How we work
There's one thing holding you back. Not nine.
Every manufacturer we meet is working three or four problems at once, and usually none of them is the one that matters.
The rule
You can't fix a stage until the one before it holds.
If you've read The Goal, you know the idea. At any moment one thing limits the whole system, and effort spent anywhere else doesn't move the output. Your office works the same way, which is how a shop ends up buying a new system when the real problem is that the two it already owns don't agree.
A close rate you can't explain isn't a sales problem if nobody agrees what a qualified lead is. A dashboard is decoration if three systems feed it three different numbers. And no AI is going to help if the records were never clean at the source.
So we start where the record is born and work the first thing that's actually broken. It's usually cheaper than what you were about to buy, and it makes everything above it possible.
Then we do it again. Every new product, customer, employee and tool breaks something you rely on, so the constraint keeps moving. That's the work, and it's why this is a standing engagement rather than a project with an end date.
The Information Pipeline Map
The nine constraints we see most often in small manufacturers.
Almost every shop we talk to has several of them, and one is doing most of the damage. Follow the path: each tier has to hold before the next one means anything, so the constraint worth fixing is always the earliest one that is broken.
Capture
Does the data get born, and get born right?
Start here
- 1Never recorded.
It happened, and nothing anywhere knows it did.
- 2No shared definitions.
Two people count different things and use the same word for both.
- 3Error-prone entry.
The record is created wrong, and everything downstream inherits it.
Connect
Does it move between systems without a person carrying it?
- 4Dead handoffs.
Someone retypes the same record into two or three systems.
- 5Shadow spreadsheets.
The real number lives in a spreadsheet beside the system that's supposed to hold it.
- 6One person is the system.
It works because of who they are, and it stops when they're out for two weeks.
Decide
Does anyone decide anything from it?
- 7Hand-built, too-late answers.
Every report is a project. Month-end takes three weeks.
- 8No cadence.
Nobody looks at the numbers on a schedule, so nothing gets decided from them.
- 9Starving tools.
Dashboards nobody opens, analytics nobody uses, and AI that gives generic answers.
“Every system gives me a different number” isn't on this map, on purpose.
Three of these produce that sentence. Which one it is, is the diagnosis.
From diagnosis to work
The map decides the order. Not a wish list.
What the work costs| Stage | Where it sits on the map |
|---|---|
| The first win | Usually a dead handoff (#4), because it is the fastest thing to count and it hands somebody hours back. That is the quick result, not a guess at what is binding. |
| The first 90 days | Capture through Connect for the thread that matters most, then stand up #7 so the answers stop being hand-built. |
| Then, continuously | #8 becomes permanent, and we keep working back down the map. One constraint clears, the next one is picked by the numbers. |
| The AI work | #9, once one through six are true enough to feed it. |
What a useful AI job looks like here.
A named job, on a named system, where somebody gets time back. Never a pilot, and never a strategy.
Quote intake
An agent that drafts a new quote from your last two hundred, so the estimator edits instead of starting from a blank page.
Exception watch
An alert the first time a job crosses its estimate, rather than at month-end when it is somebody's variance to explain.
Asking your own data
Questions answered against your actual business, because by then the systems underneath it agree with each other.
Working together
What it actually looks like.
| The question | The answer |
|---|---|
| How often do we meet? | A standing review on a schedule you pick, plus a quarterly session where we look at the numbers together and decide what's next. |
| What access do you need? | Read access first, always, to whatever holds the numbers we're connecting. Writes are scoped, deliberate, and agreed before we make any. |
| How long before I see something? | The first result lands in month one. The full instrument is live and trustworthy by day 90. |
| How much of my team's time? | Almost none to start. We give time back before we ask for any. |
| What if we're slammed that month? | Most of what we do doesn't need your people. A quiet month still moves. |
Will you work with what we already run?
Usually yes. It depends on what the system will let us read. Here's what we've worked in.
| Group | Systems |
|---|---|
| Storefront and product data | Shopify · Matrixify · Shogun (migration off it) · WooCommerce · BigCommerce |
| Feeds and merchandising | DataFeedWatch · GoDataFeed · Google Merchant Center |
| Tracking and measurement | GA4 · Google Tag Manager, including server-side · Meta CAPI · Microsoft Clarity |
| CRM | GoHighLevel · HubSpot · Pipedrive |
| Accounting | QuickBooks |
| Data pipelines and storage | Google Sheets · Google Apps Script · Supabase |
| Dashboards and reporting | Looker Studio · Databox · Zoho dashboards · custom dashboards we build and host |
| Automation | Zapier · Make.com |
| Ops and delivery | ClickUp · Monday · Asana · Harvest · Slack · Fathom · Fireflies |
This isn't exhaustive. If you run something that isn't on it, we can usually still help. Sometimes there's a prebuilt integration, sometimes we write the connector ourselves. Tell us the problem and we'll work out the fix.
Measuring it
Six questions on day one. The same six on day 90.
On day one we write down the six questions you most need answered about your own business, whether you can answer them today, and how long each one takes. Month-end included. On day 90 we ask the identical questions and write the answers next to the originals.
More of them answerable, each one faster. Measured rather than claimed.
| Can you answer it? | Day 1 | Day 90 |
|---|---|---|
| Win rate on quotes | No | Yes · 2 min |
| Margin on a finished job | 3 days | Same day |
| Which customers reordered | No | Yes · 5 min |
| What the pipeline is worth | Half a day | Live |
| Month-end close | 3 weeks | 4 days |
| What counts as a qualified lead | 3 answers | One, in writing |
Getting started
How we get started.
- 1
A 25-minute call.
Five questions about how your systems work today. No access to anything, nothing to prepare.
- 2
The Systems Snapshot.
A short form: where each function lives today, what gets retyped and by whom, and which of your own numbers you can produce. Unknown is a real answer, and writing it down is the point. It's also your day-0 baseline, and we re-run it at day 90 to show the change.
- 3
The systems audit.
The form tells us what you believe. Getting into the systems tells us what's actually true, and those are frequently different. We'll say if it isn't worth doing. If it is, you leave with a 90-day plan: implement it yourself, or the fee credits in full against the build.
If it's not a fit, you'll still leave with a clear answer and something you can act on.
Start here
Start with the call.
Twenty-five minutes, five questions, nothing to prepare.