but the whole has none, so most of the effort goes into gathering the data rather than judging it.
What works: shared meaning of the numbers and use of data in the decision. Also weaker: data preparation and early signal; you will come back to them in "What comes next, once this works".
Scattered data and the fact that only one person knows how to stitch it together are one problem, not two: as long as the data sits in several places, someone has to stitch it and becomes the bottleneck. That is why we start with access, not with preparation.
This topic is in your plans, and the first step will show results within a few weeks without new tools or extra people. We suggest starting now with the step "One place for the data for this decision".
To begin with, the first point of the step below is enough, and you can do it tomorrow.
Do you find out about problems in time?
Not always. Sometimes you find out about important things too late, and nothing warns you in advance.
More below, in "Early signal"What this really means
In practice, whoever happens to need the data is responsible for joining it, and everyone does it their own way. So every stitch-up is slightly different, and the numbers prepared for one decision are not fully comparable with those from the previous one.
The cost nobody usually counts is the questions that stop being asked: if gathering the data takes longer than the decision itself, people decide without it and nobody records that. The second is the price of every extra question from management, because each one means gathering from zero again. The third is a stitching error that surfaces only at the next decision, when the numbers do not match the previous ones.
Good news: fixing access here does not require sorting out definitions, because the numbers mean the same thing to everyone, and when differences do appear, you clear them up quickly. That usually saves the longest and most contentious part of such a change.
The smallest sensible step
The data for the decision "Purchase planning" lands in one place, in an agreed way, before it is needed.
One place for the data for this decision (2 to 4 weeks)
- Tomorrow, ask the person who prepares this data to list every place they take it from and how long assembling the whole takes. If the list has more than three items or the time is counted in hours, the diagnosis is confirmed. Write that time down: it is the starting point for measurement.
- With the team, list the numbers this decision cannot be made without; everything else waits.
- For each number, name one source and one person who keeps it current by the agreed day.
- Put them in one file or report filled in the same way every time; from the next decision use only that.
Is the step worth it: quick maths
The effects you point to in your answers: "the team loses time on manual preparation, corrections or reconciling" and "we miss opportunities because we learn about them too late". Count those first.
Count: hours of gathering and stitching per decision × the number of such decisions a year × the hourly rate of the person who does it, plus days of waiting for the complete data × what each day of delay costs in the decision "Purchase planning" (for example a quote waiting for an answer, or a job that cannot start). Compare the result with the cost of the step: usually a few days of the team's work spread over 2 to 4 weeks. If the yearly result is clearly higher, the step pays for itself in recovered time alone. If not, only make the list of numbers and their sources, and put the one place aside until the decision starts happening more often.
If the cause is different
Scattering can also come from one source that is hard to extract, while the rest of the data is at hand. You will recognise it when one place on the list from the test takes most of the time or always arrives last. Then start there: agree a standing export with its keeper or the system vendor before you build one place for everything.
Five questions for your team
The questions concern the decision "Purchase planning". If you make it yourself, the questions for the person who decides are questions for you.
- To the buyer: how many times at the last order did you have to assemble the data from scratch because someone asked for a different number?
- To the person who makes this decision: which three numbers about stock do you really use, and which do you only look at?
- To the warehouse keeper: who besides you can give the current stock and received deliveries when you are away?
- To the buyer: which order recently had to be corrected with the supplier because stock or orders in transit differed from the summary?
- To the person who makes this decision: how many hours a month go into gathering data on stock and deliveries, and how many into the decision itself?
The typical mistake at this stage
Companies at this stage most often buy a reporting tool or commission a "data warehouse" before agreeing exactly which numbers the decision "Purchase planning" requires. The tool connects to everything that exists and reproduces today's chaos in a prettier form. After a few months someone is stitching a spreadsheet on the side again, because the tool lacks what is really needed.
What it looks like when it works
The buyer opens one summary and sees stock on hand, orders in transit and the last few weeks of sales without asking anyone. Every order to a supplier is made from the same numbers, so it is possible to compare what was ordered and what came of it. When delivery data is missing, everyone knows who is responsible for entering it and by when. The purchasing meeting is about decisions, not about gathering data.
On your own or with help
The first actions, the list of numbers and naming the source owners, the team will do on its own within a week; that is a conversation, not a project. Such steps usually stall after a few weeks, when a spreadsheet someone is attached to has to go, or when it has to be settled which source is the right one. Outside help makes sense exactly there: someone from outside the team finds it easier to ask "what is this number for" and keeps the one place from growing into ten.
What comes next, once this works
Once the one place works, preparation is next: write down how the summary is produced and remove the actions that have to be done by hand every time. Then agree who looks at this data before the decision and when, so the complete set does not sit unused.
An earlier signal built on today's data will inherit its delays, gaps and manual corrections. So the order is: first the data for the decision "Purchase planning", then the signal.
When it starts costing more
- The keeper of one source changes the system or the file layout and the current stitching stops fitting.
- A second decision appears that needs the same data, and someone starts stitching it in parallel.
- Customers, jobs or employees multiply, and gathering time grows with them.
AI in your case
AI will not solve the basic problem here: there is no single place it could read from, and a model fed hand-stitched files will inherit their gaps and delays. Automation at this stage will cement today's stitching: someone will still do it, only for the automated step. AI helps later: once the data for the decision "Purchase planning" is in one place, a simple model can flag deviations or suggest what to look at first. The condition: the same complete set of data at every decision, without manual corrections. A question for a vendor who proposes AI: "Exactly which sources will it read from, and who is responsible for keeping them current?"
Early signal
The value of an earlier signal can be judged on the history of decisions, and in many companies nobody records that history. Start keeping it now, without any tool: at every decision, write down in one place what was decided, on which numbers, and what came of it. You will then be able to test any idea for a signal, your own or a vendor's, on real decisions.
Finally
This report was built solely from your answers about one process. We did not check your systems or data and make no claims about the rest of the company.
Once the step is done, come back here and answer again. Compare the new report with this one: you will see whether the weakest link has moved and what comes next.
If you want to take this step with Codino's help: "One-process review" (2 hours). A ruling on which source is the right one for each number, a short list of the numbers the decision cannot do without, and a plan your team follows to build one place for that data on its own. You can also start with a 30-minute conversation to judge whether you can take the step on your own.
Let's talk about the decision "Purchase planning"
30 minutes. Bring this report and leave with the list of numbers this decision can't do without, plus the first step planned for the week ahead. No sales pitch at the end unless you ask for one.
Get in touch