Microsoft announces lots of new things. You can't look at all of them. So I ask four questions about each one and give it a score. High scores get my attention now. Low scores can wait.
The Microsoft 365 roadmap has more items than anyone can follow. I score each one on whether it applies to us, how much it would change, how soon it's due and how firm that date is. Multiply them together and you get a rough attention score. Above one line, act. Below another, ignore.
Attention score = relevance × impact × timing × certainty, each 0 to 3, with thresholds for act, watch and ignore. Relevance is gated on cloud instance and licensing. Certainty rises with status and Message center confirmation. Scores are re-run monthly against the feed.
The problem
On 10 October 2026, the snapshot of Microsoft's roadmap feed on this site held 1,911 items. 749 were In development and 238 were Rolling out. Nobody can read all of those properly. So horizon scanning is really a sorting job: which few items deserve attention now?
I use a small scoring model for this. It's my own method, and the same shape as the other decision models on the site.
Four scores
Each item gets four scores from 0 to 3.
Relevance asks whether the item applies to us at all. A 0 means it's for a cloud we don't use, a product we don't own, or a licence we don't have. Checking the cloud instance and phase comes first.
Impact asks how much would change. A 3 means it changes how people work, what data is exposed, or what we pay. A 1 is a small interface tweak.
Timing asks how soon it's due. Within a month scores 3. More than six months away scores 1.
Certainty asks how firm the date is. In development with no Message center post scores 1. Rolling out with a Message center post scores 3.
Attention = relevance × impact × timing × certainty. Because it's a product, a 0 anywhere means ignore, which is what I want: an item that doesn't apply to us shouldn't take any time, however exciting it is.
The thresholds
The thresholds are illustrative and need tuning to how much time you have.
| Score | Action |
|---|---|
| 24 or more | Act now: test it, brief people, check settings |
| 6 to 23 | Watch: recheck monthly |
| Below 6 | Ignore until something changes |
A worked example
These scores are illustrative, for a UK organisation with Microsoft 365 Copilot licences on the worldwide cloud. The first two items are real entries from the 10 October snapshot. The third is a hypothetical item, for contrast.
| Item | Relevance | Impact | Timing | Certainty | Score | Action |
|---|---|---|---|---|---|---|
| 573220 Scheduling with Copilot (government clouds only) | 0 | 2 | 3 | 1 | 0 | Ignore |
| 570966 Audio Overview updates in Copilot Notebooks | 3 | 2 | 3 | 1 | 18 | Watch |
| Hypothetical: an oversharing control rolling out next month, with a Message center post | 3 | 3 | 3 | 3 | 81 | Act now |
The Notebooks item lists sensitivity label support for audio overviews among its changes, which pushes impact up for anyone who cares about data protection. It's still In development, so certainty is low. It goes on the watch list.
Why multiply
Adding scores would let a highly exciting item for the wrong cloud score well. Multiplying means every factor has to be there. The cost is that the scale is uneven. Above 12, the only possible scores are 16, 18, 24, 27, 36, 48, 54 and 81, with gaps between them. That's fine for sorting into three groups, which is all I need.
Where I got stuck
Impact is the score I find hardest to keep consistent. A feature that's minor for most staff can be major for an admin or a compliance team. I now score impact for the group most affected and note who that is, so I can see why I scored it that way.
Sources
- Microsoft 365 roadmap RSS feed, snapshot taken 10 October 2026 for this site: https://www.microsoft.com/releasecommunications/api/v2/m365/rss (accessed 11 October 2026).
- Microsoft 365 roadmap, item 570966: https://www.microsoft.com/microsoft-365/roadmap?id=570966 (accessed 11 October 2026).
- Microsoft 365 roadmap, item 573220: https://www.microsoft.com/microsoft-365/roadmap?id=573220 (accessed 11 October 2026).
- The scoring method is my own; the thresholds and scores are illustrative.