The target metric is agreed before anything is built, and it is almost never engagement
Click-through gets reported because it is diagnostic, not because it is the goal
Teams that optimise clicks usually discover the problem a quarter later
Assignment, exposure logging and readout get built first, so every later change has an honest measurement
It also means we can tell you when a change did nothing, which happens more often than vendors admit
Without a control group, a seasonal swing looks exactly like a win
A recommender with no exploration converges on existing best sellers within weeks and stops surfacing anything new
New items get deliberate exposure and the ranking carries an exploration budget
This is the most common reason these systems stop adding value after three months
Froggik
"DESH Team maintained effective communication throughout the project."
Thanks to DESH Team's work, the client saw increased product recognition within the cryptocurrency community. The team managed the...
Viktoriia Bernatska
Co-Founder
ChainCrafters
"I liked their corporate policy and how they turned to customers and their wishes."
DESH Team delivered the project on time, effectively improving the site's UX and flow. The team took the time to understand the cl...
Kolya Vovkun
CEO, Founder
Dropshipping
"I really like how they treat their clients."
DESH Team successfully completed all deliverables; the branding was a great fit for the client's company, and the website was done...
Tetyana Yarchak
CEO
Enough for a meaningful experiment, roughly thousands of sessions per variant per week. Below that, content-based personalisation still helps but the lift will be hard to prove statistically, and we will tell you that before you spend on it.
No. Click-through is easy to inflate and often trades against revenue. We optimise on the business metric and report engagement alongside it as a diagnostic.
Cold start is designed for explicitly, using attributes and content similarity plus deliberate exploration. Skipping it is the most common reason a recommender stops adding value after a few months.
Yes, as a service alongside your platform or embedded in your existing search infrastructure. We work with what you have rather than making a replatform a precondition.
First experiments usually run within six to eight weeks. Expect the first test to teach you something about your traffic that changes the second one. That is normal, and it is why the experiment infrastructure matters more than the initial model.
