Cut task time on Bandana’s internal tools by up to 27%

Bandana is a job search platform on a mission to make hiring more transparent, connecting people with local jobs that show pay and hiring details upfront. As the sole product design intern, I redesigned and shipped an internal dashboard and browser extension, reducing task completion time by up to 27% for 100+ contractors.

  • Web & iOS
  • Enterprise
  • Shipped 2026
Bandana’s internal tool: a job’s setup steps, with an applicant’s details open in a side panel.
Role
Product Design Intern
Industry
Job Search Startup
Timeframe
Jun–Aug 2026 (10 weeks)
Team
3 Engineers
1 Product Design Intern
Skills
Product Design
Product Analytics
User Research

Designing the tools that turn user-approved job matches into submitted applications.

Bandana applies to approved jobs on behalf of their users, and every application is filled out and submitted by one of about 100 contractors. They work in one of two lanes: Completing XML or Posting applications.

XML applicationsPosting applications
Where jobs come fromPaid feeds from job advertising providersEmployer career sites and applicant tracking systems
RevenuePaid per application or clickUnpaid
Avg Completion Time~10 min~2 min
Contractor payHourlyPer application

Operators are the Bandana employees who run the contractor team. They decide who works which lane, enforce each job provider’s rules and budget caps, and check that the work actually gets done. Two operators handled all of it by hand:

  • Handing out work: reassigning approved matches 25 at a time whenever a contractor asked for more in Slack
  • Passing along rules: announcing provider rules, like required click counts or an employer not to apply to, in Slack as they changed
  • Verifying work: checking each contractor’s screenshots and tallying their hours from Slack threads into a spreadsheet
  • Tracking expired jobs: re-checking them several times a day, since most reopen later

Operator time grew with every match, and Bandana aimed to roughly triple its volume. The result was slower turnaround, jobs expiring before anyone applied, a higher cost per XML application, and no reliable way to tell which applications went through.

The old contractor workflow

XML Jobs

$10/hr, ~8–10 min each

  1. Look up the provider’s city/state in internal, then set up an IP Royale proxy matched to that location. Add a new IP, toggle on “High-end pool”, and open a new incognito tab.

  2. Confirm the IP is in the right location using IP Royale’s online IP lookup tool.

  3. Then, use David’s plugin to check the strength of the IP, whether the IP is “clean” (e.g. hasn’t been used for the same provider in the same week), and your internet speed.

  4. Apply manually. Contractors use ChatGPT to help answer open-ended questions. If necessary, use proxy email for verification codes (additional ~2–6 mins)

  5. Screenshot the completed application and post it with the IP number to the contractor’s personal Slack thread.

  6. Mark “Applied” in the internal tool.

  7. Disconnect and delete the saved IP and close the incognito window.

Posting Jobs

$0.25/app, ~2–3 min each

  1. Go to Bandana’s internal tool.

  2. While logged into Bandana.com, use the Sync and Apply extension to auto-fill (or manually fill out, depending on the provider). The extension provides the application info.

  3. Screenshot the completed application and post to the contractor’s personal Slack thread with the applicant’s username + applied company.

  4. Mark “Applied” in the internal tool.

What I learned

I interviewed 8 people (3 operators and stakeholders, and 5 contractors) to map both workflows end to end.

“Research” in three steps. Reading and writing documentation, with the “Contractor Workflow — UXR Hypothesis & Research Plan” doc. Conducting 8 UXR interviews across stakeholders, shown as speech bubbles. Aggregating research results to inform designs, with the “Contractor Internal Tooling — UXR & Design Summary” doc.

Five problems stood out

  1. Proof of completion is manual and easily faked.

    “There have been many abnormalities about hours reporting… someone is trying to lengthen their time, not really doing postings every minute—instead of sending every minute, they send one every 30 minutes… those rules are implemented, however, they do not really stop the fraudulent time reporting later on.”

    Keizel, Contractor
  2. Contractor knowledge ranges widely across experience levels.

    “New updates are impromptu. New Contractors have a harder time being onboarded.”

    Keizel, Contractor
  3. The application completion process itself is full of friction.

    “There’s really a lot of windows open, and sometimes, it leads to mistake, especially from copying the IP address from IP lookup, pasting everything.”

    John Joseph, Contractor
  4. Application failures are invisible — the current process masks them.

    “There’s not a uniform alignment whether an application is successful or not... different contractors have different knowledge as to whether the application is actually submitted or it’s just in the initial level of questioning.”

