Data
The number was wrong. The data was fine.
When a pipeline figure gets challenged in a leadership meeting, the reflex is to blame the data and book a clean-up. Often the records are fine and the number is built wrong, for structural reasons nobody looked for.
You know the moment. A figure goes up on the screen in the leadership meeting, somebody says it can't be right, and the next 20 minutes are about Salesforce instead of the business. Afterwards the verdict is nearly always the same: the data is bad, the reps haven't updated their deals, someone needs to send another reminder or book a clean-up.
Sometimes that's true. But in a single week recently I had four calls about numbers that didn't look right, and in none of them was another nag to the sales team the answer. In several, the records themselves were fine. The number was still wrong, for boring, structural reasons nobody had gone looking for, because everyone assumed the problem was the people doing the typing.
So here's my argument.
Before you clean the data, find out how the number is built.
The data gets the blame because it's the easy suspect
Blaming data quality feels responsible. It points at something visible and comes with an obvious action: chase the team, run a clean-up, add another required field.
But a report is more than the records inside it. It's a stack of choices: which field it counts, which filter it applies, which view it's shown in, what it lumps together. Someone made every one of those choices, usually a while ago, usually for a good reason at the time. When a number goes wrong, that's where I look first, and it's rarely where anyone else has looked. Three of them came up that week.
Why is my Salesforce report wrong when the data looks right?
1. The report is keyed to a field that moves
One team asked me to build KPI reports showing, quarter by quarter, who had created new opportunities. The original dashboard credited each deal to its owner. That's fine in the current quarter, because whoever finds a deal usually owns it for a while.
Then the deal is handed to an account manager. The owner changes, and the deal drops off the originator's history. Nothing in the record is wrong. Every historic view on that dashboard is, because it counts a field that's designed to change.
My first thought was "created by" instead. Then I found one person had been creating opportunities on behalf of a colleague, because he was quicker in Salesforce and was saving her the time. Kind, sensible, and it breaks "created by" as a measure of who did the work. The answer was to key the reports to the field that records who actually sourced the deal, which nobody reassigns. A small change, and the history stays put.
The same week, a leader in a different team found his pipeline lighter than he knew it should be. A large deal sat under a colleague's name, so his view didn't count it. The real question was never "is the data right?" It was "do you want deals in your name, your team's name, or against the company they belong to?" Three different numbers, and only one answers what he was asking.
2. The view has a limit nobody knew about
A manager called me because his team's total didn't match the sum of his individual reps. He hadn't touched anything. The data hadn't changed. Add up each person's deals in a stage, compare with the team figure, and the two disagreed.
He'd built himself a Kanban view of his team's opportunities, which is a perfectly reasonable way to look at a pipeline. What he didn't know is that the Kanban view shows up to 200 records. His team had more than that, so the board was leaving deals out of its totals. Same data, different list view, right answer.
"This is a Salesforce quirk," I told him, and it is, sometimes even the best designs fall victim to the limits of the system they are built on. He built something sensible with the tools in front of him, and nobody had told him where they stop. That isn't a data problem or a people problem. It's a knowledge gap, and a cheap one to close once someone spots it.
3. Two different things are landing in one bucket
The biggest number of the week was a pipeline figure roughly ten times what it had been when I started working with that business. Nobody in the room believed it, and they were right not to.
Two things had happened. The opportunity list was doing two jobs: it held the deals the team was genuinely working, and hundreds of unqualified records that were really a lead list. The team had more than 400 between them. Nobody can see the wood for the trees in a list like that, least of all the person checking what's being prioritised. And nobody had checked for duplicates, so some deals had simply been entered twice.
The fix isn't a clean-up. It's a decision: an explicit Unqualified stage, so what's being worked is visible and what isn't is still there for marketing. It's the same reason less data, kept well, beats more: a list full of things nobody is working hides the things they are. The team also settled a question that had hung over the system for a long time. As I put it on the call: "It was always like a grey area of, do we use leads? Do we not use leads? And now, we're not using leads." Once that's decided, the bucket means one thing again.
Two cousins of this turned up elsewhere. In one business, a whole revenue line, recurring fees that make up a meaningful share of margin, wasn't in Salesforce at all. Nobody was careless; nobody had ever built a place for it, so the board number under-counted it. A place is only half the job, though. As I said at the time: "We can set Salesforce up so you've got a place to log it. You've still then got to go in and log it."
In another, a report flagging low-margin jobs was flagging jobs that were fine. The costs of a change to the job had gone into one field instead of being split across two. Correct total, wrong field. The figures were keyed in by hand, and as I said, "then you're just relying on someone not having fat fingers." A completeness rule wouldn't catch it either, because the field was filled in, which is why your CRM has always measured the wrong thing. That's a process design point, not a nagging point.
A disputed number costs confidence, not just time
When a number gets challenged in front of the leadership team, the cost isn't the hour it takes to find the cause. It's confidence. The next report gets questioned too. People start keeping their own spreadsheets, just to check. Before long, the system the whole sales operation runs on is something everyone works around.
And the usual response makes it worse. Nag the reps when the records were right, and you've told them their effort doesn't count. Pay for a clean-up when the report was built on the wrong field, and you'll get clean data producing the same wrong number. Point an AI tool at it, and you'll get the wrong number faster and with more conviction. If that's where you're heading, decide where the truth lives first.
Every fix here was small: a different field, a different list view, a new stage, a home for revenue that had nowhere to go. That won't always be true. Some causes run deep, into integrations and how data moves between systems, and those take real work. But you can't tell which kind you've got until someone has looked at how the number is built.
Three questions for the last number you argued about
Take the figure that caused the most trouble in your last leadership meeting, and ask:
- What field is it keyed to, and can that field change? Owner and team can both change as a deal progresses. If the report counts something that moves, its history will drift.
- What view or report produced it, and what does that view leave out? A filter, the person it runs as, a display limit, a list that only shows your own records.
- What two things might be sharing one bucket? Leads and opportunities. New business and recurring revenue. Two kinds of cost in one field.
If you can answer all three with confidence, the meeting can get back to the business, which is where it belonged. If you can't, you've just found the most useful thing you'll learn about your Salesforce this quarter.
In the meetings I sit in, a challenged number is rarely a verdict on your team. Far more often, it's a verdict on a decision someone made about a field, a view or a list years ago, and never wrote down.
That's exactly what a Salesforce Reset looks at. Data discipline is one of its three areas, alongside process and friction, and adoption and ownership, and it starts with how your numbers are built, not how clean the records look. Within 14 days you get written findings: what we found, what it's costing you, and what to fix in what order. Book a Salesforce Reset.
Scott Reynolds
Founder, SaaS Success. Fourteen years at the intersection of sales process and CRM; nine Salesforce certifications, two at architect level. Scott writes the way he consults: plain English, working shown.
Get the next one when it's written.
No cadence promises. Just the next piece, when there's something worth saying.

Reading about it won't fix yours.
When the researching is done, start with the Reset: a full review of your org, a written answer within 14 days, £1,000, refundable for fourteen days after your findings arrive.
Book a Reset