1 of 48

1

HI end-to-end experience

Requirements, Job Optimization, Screening and Explainability

Annie Welch

Lead UX Designer | Hiring Intelligence

March 2024

2 of 48

Carrie, COO

Healthcare

Central Florida

Carrie’s story

Prototype (link)

Intersections with

Monetization

Discussion

Deep dive topics

01

02

03

04

05

Agenda

Indeed HI end-to-end experience

2

Case study

3 of 48

3

4 of 48

4

5 of 48

Advertising Reach Concept Test (2022) - Laura Carroll and Robert Newell

5

Carrie’s story

Carrie articulates a few key pain points:

budget (paying for unqualified candidates)
time (to understand whether the candidate is actually credentialed or not)
ease of use (there is effort in having to track down the information necessary to know if the candidate is credentialed or not)

We’re not going to spend a lot of time on budget because Frontier has plans to address that aspect.

QUOTE: �Oftentimes all of the applicants that I’ve been getting since that changeover recently, like they’ll check yes, that they have the credential they have to have and 99% of them don’t.

And so I’m now having to go in and search and find are these people actually credentialed? Because I paid hundreds of dollars for people that when I talk to them, they’re like, no, I’m not certified, but I’m willing to get it.

It’s like, you’re not qualified for this job.

And it’s like just super time consuming and frustrating.

6 of 48

Advertising Reach Concept Test (2022) - Laura Carroll and Robert Newell

6

Carrie’s story

How can we solve for Carrie’s time and ease of use pain points?

�Well, Carrie has some ideas!

QUOTE:�So for example, one way of working around people that check out, yes, that they’re certified, I think ask, what is your certification number? Because they’re checking yes.

And then I realize that they are not in their resume shows that they’re not in all of these things. So I’m like, okay, well, if I ask them their certification number, then obviously they’re going to like have to answer that and I’ll know, but like, I can’t make that question a deal breaker I guess because it’s like a custom question that I’m making.

And so they, and I can’t make it required. So they just leave it blank and then it’s still like, do they just not have it or are they actually not certified?

7 of 48

7

8 of 48

8

Confidence in the reliability and accuracy of candidate’s application data

What does Carrie need?

High quality job post�Clearly communicate requirements and preferences for the role.

Screening methods evolution�Effective and equitable screening methods to evaluate job seeker qualifications.

Collect evidence�Demonstrate that candidates meet the job’s requirements.

01

02

03

What do Carrie and other employers who are trying to fill these difficult credentialed roles need?

Seems simple, but there it actually a bunch of work we need to do to deliver.

Up front

What’s in flight, what’s being investigated vs. long term directional

How does the solution solve for particular problems - candidates “lying”/candidate control over how they’re presented to employers - hills

ALISON:

“Skills & Experience” strategic hill is broken into top employer concerns… which also seem to align to the problems addressed in this deck

Experience, likes, (assessments), licenses / certifications / education

From how you described the problems we heard about in the clips, here are the problems I think you address in this deck:

Value (do they get the candidate quality they expect for what they are paying?)
Efficiency / effort: can they easily determine which candidates meet their qualification criteria? How much effort / time does this require from the user
“Lying” - perceived & actual accuracy of candidate information - CONFIDENCE IN DATA ACCURACY

Related: opportunities to improve data accuracy (job quality?) -

Transparency into what info we are showing employers
Opportunity to correct / update info
ALSO: MORE NUANCED COLLECTION OF SCREENING DATA

Provides more accurate data
Reduces perceived inaccuracy
Allows JS to create context, further improving data quality & eliminating need to ‘round up’ on part of candidates

9 of 48

2024 Hills

Indeed | Employer Connect

Hiring Intelligence

Hills are key user problems that we are focusing on.

Job Quality

Skills and Experience

Lying

Skills and Experience

we are here

M3 Horizon

Indexed Jobs

we are here

10 of 48

Prototype (link)

10

Let’s walk through our vision for how we plan to help Carrie gain confidence in the reliability and accuracy of candidate’s application data.

The goal of this prototype is to learn - so we’re building just enough to understand whether we’re heading in the right direction or not. The detail-focused work happens as we’re preparing to roll out individual tests to production - we will scrutinize everything at that stage. Consider this a sketch, not a representation of the finished product. We’ll get there!

So the first thing Carrie does is post her job on Indeed. In this case we’re starting with a new post, mostly for illustrative purposes but we recognize starting from a template is also a common use cases and will be considered in our designs.

As Carrie is entering the basic information about her job, she starts to define the location. We recently moved location requirements into this section so Carrie is able to enter all of that information in one spot, rather than having it separated by multiple pages in the funnel. As a result, we’ve seen a significant increase in the attach rates for location requirements.

For this role Carrie requires Work Authorization, but she’s not willing to sponsor a visa, and she can consider candidates who are willing to relocate but because this is a high-turnover role, she’s not willing to pay for that relocation.

