every product, regardless of its scale, has a key feature that is its foundation. such a feature, or rather a unique feature in linkmuse, was castings.

castings were a strip of cards with names, descriptions, and fees. in addition, they implied the possibility of responding to them. all castings were created and published by agents, casting directors, productions, and agencies in need of talented performers.

the functionality provided for creating castings and performer feedback was far from perfect. this was evidenced not only by the conclusions arising from the u-tests and depth tests, but also by the metrics.

monthly tech support tickets, interviews, and falling conversions pointed to problems related to script complexity, lack of notifications, and low readability.

first of all, we clustered the problems and their priority, conducted additional qualitative research and conducted a competitive analysis of sites with similar functionality.

it was important not to delay the implementation of the solution, as activity rates were falling from week to week, so the formulation of the scope of work and decomposition of the task was carried out in parallel with data collection.

most of the discussion focused on two topics:

• what we can improve;
• what we can offer

the answers to these questions were not long in coming. the improvements were to affect the stages of creating castings and responses to them, the processes of working with responses, as well as the search and filtering of content in the output.

the proposals concerned monetisation, or more precisely, paid promotion of castings from customers and feedback from performers. the hypothesis implied a significant increase in coverage and views by holding the first positions in the feeds.

conclusions from research and earlier feedback from clients on verification of performers came down to one idea - implementation of aggregation of customers, performers, castings, services, goods, and even sites.

previously, the company was already planning to move in this direction and was going to develop products independently under its own brand, which undoubtedly required a serious expenditure from the company as a marketplace.

in turn, the new data helped to understand the audience, their desire to make, promote and sell their own content. so the bulletin board was born, which was to continue to expand to locations, jobs, courses, articles, and a database of performers.

creation and publication of castings took customers a large chunk of time, from 10 to 30 minutes, which absolutely did not suit them. endless forms and steps, lack of prompts and confusing scenarios, sometimes leading to a personal account after creating a casting, potentially affecting overall satisfaction with using the service.

so we revised and prioritised all the fields and optimised the steps. in addition, we added cues based on frequently asked questions. the cues were supposed to greatly simplify the creation of casting and in the future to play an important role in gamification.

as a result, in the split test, the form with prompts outperformed the casting creation conversion by a factor of 3 to 73.2% and reduced the session by a factor of 4.5 to 2 minutes.

among other things, with the subsequent update of registration and merging of profiles, the creation of casting became available to everyone without prior authorisation, not just customers.

the first experience users had when opening the service was interacting with the casting feed. the problems were clutter rankings, broken searches, and a curve in filtering.

these problematic areas we addressed iteratively and in parallel with the you-tests. the update of the tape also affected the casting cards, their informativeness, by which the performer could estimate his chances of success at a glance.

when implementing paid services, we relied on the services of other ad sites. when testing several hypotheses, the most successful ones were:

• one-time rise to the top (every third casting a day after publication);
• weekly anchoring in the top (every 7th casting on the day of publication)

the detailed casting page was useful for performers, but more emphasis was placed on customers who had all the previously scattered functionality gathered in one place.

so the feedback left by performers became more accessible and convenient for customers thanks to the merging of scenarios and simplification of profile cards.

a few months after the release, we already had overall analytics on the upgrade. so the conversion rate to casting creation has increased 3-fold to 61.38%, and the bounce rate has dropped 8-fold to 3.1%.

at the same time, the average duration of a session in the feed increased 1.5 times to 4.5 minutes, and the average number of responses per week increased 1.9 times from ~3600 to ~6800.

another interesting thing we tracked was the conversion rate in responses, which doubled from 17.6% to 35.2%, and the conversion rate in the opening of the detailed casting page, which also doubled from 6.4% to 13.44%.