The Benefits

Never mind the trade-offs, if people aren't too optimistic or pessimistic, your agile poker (or whatever technique you use) otherwise largely works, the result should give you an impression how the total sprint points may vary.

One good thing is by adding more and more of "everyday" random variables together, the relative uncertainty shrinks, and will help you have some reasonable, quantitatively more justified confidence about your plans.

To build some intuition: it's easy to guess how 'taking more measurements affects the accuracy of the average'.

When going for the sum instead of the mean, the difference is skipping a division by the count. Whether we divide by the count, the ratios, such as relative uncertainty remain unchanged.

Of course, beyond a scale (many tasks tallied up) the nature of such projections changes. Yet for shorter predictions, such as those covering a single sprint, with organisational, market and design changes less likely interfering than in the long run, it is more plausible to work out.

For instance, 3 hypothetical tasks may roughly add up as on the following chart:

Here, the individual relative standard deviations of 30%, 31% and 25% totalled into an approximately 17% final result.

Another good thing is that these intervals come at almost no additional cost around sprint planning, enriching your prioritisation repertoire.

And it is sometimes easier to settle on an interval than on a single, misleadingly exact-looking value. It just comes more naturally - and the impression of less friction can be a good sign.

Less friction means fewer challenges robust crowd wisdom has to overcome before it reaches the surface, that is, team members agree. Lengthy negotiations are more exposed to being imbalanced by opinionated thought leaders, and be overaffected by their bias. The opportunity to shorten the length gives you another chance to reduce bias, as well as empower, wake up and utilize less exploited talent, and an effortless means of cutting back on less effective arguments.