The first stage in setting up sales funnel reporting is to understand your sales process fully (perhaps using a technique like flowcharting). While sales processes are often quite similar from company-to-company, there can be points of difference depending, for example, on the size of the order and the consequences to the client of making the wrong purchase decision.
I'm not sure how the algorithm works (or if there's one at all), but while I was filling it out, I had some anxiety about not qualifying. Once I found out I did, I was excited to fill out the form, which I'm sure most people who are in debt and using this tool are. By making this offer seem more exclusive before the form appeared on the landing page, I'd bet that Bills.com increased conversions pretty significantly.
If a sales opportunity does not move down the funnel, the sale will not happen and the opportunity should be removed, hence the “leaky” funnel. A leaky funnel is not necessarily bad; as a salesperson, you want to focus on opportunities that are likely to yield results. It is the nature of sales to have to remove an opportunity from your funnel. It does not mean that you will not sell to that account (a positive action by the customer can put them back into the funnel), but for the time being, you should centre your attention on opportunities that remain in the funnel.
Full disclosure: IMPACT is a HubSpot partner -- but that's not why they're included here. IMPACT's landing pages have long been a source of design inspiration. I love the simple layout of the page, from the large headline copy and detailed featured image, to the outline that surrounds the form, to the colors and fonts that are very pleasing to the eye.
Open-ended experimentation. This approach is similar to closed-ended experimentation except that more variations will be added for testing and experimentation will not stop when a winner is found. This method is used by large corporation to dynamically improve their conversion rates and improve user experience. Landing page can also be adjusted dynamically as the experiment results change to further enhance user experience.