A computer can price a deal in a few milliseconds. In a lot of companies a quote still takes days.
Those days go into waiting. A rep waits for the page to recalculate. The quote sits in an approver’s queue. Then somebody retypes the numbers into the contract system, and later into billing.
I led quote-to-cash on Salesforce CPQ at LinkedIn, though very little here depends on Salesforce. Most of it comes down to two questions: where the deal’s facts live, and how many times a person has to touch them on the way to an invoice.
Six stages, five handoffs
Quote-to-cash is the path from “the customer wants a price” to “the money is in the bank”. It usually crosses sales, finance, and legal, and three or four systems along the way.
- A rep builds a quote from the product catalog: products, quantities, terms.
- Pricing rules turn that configuration into numbers: list price, volume tiers, discounts, proration.
- Anything outside policy goes to approvers.
- The accepted quote becomes a signed contract.
- The contract becomes an order that provisioning and finance can act on.
- Billing invoices the customer, and finance recognizes the revenue.
I care less about the stages than about the handoffs between them. Each handoff is a place where data gets retyped, reformatted, or waited on.
The quote owns the numbers
The rule I’d hold every design to: the quote is the single source of the deal’s commercial facts. Every later stage reads from it. Nothing downstream re-enters it.
The opportunity is usually where this breaks first. Forecasts live on the opportunity, so people want to edit the amount there.
Don’t let them. Derive the opportunity’s amount and forecast from the primary quote and make those fields read-only. Two editable copies of the same number will disagree by the end of the quarter, and someone in finance will get to reconcile them by hand.
Past the opportunity, hand off with events. When a quote is accepted, publish that fact once. Contracting, order management, and billing each subscribe and react on their own. Nobody polls and nobody retypes. If you’ve got this right, finance can trace any invoice line back to the quote line it came from. That’s the check I’d use to know the plumbing works.
Pricing is a function, so treat it like one
Given a quote, a set of rules and a pricing date (plus exchange rates, if you sell in more than one currency), the price should come out the same every time, with no side effects. Once pricing behaves like a pure function you can test it with ordinary unit tests. Better still, you can rerun last quarter’s quotes, at their original pricing dates, against a proposed rule change and see exactly which deals move before you ship it. I wouldn’t change pricing without that.
The other half is where the function runs. Recalculating on the server after every edit is the usual reason quoting feels sluggish. Run the same pricing code next to the rep, in the page they’re typing into, and keep the server as the final check before anything is saved. The spinner goes away, and the rep can give a correct price while the customer is still on the call. If the browser and the server use two implementations, they’ll drift, and the final check will start rejecting prices the rep already quoted.
Catalogs grow by accident
Every special deal wants to become a new product.
A typical case: a large customer negotiates a bundle with a custom support tier, and the fastest way to quote it is a new SKU. Do that for every large customer and nobody can find anything in the catalog. Worse, the pricing rules get quietly copied per product, and they drift apart.
Keep the product list small and model variation as attributes and rules on those products. Then put a catalog review on the calendar, because left alone it will grow back.
Approvals belong in a table
Approval queues are where quotes tend to sit, and hard-coded approval chains make every policy change a deployment. Put thresholds and approvers in a table that finance owns and can edit themselves.
A few rules on top of that go a long way:
- Standard deals inside policy skip approval entirely. If most quotes need a sign-off, fix the policy before blaming the reps.
- Independent approvals run in parallel. Legal shouldn’t wait for the deal desk to finish.
- When a quote changes after approval, re-trigger only the approvals the change actually touches. Updating the billing contact shouldn’t send a discount back to a VP.
Amendments are the real test
A new sale is the easy case, and it’s the one every demo shows. The hard cases are the mid-term upgrade, the co-termed add-on, and the renewal that comes with a price change. In a mature business they make up most of the volume.
If the data model only really works for a fresh deal, you’ll find out the first time someone adds seats halfway through a contract. I’d design for amendments from the start and treat the new sale as the special case.
Time it before you fix it
Before changing anything, measure how long each stage takes on its own. The delay tends to be an approval sitting in a queue, or a calculation that reruns far more often than it needs to. The system everyone complains about often turns out to be fine.
If you want to try this on your own process, start small. Take ten recent deals and put a timestamp on every stage. The slowest stage is almost never the one people blame.