Find problems your research confirms twice

Why connections matter

One source on its own can mislead you. A heatmap shows what visitors do, but not why. A survey tells you what visitors think, but not where on the page it goes wrong. When both point to the same problem, you can be much more sure it is real.

An example: in your survey, 38 of 212 answers ask "How do I return this?". On the same product pages, the returns info sits at 92% of the page, and the scroll heatmap shows that only 11% of visitors get that far. Two independent sources, one problem: visitors want to know about returns and can't find it. That is a confirmed problem.

Because all your research lives in Varly, per site, Oran can look for these links every day. Separate tools for heatmaps, surveys and testing can't.

How strong is a finding?

Each finding shows how many independent sources back it up:

Label	Sources	What it means
Signal	1	Worth a look, but nothing else confirms it yet.
Confirmed	2	Two different kinds of research point to the same problem.
Strongly confirmed	3 or more	Three or more sources agree. Test this first.

These count as different sources: Survey (what visitors tell you), Heatmap (clicks, rage and dead clicks, and scroll depth), Click to revenue (clicks linked to orders), Recordings, Funnel and Speed & errors. Rage clicks and scroll depth are both heatmap data, so together they count as one source.

What Oran looks for

• Questions the page doesn't answer where visitors look. Oran groups survey answers into themes, such as returns, shipping costs, delivery time, sizing, price, trust and reviews, payment options, discount codes, stock, product info and contact. Varly recognizes these themes by itself in English, Dutch, German, French and Spanish. With your AI key, Oran also groups answers in any other language. Then it finds that theme in the text of the page (in those five languages), checks how far down it sits, and how many visitors scroll that far. It also notices when the info is only in the footer, hidden behind a click, or missing.
• Frustration about the same thing visitors ask about. For example rage clicks on "Shipping costs" while survey answers complain about shipping costs.
• Valuable elements few visitors reach. Visitors who click an element buy more often (click to revenue), but it sits low on the page where few visitors scroll.
• Funnel drop-offs with errors. A big drop-off after a page that has JavaScript errors or rage clicks.
• Recordings of the same problem. When recorded visits show the same frustration or errors on that page, they are added as a source.

With an AI key, Oran also sorts survey answers that don't use the usual words, for example "can I send it back if it doesn't fit?", during the daily research round.

Read a finding

Each card shows:

• the strength and the sources, as colored labels;
• the problem in one sentence, with the numbers;
• a small picture. The page strip shows the page from top to bottom: the darker a part, the more visitors see it. The dashed line is the end of the first screen, and the dark bar is where the info sits. Next to it you see a real answer from your survey and how many answers say something similar;
• each source with its number. Click it to open the survey, the heatmap or the recordings behind it;
• At stake: an estimate of the extra revenue a month if a fix lifts conversion on that page by a typical amount (4% for a signal, 7% for a confirmed and 12% for a strongly confirmed problem). It needs connected orders. It is an estimate, not a promise; the test gives you the real answer.

Turn a finding into a test idea

• Turn into test idea creates an idea with the problem, a suggested change and all the evidence of the finding. You go straight to the idea to edit it.
• Add to an existing idea adds all the evidence to an idea you already have, for example one you wrote yourself.

After that, the card shows which idea it belongs to. On the idea's page, Research behind this idea groups the evidence by source and shows how many sources confirm it.

During the daily research round, Varly also adds the evidence of confirmed findings to ideas about the same element, and turns confirmed findings without an idea into new ideas.

How it changes the priority score

PXL has a question Confirmed by data? that Varly fills in: 2 points with 3 or more sources, 1 point with 2 sources. In ICE, Confidence goes up when more sources agree. For your own framework, add Confirmed by data? from the ready-made questions, or set any question to Fill in automatically → How many sources confirm the idea.

In Optimize → Prioritization, ideas with two or more sources show Confirmed by 2 sources (or 3, 4…) under their title, so you can see why they rank higher. You can always click a different answer to overrule Varly. See Prioritize ideas.

Common problems

• Nothing connects yet. A connection needs at least two kinds of research about the same page. Start a survey on your busiest pages, and give the heatmaps a few days of visits.
• A survey theme doesn't show up. Oran needs at least 3 answers about a theme, and at least 3% of all answers.
• The position on the page looks off. Varly measures where the text sits when it captures the page for the heatmap, once a week. Right after a page changes, open the heatmap and click Capture the page again.
• Oran says a page doesn't mention something, but it does. Oran reads the text on the page. Info that is only in an image, or worded very differently, can be missed. The finding is then a Signal, never confirmed by the page alone.
• The finding is gone. Findings follow your latest data. Click Look again to check with the newest data.