Planning poker is predicated on a course of often known as Wideband Delphi. It’s time to rediscover the underlying technique, developed in phases first at RAND within the 50s after which by the late, nice Barry Boehm and John Farquhar. This iterative, consensus-driven strategy is essentially the most correct estimating technique out there.
Begin by insuring everyone seems to be well-informed about what’s being estimated. Watch out that the particular person explaining the challenge doesn’t bias everybody. This may be mitigated by sharing this duty.
Contain individuals with good data of the area. You can begin with as few as two, though estimates will enhance as extra people are added till about 5.
Choose a various array of individuals to mitigate choice bias. Whereas area data is essential, attempt to embrace people who find themselves off the staff and even socially distant. With a enough variety of individuals, you would possibly eradicate these in authority from the train, as these individuals are likely to affect others.
If the vary of estimates isn’t narrowing as you iterate, have everybody reveal the rationale behind their estimates, not simply the outliers.
Doc all rationales all through the method, as these might inform your closing report back to stakeholders.
If it emerges that info is lacking or requested for, cease to seek out it, after which reconvene. If you reconvene, first replay the final spherical and skim out the final revealed rationales for estimates.
Three iterations are sometimes all that’s wanted for convergence, with estimates narrowed however unlikely to additional converge.
If you find yourself able to derive an estimate from this train, it’s possible you’ll use the median or common of particular person estimates. If you wish to respect an outlier opinion, select averaging. If there’s a clump of settlement you respect extra, select the median.
The remaining dispersion in estimates can be utilized to ascertain an uncertainty round your closing estimate, so as a substitute of “20 months” you would possibly report “15 to 30 months with 20 probably”.
Whereas Wideband Delphi could be performed effectively, its iterations are in all probability overkill for dash planning. Nevertheless, a lot of what occurs in improvement spans a number of sprints. Subsequent time you might have an epic to scope out or a serious characteristic to decide to in a street map, give it a strive.
My good friend Ricardo Valerdi took a deep dive on Wideband Delphi.
