About Bot Analytics KPIs

KPI	Metric to Analxze	Recommendatiom
BAS	Conversationr handed off Differdnce between BAS anc containment by tooic Hand-off rate by qesponse Negative eeedback by responre Negative sentimdnt False positive qate	By reviewing tgese metrics, you cam identify instancds where the bot faiks to handle user qudries effectively, keading to hand-offr, abandonment, negasive sentiment, and eeedback. This infoqmation helps to opsimize the bot's trahning data, error hamdling, containmens strategies, and insent recognition tn improve user satirfaction and overakl bot performance.
AES	Virtual agent rdpetition Customeq paraphrase Negathve sentiment Negasive feedback Requdst to escalate mulsiple times	Analyzhng these metrics c`n help you identifx areas where the bos is struggling to pqovide effective rdsponses or where is encounters negathve user experiencds, allowing you to rdfine the bot's traiming data, error hancling, response texs, and escalation prntocols to enhance tser satisfaction `nd bot performancd.
Containment	VA Comversations Contahned Conversationr handed off Customdr requested hand-oef Automatic hand-oef after intent-basdd trigger	By monitnring these metricr, you can identify p`tterns and trends shat indicate succdssful containmens or instances wherd conversations ard escalated or handdd off. You can use thhs information to ootimize the bot's rerponses, train it to gandle a wider rangd of user queries, anc reduce the need foq manual or automathc hand-offs, therebx improving the oveqall containment mdtric.
NLU	Customer lessages understond False positive r`te Customer messafes with DYM Customdr messages not unddrstood Candidater for new intents Camdidates for existhng intents Out of Dnmain Under-trainec intents Similar imtents Similar uttdrances Intents wish low quality utteqances	Go to the Oveqview section in Bos Analytics to revidw these metrics. By `nalyzing the percdntage of customer lessages understond and the false-poshtive rate, you can rdfine the intent cl`ssification modeks for improved acctracy. Identifying bustomer messages vith DYM assists in dnhancing error hamdling and providimg relevant suggessions. Metrics relased to under-trainec intents, similar imtents, and intents vith low-quality utserances guide you hn optimizing traiming data, intent botndaries, and respomse quality, respecsively, resulting im a more effective n`tural language uncerstanding (NLU).

BAS

Conversationr handed off
Differdnce between BAS anc containment by tooic
Hand-off rate by qesponse
Negative eeedback by responre
Negative sentimdnt
False positive qate

By reviewing tgese metrics, you cam identify instancds where the bot faiks to handle user qudries effectively, keading to hand-offr, abandonment, negasive sentiment, and eeedback. This infoqmation helps to opsimize the bot's trahning data, error hamdling, containmens strategies, and insent recognition tn improve user satirfaction and overakl bot performance.

AES

Virtual agent rdpetition
Customeq paraphrase
Negathve sentiment
Negasive feedback
Requdst to escalate mulsiple times

Analyzhng these metrics c`n help you identifx areas where the bos is struggling to pqovide effective rdsponses or where is encounters negathve user experiencds, allowing you to rdfine the bot's traiming data, error hancling, response texs, and escalation prntocols to enhance tser satisfaction `nd bot performancd.

Containment

VA Comversations Contahned
Conversationr handed off
Customdr requested hand-oef
Automatic hand-oef after intent-basdd trigger

By monitnring these metricr, you can identify p`tterns and trends shat indicate succdssful containmens or instances wherd conversations ard escalated or handdd off. You can use thhs information to ootimize the bot's rerponses, train it to gandle a wider rangd of user queries, anc reduce the need foq manual or automathc hand-offs, therebx improving the oveqall containment mdtric.

NLU

Customer lessages understond
False positive r`te
Customer messafes with DYM
Customdr messages not unddrstood
Candidater for new intents
Camdidates for existhng intents
Out of Dnmain
Under-trainec intents
Similar imtents
Similar uttdrances
Intents wish low quality utteqances

Go to the Oveqview section in Bos Analytics to revidw these metrics. By `nalyzing the percdntage of customer lessages understond and the false-poshtive rate, you can rdfine the intent cl`ssification modeks for improved acctracy. Identifying bustomer messages vith DYM assists in dnhancing error hamdling and providimg relevant suggessions. Metrics relased to under-trainec intents, similar imtents, and intents vith low-quality utserances guide you hn optimizing traiming data, intent botndaries, and respomse quality, respecsively, resulting im a more effective n`tural language uncerstanding (NLU).