How HouseFacts built a data system they trust, from day one, with Convex Works

The company - Applying AI to help homeowners manage their homes

HouseFacts is using AI to help homeowners better manage their home. You can’t keep track of all the information in your home by yourself, and it’s just way too much.

So HouseFacts takes all that unstructured information, structures it, and helps you with the next decision, saving you time and money: appliance tracking, service history, inventory, cost basis, all enabled through AI for the homeowner.

It surfaces all that proactively, because otherwise people don’t know, and the reactive problem is catastrophic: a six-hundred-dollar job you miss can turn into a thirty-thousand-dollar one.

What differentiates HouseFacts is that it serves the homeowner, and the data is private to the homeowner. It does not sell the homeowner’s data. The reason you pay for a subscription is because the homeowner is the client, and not the product.

The challenges

1. Consent from day one
   HouseFacts really cares about data privacy, so there was the need to handle consent and collection correctly from the beginning, while the site still had almost no traffic.

2. Data lost in the browser
   HouseFacts collected its data through the browser, where much of it was lost before it arrived. Ad blockers removed the analytics scripts, Safari and iOS restricted them, later purchases could not be connected to the visit that caused them, and each tool kept its own partial copy. Because the reports still looked complete, the loss stayed invisible, and the advertising platforms were bidding money on incomplete signals.

3. Manual and automatic decisions
   HouseFacts needed its data to support both manual and automatic decisions. The team needed quality dashboards to make product and marketing decisions. The advertising platforms and the notification infrastructure needed the same data to decide who to bid on, which ads to spend money on, and which notifications to send and to whom.

That’s why we want to work with somebody like you from the beginning of making sure that we’re getting consents, making sure that the data is private, making sure that the information is getting piped appropriately. And we wanted to work with somebody who had experience with, say, GDPR, because we want to be thinking about that up front, as opposed to having to be reactive about that.

The solution

We built HouseFacts one connected data system, from consent through to the tools that use the data. Each part solved one problem:

• Data compliance
  We collect data only when the homeowner has given consent. Consent is handled server-side through iubenda and Cloudflare Zaraz, and it was in place before the site had real traffic.

• Data collection from all sources
  We hold one consistent record of each homeowner, from the first anonymous visit through to signup and product use. It is collected server-side and first-party through Cloudflare Zaraz, and Segment keeps the events and identifiers consistent across every tool.

Anyone on the team needs to be able to open the numbers and know they are real, without me as the filter. The data has to be trustworthy on its own.

• Data loss detection
  We measured what each tool actually received and found that ad blockers and browser limits were dropping a large part of it. Collecting server-side instead of in the browser recovered most of it and made the site lighter.

• Data ownership
  We keep a copy of everything in HouseFacts’ own BigQuery warehouse. The record does not live only inside each separate tool.

• Data distribution
  The same data reaches every system that acts on it. Google and Meta bid on real conversions, confirmed on the backend and counted once. ActiveCampaign gives each homeowner the follow-up email that matches how they arrived, and re-engages those who signed up but have not added a property.

I think we could have gotten away with some of the internal dashboarding, but then obviously I have to maintain it, and you don’t have the same granularity all the time. There’s a lot of value in being able to show it with other people.

Data stack

Collection
Website & forms	The site and its forms, the first-party source of every visit and signup.
Product backend	Server-side events from the product itself, so what people actually do is captured directly rather than inferred from the browser.
Server-side collection layer	First-party, server-side capture and tag firing. Because the events are collected on the server, ad blockers and browser limits cannot drop them before they arrive.
Shared event & identity layer	One consistent record of each homeowner, from the first anonymous visit through signup and product use, routed to every tool so they all read the same events and identifiers.
Consent management	Collection only fires once the homeowner has given consent, handled server-side and in place before the site had real traffic.
Cloud data warehouse	A full copy of everything in a warehouse HouseFacts owns, so the record does not live only inside each separate tool.
Distribution and activation
Product & web analytics	The team can open the numbers and trust them without a filter, across product funnels, retention, and web reporting.
Ad platforms	Server-confirmed conversions, counted once, so the platforms bid on real signups instead of the fraction a browser pixel would drop.
Lifecycle messaging	Follow-up matched to how each homeowner arrived, re-engaging those who signed up but have not added a property.
First-party proxy	A custom proxy so analytics runs first-party on housefacts.com rather than a third-party domain, keeping the data harder to block and owned entirely by HouseFacts.

The results

HouseFacts moved from scattered, partial data to one system it owns and trusts. Because this was a first build rather than the tuning of an existing setup, the result is capability more than a single number.

We onboarded a head of engineering, and he was like, ‘Well, how does this work?’ And I was like, ‘Let me provide that for you.’ It went into the config, checked every single box, and created full documentation: how it works, what the pipeline looks like, all the triggers. If this was fragmented you couldn’t do this. This is why you want it all in the same place.

The data that used to be lost is now captured, the advertising platforms work from real signups instead of partial ones, and the same record serves marketing, product, and customer follow-up.

The system also keeps growing with the company: the next step is a notification infrastructure that reaches each homeowner at the right moment, built on the first-party data already in place.