How to measure a branded Roblox world (and what to ignore)

A brand director once showed me a launch deck for a world we could all see people loved, and the headline slide reported impressions. Millions of them. Wrong number on work that had earned the right one, and it nearly got the project killed by a CFO who read “impressions” as “banner ad that underperformed.” Measure a place the way you measure inventory and you bury your own best work.

Report these three, in this order

Start with attendance. Nothing else means anything without it.

Player visits. The count of times a real person opened your world because they wanted to be there. That is a different thing from a reach estimate or an impression served. Across our two Lenta worlds that number passed 1.6M+ player visits, and someone chose to be there for every one.

Average session time. This is the number I trust most, because a media buy cannot fake it. It tells you whether the thing you built is any good. A player who stays eight minutes has told you more than any impression can. A player who leaves in forty seconds has told you something as well, and you want to hear it while there is still time to act on it.

The downstream action you wired in. A receipt code redeemed, a loyalty account linked, a sign-up completed. With Lenta, a spin in the game and a 1,700₽ basket at the till landed in the same customer record. That is the line the CFO cares about, so put it on the slide.

The vanity numbers to drop

Impressions and CPM belong to the ad model you left behind. Report them on a world and you drag the whole conversation back into cheap reach versus expensive reach, an argument no experience wins against a display network.

CPA as a headline is nearly as bad. Frame a world as a cost-per-acquisition machine and you start optimizing it like a performance ad, stripping out the parts people enjoy because they do not convert in the first session. The value here is time and return visits, and CPA framing punishes both. Be suspicious of anyone who leads with it for this format.

Launch-stream view counts and follower spikes photograph well and prove little. They count who watched the launch. They say nothing about who stayed in the world afterward.

Set the target before anyone builds

Worlds get measured badly because nobody agreed on the number up front, so at the end everyone reaches for whatever looks biggest. Name the KPI on day one instead. How many visits count as a win, what average session, which downstream event. Write it down before a single asset exists, and the go/no-go at the end is arithmetic rather than a shouting match.

Every pilot we run opens with that KPI plan for this reason. In 4 to 6 weeks you get a working world and a number you agreed to in advance, which is the only kind of number a board trusts. See what we’ve shipped, or start a pilot and we will set the metrics with you before the first build day.