When people stop trusting the numbers, they build their own.

Private spreadsheets. Fixes nobody else knows about. Executives re-checking every number before the board meeting. I find out why your organization stopped trusting its data, teach your team the habits that rebuild that trust, and then I leave. It takes one to two weeks, for a fixed fee.

Tell me what isn’t adding up Take the one-minute self-check
Board deck
$48.2M
Dashboard
$47.9M
Finance’s spreadsheet
$48.6M

Three numbers. One metric. No owner.

AI doesn’t quietly catch bad numbers. It confidently repeats them.

Until recently, a bad number meant an awkward moment in a meeting, and someone usually caught it before it did real damage. AI tools don’t work that way. They read whatever data is lying around and act on it instantly, at scale. They never see the corrections your people made in their own spreadsheets, so they use the version everyone already knew was wrong.

If your people don’t trust the numbers, your AI shouldn’t either.

It’s almost* never the technology. It’s the people and the process.

So the fix isn’t another tool. When people stop trusting the numbers, they work around them, and your data team spends its week defending last week’s work. In twenty-five years, I’ve seen it come down to the same three gaps almost every time. Banks and other regulated industries were forced to close all three years ago, with habits, not software. Most companies never were.

The gapThe habit that closes it
Nobody owns the definitions, so every department has its own formula for “revenue.”
Every metric gets a unique name, a written definition, and a named owner.
Fixes happen quietly, in someone’s private copy, and never make it back.
Definitions are centrally stored and transparently communicated.
Clients find the incorrect numbers before your team does.
People are rewarded for raising problems loudly, early, and often.

Because these are habits, they can be fixed in weeks, without buying anything.

The research agrees.

67%

of organizations don’t fully trust the data they make decisions with.1

71%

of organizations with formal data governance trust their data. Without it, about half do.2

*I almost never make sweeping generalizations. Given enough time, “never” never happens.

1Source: BlastX

2Source: Quirks, comparing organizations with and without a formal data strategy and governance

Do you actually believe your numbers?

Everyone says their business and data teams talk to each other. These six statements are harder to wave away. Check the ones that are true where you work.

0 of 6

Two times trust broke, and what brought it back.

The business didn’t trust the data team.

A data team I led was doing everything right by its own measure: fast delivery, bugs fixed before new features. The people who used its reports saw it differently. They never knew when a number might change, so they quietly saved their own snapshots to keep this month’s report consistent with last month’s.

One analyst adopted two habits. Before releasing a change, he showed it to the people who would use it. And he checked that everything he didn’t mean to change had stayed the same. Within a month, those same people were praising his work. Within six, executives stopped double-checking the team’s numbers before board meetings, and that held after I left.

The analysts didn’t trust the automation.

Every Monday at 4:30 a.m., six analysts at an ad agency hand-copied 53 numbers from an analytics tool into Excel, then into PowerPoint. An automated system could already do it. Nobody trusted it, because each analyst had been quietly fixing definitions in their own account and workbook, and the system never got those fixes.

We set up one shared place for metric definitions and a review process for changing them. That made automation possible. Along the way, we found thirty to forty historical numbers that were simply wrong: copy-paste mistakes and typos nobody had caught. If someone promises you a cleanup with no ugly surprises, they aren’t looking hard enough.

Same lesson both times. The technology was fine. Smart, well-meaning people were working around a process nobody had designed.

Designed to end.

A fixed fee, a fixed end date, and nothing to sell you afterward. Most engagements take one to two weeks. Your people probably* aren’t the problem; nobody ever showed them what good looks like. That’s fixable without firing anyone.

  1. Map Optional

    You run a small script on your calendar and messaging metadata: who meets and messages whom, never what they say. It shows who actually carries information, which is rarely who the org chart says.

  2. Interview

    I interview the people the map surfaces, or the people you point me to.

  3. Diagnose

    You get a written diagnosis. Every finding ties to a specific number and a specific conversation, or it stays out.

  4. Train

    I teach your team the habits that keep trust once it’s back: clear definitions, named owners, change control, and review.

  5. Leave

    I leave while things are working. No retainer, nothing to renew. Your team keeps the habits.

Why me

Hi, I’m Jeff. For twenty-five years, companies have hired me to fix the same problem: distrust and miscommunication where data, technology, people, and process meet. Much of that work was in places where one small error could compound into millions gone, or lives lost.

In every one of those jobs, my biggest impact was felt early, either turning things around or building them right from the get-go. So that’s the part I offer now.

Twenty-five years in, I’m not here to manage your feelings. I’m here to tell you what I see.

A consultant who needs you to like the answer will soften it. I don’t need you to like it.

Prefer the unfiltered version? More on LinkedIn

Hard questions

The ones people ask once they’re serious, answered straight.

How much does it cost?

One fixed fee, quoted up front after a first conversation and based on the size of your organization. It’s lower if you run the optional map, because then I interview only the handful of people who matter instead of finding them the slow way. No hourly billing, and nothing to renew.

What if we’re actually fine?

If you thought you were fine, we wouldn’t be talking. If it’s the technology, I’ll say so. If it’s people or process, it’ll show up in the report.

And if everything checks out but you still can’t trust the numbers, the work is about transparency and how your numbers get defined. I’ll coach you through that.

How do we know you won’t invent problems to sell us more?

There’s nothing to sell. The fee is fixed, and I leave when I’m done. No software to resell, no methodology to license. Every finding comes with its evidence, or it doesn’t go in the report.

I’m here to figure out what’s actually going on and teach your people to keep it from happening again.

What if it really is just the technology?

Then I’ll tell you, and help your team find a better option. But I won’t stop there, because the next question is why you don’t have it already.

If the right tool is free and nobody uses it, that’s an adoption problem, and you probably need an internal champion. If it’s expensive and the budget never gets approved, the organization has decided that bad numbers cost less than good software. Either way, that tells us what it will take to change.

Why not let our team figure it out eventually?

If they could and haven’t, they don’t have the time, budget, or cover. If they have all that and still haven’t, they may be afraid to tell you the truth. If they can’t, nobody showed them what good looks like, which is exactly what I do.

The real question isn’t whether your team could get there eventually. It’s how long you can afford to make decisions with numbers you don’t trust.

What does this do to our data team?

If you don’t trust their work, morale is already suffering. I’m naming the problem so it gets fixed. Usually I find they were set up to fail by a process nobody designed, and hearing that out loud is a relief.

What if the problem is the CEO?

Yep. That happens. If you’re afraid of what I might find, don’t hire me.

What if we disagree with the diagnosis?

Then I leave, which was always the plan. You’re paying for evidence, not agreement, and every finding shows its work.

Why would we let a script near our calendar and Slack?

You don’t have to; the map is optional. If you do run it, it reads metadata only (who meets and messages whom), never content. You run it yourself, you see the output first, and your IT or legal team can review it before anything runs.

Tell me what isn’t adding up.

Describe one specific time a number wasn’t trusted. A sentence or two is plenty. I read and answer every message myself.

Which number was it, and what happened? A sentence or two is plenty.

Has anyone told the person who owns that number, directly, that it was wrong?

Either answer is fine. If it’s “not yet,” that’s already a finding.

I’ll use this only to reply to you. No mailing list, no sharing.

What happens next

  1. I read your message myself.
  2. If it sounds like something I can help with, we set up a first conversation.
  3. If it doesn’t, I’ll tell you that too.

Rather skip the form?

Former colleague? Just send me a note.

Turn on blunt mode?

Same advice, said the way I’d say it in person: fewer cushions, and some swearing. You can switch back any time from the top of the page.