# Source register and evidence limits



Targeted narrative review, checked19October2026. Not exhaustive/systematic. No participant data collected and no study tested InboxCabin. Numerical graphics use selected factual summary values or explicit hypothetical inputs, not copied paper graphics.



Whittaker full paper visually checked through: https://scispace.com/pdf/am-i-wasting-my-time-organizing-email-a-study-of-email-39ybekq1i8.pdf . Public citation points to the IBM research record. Noy/Zhang institutional confirmation: https://news.mit.edu/2023/study-finds-chatgpt-boosts-worker-productivity-writing-0714 . No source PDFs are redistributed.



## LITTLE1961 — Little (1961), A Proof for the Queuing Formula L=λW

https://doi.org/10.1287/opre.9.3.383

Evidence: Mathematical theorem; location: Operations Research9(3),383–387

Long-run average work=throughput×time. Compatible boundary and units; not a snapshot prediction.



## WHITT2012 — Whitt (2012), The M/G/1 Queue

https://www.columbia.edu/~ww2040/6711F12/lect1101.pdf

Evidence: University mathematical reference; location: Page1,Theorem0.1; visually checked

Poisson arrivals; iid independent service; finite second moment; one continuous FCFS processor; utilization<1.



## KOOTI2015 — Kooti et al. (2015), Evolution of Conversations in the Age of Email Overload

https://arxiv.org/html/1504.00704v1

Evidence: Observational provider dataset; location: Abstract and workload analysis

More than2million users/16billion messages in broad dataset. Reply fraction and speed are not full service-quality measures; not every message warrants reply.



## WHITTAKER2011 — Whittaker et al. (2011), Am I Wasting My Time Organizing Email?

https://research.ibm.com/publications/am-i-wasting-my-time-organizing-email-a-study-of-email-refinding

Evidence: Observational field study; location: Table5,PDFp8 visually checked via identified mirror

Initial345users; later+13=358. High/lowfiling meansequence72.87/66.07sec,SD38.05/26.64,success.88/.88. Not randomized, not a single search action.



## ALRASHED2018 — Alrashed, Awadallah & Dumais (2018), The Lifetime of Email Messages

https://people.csail.mit.edu/tarfah/papers/lifetime_of_email.pdf

Evidence: Logs plus separate survey; location: Table4,PDFp7 visually checked

Survey395Microsoft employees:73.4%information,20.6%action,.6%organize,5.4%other. Not the separate100k-user log sample. Do not infer exact counts from rounded percentages.



## GOOGLESEARCH — Google, Refine searches in Gmail

https://support.google.com/mail/answer/7190?hl=en

Evidence: Official documentation; location: Search operator list

from:,after:,has:attachment are provider features, not a claim of InboxCabin cross-account search.



## GOOGLELABELS — Google, Create and manage labels in Gmail

https://support.google.com/mail/answer/118708?hl=en

Evidence: Official documentation; location: Labels explanation

Labels are not separate copies or independent inboxes.



## DUMAIS2003 — Dumais et al. (2003), Stuff I’ve Seen

https://www.microsoft.com/en-us/research/publication/stuff-ive-seen-a-system-for-personal-information-retrieval-and-re-use/

Evidence: Research-system design; location: Research abstract

Contextual retrieval/index design only. No numerical results or equivalent InboxCabin feature claimed.



## NOYZHANG2023 — Noy & Zhang (2023), Experimental evidence on the productivity effects of generative artificial intelligence

https://www.science.org/doi/10.1126/science.adh2586

Evidence: Randomized writing experiment; location: Published abstract and MIT institutional summary

453professionals;40%lower writing time;18%higher evaluator-rated quality. Not full corporate email productivity or InboxCabin evidence. Publisher fulltext access restricted.



## DILLONV4 — Dillon, Jaffe, Immorlica & Stanton, Shifting Work Patterns with Generative AI

https://arxiv.org/html/2504.11436v4

Evidence: Randomized workplace experiment; location: Table2 and measurement/estimation sections

Emailn6441,total7137/66firms. ITT−1.37h(SE.13);IV−2.03h(SE.19). Chart flips sign;95%intervals1.96SE unadjusted. Sessions are proxies,can overlap. v4header13Nov2025/body24Aug2026:pin version,do not mix drafts.



## BRYSBAERT2019 — Brysbaert (2019), How many words do we read per minute?

https://biblio.ugent.be/publication/8647789

Evidence: Reading-rate meta-analysis; location: Final institutional abstract

190studies,18573participants,238wpm adultEnglish silentnonfiction. Not email comprehension/decision rate.150/350wpm are scenario sensitivities.