If the board doesn't trust the pipeline or the revenue figures, the fix is usually in the data underneath. I set up how your HubSpot data fits together around how your business sells and reports, so the numbers are right and HubSpot's AI has something useful to read.
Data
Objects, properties and associations designed around how you actually sell, then the duplicates and gaps cleaned out.
Data and CRM architectureIntegrate
ERP, help desk and delivery tools talking to HubSpot, so nobody rekeys the same customer twice.
HubSpot integrationsAutomate
Approvals, renewals and handoffs that run on their own, without anyone chasing.
HubSpot implementationApps
Custom cards, app pages and portals, so your team works in one place instead of five tabs.
Custom apps and cardsAI
HubSpot's AI and agents working on data they can trust, plus Claude and ChatGPT set up for your team.
HubSpot AI readinessHubSpot CRM architecture is how your data is set up: which records you keep, how they link together and which fields exist, so your reports add up. I'm an independent HubSpot Gold Solutions Partner. I design custom objects, clean and deduplicate existing data, and set simple rules so it stays clean.
When a board stops trusting the pipeline, the report is usually doing exactly what it was told. The problem is the data it's reading. The deal amount is actually a three-year total. The close date is being used as the invoice date. One customer with twelve sites is stored as twelve unrelated companies.
A new dashboard on top of that shows the same wrong numbers. The fix is in the data model, which on most portals grew as the business changed rather than being designed. I work data first, then automation, then reporting, because each one depends on the layer underneath.
A rough rule I use: if a thing can appear many times per company, has its own dates and status, and someone wants to report on it, it's an object. If there's one value per record, it's a property.
The giveaway is numbered fields. "Site 1 address, Site 2 address." "Test 1 date, Test 2 date." That stops working the moment you need a sixth one, a history, or a report across all of them.
Custom objects need Enterprise. Before you upgrade for them, I'll check whether something you already have would do. Tickets, a separate deal pipeline or line items sometimes cover it at no extra cost.
A UK manufacturer, one of the largest in its field worldwide, tests the products it installs at customer sites every year. HubSpot had been set up when the programme was much smaller, so each product held its results in a fixed set of fields. The history couldn't grow and there was no forward schedule.
I designed a model of company, then installed product, then a test record of its own, and moved every past result into it.
The team now reports, schedules and produces customer reports straight from HubSpot, in about 30 seconds each, and has two weeks of work a month back. There's more in the case study.
An international technology firm was moving its support desk from Zendesk into HubSpot. Both systems mixed customers and their individual sites together, so moving the tickets as they were would have carried the problem across. The data model was agreed first, with customers, venues and assets as their own records, and every uncertain match listed for the client to decide.
A clean-up fixes what's there today. Keeping it clean is mostly HubSpot settings you already have:
I use HubSpot's own duplicate tools where they're enough, and my own scripts where they aren't: find likely matches, produce a review workbook, then merge only the rows you've approved. Merges in HubSpot can't be undone, so nothing is merged without that review.
| Symptom | Usual cause | Fix |
|---|---|---|
| Pipeline value doesn't match finance | Deal amount holds a multi-year total, or close date is used as invoice date | Line items with billing start date and term, reported by billing schedule |
| One customer shows up as many companies | Sites or branches stored as separate companies | Parent and child companies, or a site custom object |
| Duplicates return after every import | Matching on name rather than domain or ID | Fix the import or sync rule, then merge |
| Reports can't be built across records | Data held in numbered properties | A custom object with proper associations |
On most portals this just means a named owner, short naming conventions, a property list with descriptions, and changes tested in a sandbox first. I write it all down in plain English and hand it over, so the rules still get followed after I've gone. The training and adoption page covers that handover.
Contacts, companies and deals. Bookings live somewhere else.
Bookings, customers, courses, trainers and their expenses, each with the properties the team actually uses.
Every booking knows its customer, its course, its trainer and what it cost to deliver.
Customer codes from the back office, invoices from the ERP, sessions from the virtual classroom.
Customers due a refresher, bookings not yet invoiced, pre-paid balances still to use.
Booked, delivered and invoiced in one place, with no separate booking platform.
Sales, operations and finance, all working from the same live data.
Illustration with fictional data.
It's a record type you define yourself, alongside contacts, companies, deals and tickets, with its own properties, links and reports. Custom objects need an Enterprise subscription on at least one Hub.
Usually, yes. I find the problems, propose fixes in a workbook you can review, then apply only what you approve. Starting again is rarely the right answer.
Mostly imports and integrations. A sync that matches on company name instead of website domain will create a new company every time a name is spelled differently. I fix the source as well as the records.
Not if it's done carefully. Every merge is a dry run first, with a review list, and I only merge what you've approved. HubSpot merges can't be undone, so that review step matters.
Yes. A lot of my work is on Enterprise portals, including custom objects holding tens of thousands of records, with every change tested in a sandbox before it reaches the live portal.
Not sure where the data problems start? A fixed-price review finds them.
Two weeks of work back every month, with over 80,000 product tests in one history.
What Breeze and HubSpot's agents need from your data before they're worth switching on.
Book a short call. Tell me what the reports say and what you think they should say, and I'll tell you where I'd look first.