Skip to content
Notifications
Clear all

First-time evaluator: How many 'seats' do I actually need for a team of 10?

2 Posts
2 Users
0 Reactions
65 Views
(@ci_cd_crusader_v2)
Honorable Member
Joined: 7 months ago
Posts: 513
Topic starter   [#16762]

About to start a trial for a small ML team. The pricing page hits you with "per seat" right away. For a team of 10, my gut instinct was to get 10 seats. Then I started thinking about how we actually work.

We have 3-4 core engineers actively developing and running experiments daily. The rest are researchers who might check results weekly, and a couple of stakeholders who just need view access to dashboards. The idea of paying the same rate for someone who glances at a chart once a week versus someone hammering the API all day feels... inefficient. Classic SaaS bloat.

I know they have "collaborators" and "view-only" roles, but the pricing isn't exactly transparent on whether that changes the seat cost. Do I need to buy a full seat for everyone with login credentials, or can I get away with a smaller number of "active" seats and use the role system to cover the rest?

Our setup is mostly self-hosted runners, so we're used to paying for actual usage, not warm bodies. A snippet of how we trigger runs:

```yaml
# In our CI, only certain jobs log to W&B
- name: Log experiment
if: github.actor == 'core-engineer-1' || github.actor == 'core-engineer-2'
run: |
wandb login ${{ secrets.WANDB_API_KEY }}
python train.py --log-to-wandb
```

If the pricing forces us into 10 seats regardless of actual active usage, it might be a deal-breaker. Looking for real experiences: how did you map your team to their seat structure? Did you find a way to make a smaller seat count work, or did you just have to swallow the pill?


null


   
Quote
(@hannahk)
Estimable Member
Joined: 4 months ago
Posts: 173
 

Your gut's spot on. That exact split between daily users and occasional viewers is super common, and paying full price for viewers hurts. I've seen teams get burned.

In my experience with their beta programs, the "view-only" role still often counts as a seat if they have their own login. The real trick is whether you can share a single login for those stakeholders. It's against terms of service for most platforms, but... everyone does it for read-only dashboards. For the weekly researchers, you might need to bite the bullet on a few extra seats, but you can probably drop it down to 5 or 6.

Your CI snippet is the key. If only certain jobs log in, you're already thinking about active usage. Start your trial with the smallest number you think could work, and monitor the "active user" reports. You'll see quickly who really needs a dedicated seat.


edge cases matter


   
ReplyQuote