I want to point out the tip at the bottom of this page, which is currently in production. It has one of our highest acceptance rates (75%), so we know that we’re headed in the right direction with these optimization tips.

<Next>

So if I’m Carrie, I enter basic information about the job: job type, pay.

When it comes to the job description, Carrie has a document in google drive that she uses to collaborate with her colleagues, so she can copy/paste that content in here.

As you can see, Indeed is “reading” the job description and pulling out requirements and preferences. There are three phases to this work:

Extracting from the job description
Pulling in customizations that Carrie has made to past job posts
Suggesting requirements/preferences that Indeed knows are common for the job/industry.

The goal is to do the work for Carrie - she’s spent a lot of time agonizing over the job description, and we want to honor and leverage her work. Long term, once our models are performing well and employers no longer need to make changes to our extractions, we can move this content into more of a “review” state where employers can fix error in the rare case where we get it wrong. For now, as we’re actively improving our models, we want to give employers transparency and the ability to make changes to our extractions.

The job optimization team has identified three layers to job quality:

Completeness - job includes all necessary information. As you saw, by moving location requirements next to the location component, we saw higher attach rates for those requirements and that helps with completeness.
Consistency - job posting accurately reflects the role & the quality of candidates that employers expect to receive. By moving requirements and preferences next to the job description, we’re hoping to do two things: keep our structured data in sync with the employers’ prose and suggest improvements that can be accepted and incorporated into both places. Today these related elements are separated by multiple pages in the funnel, so we’re excited to build upon the successes with location and connect these experiences for employers.
Competitiveness - job posting is competitive & compelling in the market. We’re still working through the Preview candidate pool experience, but the goal here is to give early feedback to the employer in terms of how their requirements and preferences match the market - this will pull up a list of candidates from our sourcing product so employers can get a quick idea of who they are likely to see once they post the job. With this approach, employers don’t have to wait and burn through those critical first days before they start getting results

<Next>

We can speed through the review page

In this prototype we haven’t yet included monetization or screening, but we will go over those areas when we get back to the deck. The intention with screening is that Indeed steps into the role as the expert in screening methods, so employers don’t need to figure that out for themselves. In Carrie’s example, she tells us what’s required, and Indeed will apply the appropriate screening methods. For some of our higher end tiers, we might offer customization for screening but the goal is that our defaults work well for most requirements. Again, we want to do the work for Carrie.

Once Carrie posts her job, she immediately sees potential candidates through our sourcing product and she can start reaching out and connecting with the candidates that meet her requirements.

After a few days, candidates start to populate in the “In-progress candidates” section - these are Carrie’s applications.

As she’s exploring these applications, one of the first things she’ll see are groups for candidates who meet all of her requirements from candidates who meet some, or none. This helps Carrie focus her attention on the best fits for her role.

Depending on the kinds of candidates Carrie receives, we’ve heard from employers that having access to the job seekers who fit into the second two categories (meets some, and meet none of the criteria) are still valuable. The “some” category represents the candidates that the employer might consider if they’re willing to flex on some of their requirements - maybe they’re not getting enough people in the first bucket, or they realize that a requirement/preference is either more/less valuable than originally thought. The “none” category represents candidates who expressed an interest in her company, and even though they are unlikely to be a good fit for this specific job, they might be worth considering for a different role. When we think about the CRM strategy, it’s more effective to reach out to someone who has previously applied to a position, or engaged with you, than it is to do a cold outreach to a random candidate through sourcing. So this is just a way for employers to build a stable of candidates that they could connect with in the future.

Clicking on the first candidate, she sees that this person meets her required qualifications, but how does she know for sure?

By clicking on the RN license, Indeed shows her evidence - in this case we’ve collected the candidate’s RN license number, we’ve checked the candidate’s knowledge with more sophisticated screening questions, and we’ve highlighted keywords in the resume that indicate this person has the relevant experience.

For Carrie - this is the Ah-Ha moment, she no longer needs to guess whether the candidate meets her qualifications, she doesn’t need to spend time and effort following up on their initial application - she has all the evidence she needs to understand what this candidate brings to the table.

And then she can decide how we wants to proceed. Typically this is where employers will do a deep dive with candidates and start evaluating them for things like culture fit, onsite skill evaluations, etc. This is the part of the hiring process that employers typically enjoy, and we’re getting them there faster, reducing the pain of those routine checks that are currently a stumbling block for employers like Carrie.

11 of 48

Intersections with monetization

11

12 of 48

12

Capture requirements always

01

Flexible architecture to support tiers

02

Feature rollout strategies to avoid thrash for users

03

Intersections with monetization

13 of 48

Job seekers deserve transparent, high quality job posts.

