# What affects confirmation rate, and how can it be improved?

Currently, confirmation rate of transactions is sitting at about 30-40%.

What goes into determining this number and why is it so low?

What sort of technical adjustments can be made to improve this? In an ideal situation, what confirmation rate would IOTA aim to have, if 100% isn't feasible?

• What is your source? And would be interesting to see how that "confirmation rate" is defined exactly (e.g. are spam tx counted, are invalid tx counted, how are re-attached tx treated, etc). – Phil-ZXX Dec 21 '17 at 17:55
• I don't have a formal source, I'm going off of what devs and admins on slack have said. I will try to follow up. – aboose Dec 21 '17 at 17:56
• @Phil-ZXX you can see raw confirmation rates (for the last few minutes) on analytics.iotaledger.net/stresstest.table. Keep in mind that this will include all transactions, even invalid and duplicate ones (which can never get confirmed). But 40% is definitely low, considering that iota had around 90% earlier. – mihi Dec 21 '17 at 23:36
• One thing that has generally improved confirmation rates is increasing the frequency of milestones. – Jason Kim Mar 20 '18 at 13:04

IOTA tip selection mechanism is based on a weighted random walk. A parameter α>0 dictates how important a transaction’s cumulative weight is. The transition probability P is proportional to $e^{-\alpha\cdot w_2}$, where $w_2$ is the difference between the cumulative weights of the site and the tip. If the tip is a lazy tip, $w_2$ will be big and P will be small. Thus, it is less probable that a lazy tip will be selected (which is an important characteristic of a good tip selection mechanism).