    Keizel, Contractor
  5. Manual assignment is the central bottleneck, and it causes delays on both sides.

    “…the adding jobs part… because yeah, sometimes [David & Joseph] are busy as well.”

    Orland, Contractor

Contractor portal + extension

How might we help contractors complete more applications, correctly, with less operator oversight?
ProblemWhat I designed
Too many tabsOne flow, with the job, IP details, applicant info, and resume docked beside the employer’s page
Resumes slowed applications downA resume panel with a quick view and a copyable text view
Knowledge lived in SlackA setup tab and FAQ that gather the rules in one place
Proof of completion was easy to fakeSession and application timers, plus an automatic screenshot when an application is marked done
Failures were invisibleClear outcomes: mark as done, or mark as incomplete with a reason
Before and after: resumes. Before: an employer’s application form beside a downloaded resume PDF. The application completion process itself is full of friction; contractors had to download, read through, and copy resume info. After: Bandana’s docked panel with the Resume tab’s Quick View. Resumes are easily viewable inside the plugin with an option to copy and paste info from them.
Before and after: setup rules. Before: scattered Slack reminders, like “DO NOT refresh the page on the same IP.” Contractor knowledge ranges widely across experience levels. After: the Setup tab, with IP Setup, Setup Reminders and Provider Rules in one place. The informational set-up page and FAQ tab aggregates knowledge in an accessible way.
Before and after: proof of completion. Before: a Slack channel where contractors post screenshots and IPs in threads. Proof of completion is manual and can be faked. After: the session timer and a “Mark this application as done?” dialog that takes a screenshot as a record. Session timers, application timers, and screenshots are automatically captured and logged.

Key decisions

  • The application timer can’t be paused. XML contractors are paid hourly, so time ties directly to pay.
  • A new contractor should finish an application correctly without asking for help, and a veteran shouldn’t be slowed down by the beginner experience.

In testing, contractors completed applications 17% faster for postings and 26.5% faster for XML.

17% decrease in average time to successfully complete a posting application. 26.5% decrease in average time to successfully complete an XML application. Footnote: averages were compared between June 2026 and July 2026. Posting application completion time decreased from 2.27 min to 1.88 min (17% decrease) while XML applications decreased from 10.2 min to 7.5 min (26.5% decrease).
Early results from a partial rollout; not every contractor had switched yet.
LaneBefore (June)After (July)Time savedMore applications per shift
Posting2.27 min1.88 min17%~20%
XML10.2 min7.5 min26.5%~36%

Two early testers completed 329 posting applications at about 50 seconds each, with a 93% success rate, and 218 at about 1.5 minutes, with an 87% success rate.

More applications per shift means fewer expired jobs and a lower cost per XML application. Hours and proof of completion are now captured automatically, and failed applications are logged with a reason, so the failure rate can finally be measured.

Contractor testimonials also resonated with the new workflow changes.

  • Martin

    “Really like the interface, minimalistic and easy to operate in. When you are in an application, the bar on the right is really easy to find information you need about a user.”

  • Marco

    “The new workflow is so much smoother than before because all the information is displayed on the side of the screen. This makes it easier to navigate and help reduce mistakes. Overall, it’s a significant improvement over the previous interface/internal system we used.”

  • Cathrina

    “[The new contractor workflow] is much easier and quicker to navigate now. I like that it’s straightforward and almost all info are there already.”

What I’d carry forward

  • Product sense comes from asking. Much of the process wasn’t written down, so I learned to ask “dumb” questions early and find the right people.
  • Startups don’t move in straight lines. Requirements changed often, so I learned to scope tightly and iterate without getting attached.
  • Designers advocate for users. Bringing contractors’ own words to engineering helped me defend decisions, and adjust them when needed.

Other work from the summer

“Odds & Ends”: a grid of tiles for other summer work. Marketing + BizDev Assets (SHRM, LinkedIn, etc.), with a “Spin to win!” flyer and a “1M+ followers across socials” graphic. 10+ UXR interviews. Remote Jobs Designs, with Bandana’s remote job search on iOS and web. New Internal Design System Components. AMB Upsell & Rich Link Designs, with a text-message job alert. UXR Interviews Wrapped Website, titled “Bandana’s Users, Wrapped”. iOS Push Notification Experiment Initiatives, with For You push notification copy experiments.

Next Case Study