codeforamerica / health

Meta repo to view all health project issues
12 stars 4 forks source link

GCF vs. MBCW vs. C4 Evaluation #492

Open alanjosephwilliams opened 8 years ago

alanjosephwilliams commented 8 years ago

Recap of thinking about this project to date, piped over from private repos:

Originally spec'd out here

https://docs.google.com/document/d/1EQuk9JVVFQ1df6Ui_GybsnV-E44BK7Y7K22k8QHCGOs/edit#heading=h.wqr0jsxjwkip

Quicker turnaround scope originally intended to be delivered before Reid pitch

By 22nd we can might be able to get a good baseline & comparison data for:

Roughly here's how we would do it:

Open Qs

Flow gcf-eval

High Level Summary of Activities & Costs

Summary

Cost

alanjosephwilliams commented 8 years ago

Relevant data from past studies and current analytics

I'm going to recap what we know already, from our own data on GCF and Caroline's research on MBCW, as we prep to share with others in the org and scope the next tranche of work.

First up, GCF data: (All of Jan/Feb 2016, excluding Feb 29)

Step Users Data Source
visitors to /apply from /county 1.3k (66%) GA
visitors that drop off at /county 677 (34%) GA
typeform sessions initiated (saws-driver IDs created) 1018 saws-driver
typeform apps submitted (dupes removed) 449 GCF-cases
typeform conversion rate 44.1% saws-driver/GCF-cases

MBCW data, from Caroline's research: https://docs.google.com/presentation/d/1A5l1x0QiahDJlov3qHyEb_uP2NszdGnRTwud_ueOlrA/edit (Summer 2015)

Population is defined as those that responded to a survey request and follow up interview.

Step Users
Shown screener 2,222
Screener responses 111
Phone Interview (1 hr to 2 days 30
30-Day Follow Ups 15
daguar commented 8 years ago

To clarify the SD #, I'm pretty AdWords are off so I'm not sure that we know what max'd out AdWords blitz on SD would yield.

daguar commented 8 years ago

(Ah I misunderstood, that was copy/pasted; but the general point that we could blitz SD [maybe SF too?] and get a fair # is right to me.)

daguar commented 8 years ago

FYI — I just realized that our numbers from Typeform unique IDs are an underestimate of our conversion since right now both (1) our manual testing of the site [locally or on dev], as well as (2) our automated test suite's integration tests, actually visit the Typeform, which increments that #.

If we want/need to change this to get numbers, let's make it a priority in the study.

alanjosephwilliams commented 8 years ago

@MatthewLoveless this is the public issue we have collecting our thinking to date.

daguar commented 8 years ago

Web app is up and live

daguar commented 8 years ago

GSpreadsheet is also :)

alanjosephwilliams commented 8 years ago

Matthew is going to give us a mid-flight update tomorrow