Introducing Curve Analyst: Privacy-Safe AI Chat for Healthcare Marketing

We spent the past year building the tracking and attribution layer healthcare marketing never had. Today we are shipping the part we actually wanted to build all along. Curve Analyst is a chat window inside Curve. You type a question about your marketing, in whatever words you would use to a colleague, and it answers from the data already sitting in your account.

Which campaign is filling consults this month? Where do people leave the booking funnel? How is AI search traffic looking? You ask, it queries your account, you get a number and a chart.

Why we built it

Most marketers outside healthcare solved their reporting problem two years ago. They paste an export into ChatGPT and ask it what changed. If you market a clinic, a telehealth service, or a med spa, you cannot do that. The export has page paths, campaign names, and conversion events tied to specific health services, and the tool on the other end has never signed a BAA. Your privacy officer would call that a disclosure. So would a plaintiff's lawyer.

So healthcare teams kept doing it the slow way. Google Ads in one tab, Meta in another, web analytics in a third, and a spreadsheet to argue with. A simple weekly answer to "what is working" takes an afternoon, and half the afternoon is reconciling numbers that were never going to agree.

We were already holding all of that data. Curve receives the events, does the attribution, and pulls campaign performance from the ad platforms into one report. The obvious move was to put a model on top of it, inside the same walls. That is what Curve Analyst is.

What you can ask

Analyst reads from three parts of your account.

Website analytics

Visitors, sessions, and pageviews over any period. Top pages. Traffic sources and channels, including referrals from ChatGPT and Perplexity, which are showing up in more healthcare accounts every month. UTM breakdowns. Devices, browsers, locations, custom events, and who is on the site right now.

• How did visitors trend over the last 30 days compared with the 30 before that?
• How is AI search traffic looking this month?
• What were my top sources this quarter, and which one converted best?
• Which pages does paid traffic land on most?

Goals and funnels

Every conversion goal you have set up in Curve, how many times it fired, and how that compares with the period before. Funnels step by step, so you can see where a visitor stops between the landing page and the booked appointment.

• Which goals converted the most this month?
• Walk me through the booking funnel and tell me which step loses the most people.
• Did form submissions drop after we changed the landing page?

Campaign reporting

Spend, conversions, cost per acquisition, ROAS, clicks, impressions, and click-through rate by platform, campaign, or ad group. Spend and conversion trends over time. And a reconciliation view for the question every healthcare marketer eventually asks: what Curve sent to Google or Meta, next to what the platform decided to credit.

• Which campaign is performing best this month?
• Is Google search cheaper than Meta for booked consults?
• Is spend up versus last month, and did conversions follow?
• Why does Google Ads show more conversions than Curve recorded?

You get charts, and it knows which screen you came from

When an answer has a shape, the dashboard draws it. Trends come back as line charts, ranked lists as tables, totals as stat cards, funnels as step charts. Ask for a graph and you get one next to the explanation.

Open Analyst from a report or an analytics view and it keeps the date range and filters you already had applied. "What drove the change here" means the screen you were just looking at, with the same filters. The suggested questions change to match. This sounds like a small thing. In testing it was the difference between people using it on day one and people forgetting it existed.

It will tell you when it does not know

The thing we worried about most was an assistant that confidently makes up a conversion number. A wrong number in a chat window looks exactly as authoritative as a right one. So Analyst has one hard rule: every figure in an answer comes from a query against your account. It does not estimate and it does not fill gaps.

If a metric comes back empty, it says so. If a data source is stale or still syncing, it says that before it says the number. It treats "unknown" and "zero" as different things, because telling a practice it got zero conversions when the data simply is not in yet is the worst thing a reporting tool can do. And for ad platform figures it says which view it is quoting, Curve's attributed conversions or the platform's own count. Those two numbers almost never match. Analyst explains the gap rather than picking one and hoping you do not notice.

Where the data goes (nowhere)

Analyst runs on Claude through Amazon Bedrock, inside infrastructure covered by our Business Associate Agreement with AWS. Curve signs a BAA with every customer, on every plan. The model reads the same US-hosted analytics and reporting layer the rest of your dashboard reads. Your data is not shipped to a separate AI product and is not used to train anything.

We also kept it narrow on purpose. Analyst is read-only. It cannot change tracking, edit a destination, or take any action in your account. It will not look up an individual or answer questions about a specific person. It answers about audiences, campaigns, pages, and funnels, which is what a marketing decision needs. We think that restraint is what makes it usable in healthcare at all, and as far as we know it makes Curve the first HIPAA-compliant marketing platform where you can point a current AI model at your performance data without creating a compliance problem in the process.

How to start

Analyst is rolling out to Curve customers now. Log in, open Curve Analyst in the navigation, and either pick one of the suggested questions or type your own. The best first question is whichever one you would otherwise spend an hour building a report to answer. For most of our customers that is one of these:

• Where did last month's booked patients come from?
• Which campaign is filling consults, and what does each one cost?
• Where are people leaving before they book?

Not on Curve yet? Book a demo and bring the reporting question that annoys you most. We will answer it live on real data.

Frequently asked questions

Does Curve Analyst send my data to an AI company?

No. It queries the analytics and campaign reporting already stored in Curve and runs the model through Amazon Bedrock under Curve's BAA with AWS. Nothing is exported to a separate AI product and nothing is used for training.

Is it included in my plan?

It is being switched on for Curve accounts as part of the rollout. If you do not see it in your dashboard yet, ask your Curve account manager or reach out to the team.

Can it change my tracking?

No. It can explain your goals, funnels, and campaign performance, but every configuration change still happens in the dashboard, by a person.

Why do its ad numbers differ from Google Ads or Meta?

Curve reports its own attributed conversions, credited by Curve's attribution engine using the model you chose. Google and Meta count with their own windows and their own modelling. Analyst tells you which figure it is quoting and can show both side by side.

What is next?

Questions about tracking setup and connector health, deeper cross-platform reconciliation, and ways to act on what Analyst finds. If there is a question you wish it answered today, tell us. We are deciding the next few months based on what customers ask it in the first one.

Curve is the HIPAA-compliant tracking, attribution, and analytics platform for healthcare. Book a demo to see Curve Analyst on your own data.