If you cannot tell your vendor whether it will receive 10,000 invoices or 20,000, but you expect every invoice processed on time, what exactly do you think he or she is going to put in the rate?
The cost of being ready for you, aka perhaps the #2 or #3 most ‘unsolvable problem’ that we had when I worked at google (not a great comparative example for every company out there, but its what i got 🤷).
That might mean keeping capacity available, paying overtime, moving people from another account or scrambling to train extra staff. A vendor that understands this risk before it prices the service has to decide how it will cover it. Your uncertainty can become part of the rate you pay on every unit of work.
Then procurement comes along and asks for a discount.
Last week’s ProcureCon article was about ownership after signature. Forecasting is where that question gets expensive. Someone makes a business decision that creates work. Someone else forecasts it. Another team manages the vendor. The bill lands somewhere between them.
I spent a large part of my career in that space.
We had a centralized forecasting function. We did not have a centralized business.
At Google, we spent a lot of time and money every month trying to provide accurate forecasts to our vendors.
We were trying to coordinate with product managers so we could find out about a promotion for Pixel sales in Japan before it started, ideally more than 24 hours after it had already started. We were trying to consolidate inputs from dozens of internal teams whose decisions could influence how many support calls arrived.
There was a centralized demand-planning function. The processes creating the demand were spread across the company.
The people responsible for forecasting were not the people responsible for managing the spend on the services using those forecasts. Product areas had powerful leaders with their own priorities. Another million dollars to have a vendor work around the clock was not necessarily the thing that got their attention.
Customer satisfaction did. But connecting a bad forecast to the scramble to protect CSAT, and then connecting that scramble to what we paid, was much harder.
You could end up with a successful promotion, acceptable customer satisfaction and an expensive vendor operation. The teams involved were looking at different results.
A centralized planning team cannot fix that by asking everyone to fill in the same spreadsheet more enthusiastically.
If it happens every month, stop calling it a surge
There are real surprises. There are also recurring failures to tell the vendor what is coming.
Imagine a provider handling invoices. It is told to expect 10,000, receives 20,000 and is still expected to meet the processing target. That is a simple hypothetical, but the operating problem is familiar: somebody has to find the capacity.
Now imagine the same thing happens the following month. And the month after that.
The vendor learns that the client’s forecast is unreliable. At the next pricing discussion, it has little reason to assume the neat number in the RFP represents the work it will actually have to deliver.
A premium for uncertainty can become a permanent feature of the relationship. You pay for flexibility because you have not given the provider a credible basis for planning something cheaper.
Knowing that 20,000 invoices are coming gives the vendor a different staffing problem from discovering it as they arrive. The volume may be identical. The notice, staffing choices and delivery risk are not.
That is the commercial value of a forecast. It lets the provider make decisions while it still has useful options.
The vendors could forecast their own work better than we could
I tested this at Google many times. We let vendors produce forecasts alongside our formal internal forecasts, drawing on their long experience running the operations.
The results were almost embarrassing. Across virtually every workflow we tested, the vendors predicted volume materially better than we did internally: customer service, content moderation and what we called “human judgement” work, much of which we would now describe as AI model training.
They had better records of the work they were actually doing. They understood their scope, the recurring patterns and the things that made an apparently ordinary week difficult.
We were spending a great deal of effort forecasting work that someone else had been observing and recording for years.
Historical operating data is the best place to start. What arrived, when it arrived, what kind of work it was, how long it took and what happened next. A forecast should be able to explain why it departs from that history.
If you regularly receive 20,000 invoices, a forecast of 10,000 needs an explanation. Repeating the same surprise does not make the underlying pattern less visible.
Then add the information history cannot know yet: the promotion, product launch, policy change or migration somebody inside the company is planning. Get it to the people making capacity decisions while they can still use it.
The vendor may have the better baseline. The buyer should have the earlier warning about what its own business is going to change.
The catch: every vendor only sees its piece
The vendors’ advantage did not solve our portfolio problem.
Each could see the work inside its own scope. None of them could give us a complete view of everything we were buying, all the internal teams creating demand or the effects across other providers.
One business decision can affect several services. A promotion can increase support contacts and work further along the order process. A policy change can alter both the volume and difficulty of moderation work. Moving work between providers changes each vendor’s forecast without necessarily changing total demand.
That is the view the buyer has to assemble.
We need to connect the historical work to the service, the business using it, the supplier delivering it and the commercial terms that turn it into spend. Then connect the planned business changes to the services they could affect.
This is where invoicing, demand management, operations and rate negotiation intersect. Invoicing tells us what we were charged. Operating records tell us what work arrived and how it was handled. Demand planning adds what may change. The agreement tells us how volume, capacity and service expectations become a price.
Those connections do not appear just because the demand-planning function is centralized. Outsourced services can still sit outside its coverage, or inside it with important inputs missing.
And a vendor total will not tell you which part of that relationship is carrying the uncertainty. You need to get down to the actual service.
Give the people creating demand a view of its cost
A product manager should be able to see that the timing of a promotion affects the capacity we need and the choices available to provide it.
That does not mean procurement gets to veto the promotion. It means the business can make the decision with its delivery consequences visible.
Take a recent miss and follow it through. What did we expect? What actually arrived? When did someone inside the company know it might change? When did the vendor find out? What did the provider do, and what did that cost? What happened to service?
Sometimes the vendor protects the customer experience by spending more. CSAT holds, and the business never sees the failure that nearly happened. The extra cost deserves to be part of that conversation, alongside the revenue or other benefit the business decision created.
Sometimes service suffers as well. Establish the connection using the actual timeline and operating records; a forecast miss is not proof that every subsequent performance problem was caused by the buyer.
The person who owns the business change, the person producing the forecast and the person managing the vendor spend need to be looking at the same event.
Take the uncertainty into the rate negotiation
Before asking the provider for a better rate, ask what uncertainty it has assumed in the current one.
What demand range is it planning for? How much notice does it need to change staffing? What capacity is included? What happens when the work arrives outside that range, or with a different mix of skills?
Then ask the useful commercial question: if we improve the information and notice we give you, what can you price differently?
A credible forecast can support a specific trade. You provide an agreed planning window, a usable range and earlier notice of changes. The provider explains what that lets it change in its capacity plan and what benefit comes back to you.
Put the treatment of expected demand and genuine exceptions into the commercial discussion. Otherwise, you can improve the forecast and continue paying the old price for uncertainty.
Better forecasting does not automatically entitle you to a discount. The benefit depends on what cost or risk you actually remove, and what the provider agrees to give back. Nor should you promise a volume commitment your own business will not stand behind.
The point is to negotiate something operationally real. “Trust us, next month will be different” is not much of a concession.
Start with one forecast the vendor does not trust
Bring your forecast, the vendor’s forecast and the actual work together for the same service and periods. Compare what each side knew before staffing decisions had to be made, not a revised forecast produced after the work arrived.
Use enough history to see the pattern, and the daily or weekly detail that matters to the operation. A monthly total can look accurate while the work arrives in a completely different week. Include work mix where it changes the capacity required.
Look for the recurring miss. Find the business input that was absent or late. Ask the provider what earlier, more reliable information would let it do differently.
The Forecast-to-Rate Review with this edition gives you a short set of questions for that conversation. It should end with one change your side will make, one response from the provider and a way to check whether the service or its cost improved.
At ProcureCon, I kept coming back to who owns the outcome. Here is a very practical test: when our forecast makes the service more expensive, who can connect the business decision, the vendor’s response and the bill?
Until we can do that, we will keep negotiating rates that include the cost of our own uncertainty.