And Indeed requires accurate job data to fuel the system

Intersections with monetization

Qualifications

Flywheel

Step 1

Create/update/recommend

job requirements

Step 3

Source applicants

Step 4

Gather missing candidate data

Step 4

Generate candidate recommendations

Step 5

Deliver

candidates

Step 6

Customer decision

(Y/N/M) gives feedback

Optimizations�(Pre and post job creation)

Evaluation

(Prominent and

Explainer Platform)

SERP, MRP,

Resume

Screening

(Questions, Assessments, Interventions)

Step 2

Configure

screening workflows

Qualifications

(Requirements and preferences)

In trying to decide what might be worth monetizing, we started thinking very carefully about the purpose of each element in our experience. We started to break the job posting experience down into the the “What” - this describes the job (salary, location, job title, requirements- what does the employer need and what are they offering), the next part of the funnel is “How” - the methods for getting what the employer needs (this is where monetization comes in to offer higher visibility, better matching, more screening methods to ensure the candidate has what the employer needs)

Conclusion: capture requirements always

We think it’s important to capture the “What” about the job for all jobs - both free and paid. It’s important for job seekers to know what the job requires when they decide to apply, and it’s important for indeed to have high quality information about all jobs. And if we think about the downstream implications on monetization - knowing what the job requires, gives us the ability to upsell later. We can't upsell quality if we don't know what we're looking for.

So requirements and preferences are not something we plan to pay-gate.

The “how” is where we start to think about monetization - employers can pay more for methods that will help them fill the role - which includes how we screen candidates.

14 of 48

14

Intersections with monetization

Packaging example: Frontier tiers

Free includes organic traffic, with little help from Indeed to validate candidate’s qualifications.

Relevant includes matches + minimal help from Indeed to avoid irrelevant candidates

Exact match includes matches + a full suite of features from Indeed to validate candidate’s qualifications

Place screenshot here

15 of 48

Close Coordination w/Frontier Team

Governing GTM based on ability to deliver quality

15

Intersections with monetization

Example: How do we know if/when we can offer “Right Fit” for [Heavy Tractor Trailer Drivers]?

[JSU] Do we have a sizable population of JS in this occupation? → Yes
[JSU] What about JS w/most requested attributes? → Not as good, about 1 in 4 have CDL-A
[Screening] Can we close the CDL gap with screening?

→ Based on the data, we’ll prioritize CDL screening (2 variants in queue to test)

16 of 48

16

Intersections with monetization

Screening package is only available for exact match tier

Features like

Custom screening questions
Configurable screening packages
AI-enabled capabilities
Etc

are all pay-gated.

17 of 48

Preceding user experience adapts based on tier

17

Intersections with monetization

Free

Right fit

18 of 48

How do we decide what to build?

First, focus on utility and value to users - if it helps Carrie, it’s a good bet. As a product team our primary motivation is building valuable features.

Second, assuming our work generates value, the user experience and availability of those features should be able to adapt based on packaging determined by monetization.

18

Intersections with monetization

Carrie, COO

Healthcare

Central Florida

Case study

19 of 48

Place screenshot here

Test gently, but rigorously

Need to test new features on a broad (free/sponsored/both) population:

Generates enough sample quickly
Avoid selection biases & maintain monetization optionality

But we don’t want to thrash users with capricious UX. We’ll mitigate risk by:

“beta” tagging experimental features
Coordinating each test with IndiMon
Testing on the smallest % possible

19

Intersections with monetization

Balance our need to get data on efficacy of our features vs creating thrash for users

In order to prove validity of our screening method and evaluate it for disparate impact on jobseekers, we need to pilot with a certain volume of jobs before we can launch in production. And we want our test population to be representative of the overall population, to ensure we’re getting high quality data

In addition to giving us more representative results, when we test with the full population, we can test with the same # of people, but in a much bigger pool. Roughly 1 in 10 jobs are sponsored, so testing with 1% of everyone is equivalent to testing with 10% of sponsored.

On the other hand, we’re sensitive to the thrash users might feel if we give/take features too frequently.

The intention is to frame these tests as free beta access, which hopefully primes the user base to get excited for a full launch.

Ultimately, we want to get the data we need with minimal disruption.

20 of 48

Don’t forget the upsell

It’s useful to build in a hook that leads users into a paid features through discovery and natural product usage.

Tease screening packages
Tease evidence on candidate profiles
Generate excitement through betas

Execution of an upsell can be tricky, but effective if done well

20

Intersections with monetization

21 of 48

Discussion

21

22 of 48

Deep dive topics

24 Definition of terms

25-26 High quality job post

27-31 Screening methods

32 System overview

33-35 Candidate evidence

36 What does Carrie need?

37 Hypothesis

38-42 FAQ

43-48 Execution strategy

Slide 23

23 of 48

Preference A
Preference B
Preference C
Preference D
Preference E
Preference F

Requirements

Generally predetermined & inflexible

They matter during job posting

Preferences

Fluid and based on candidates in context

They matter during candidate management

REQ 1

REQ 3

REQ 2

Candidate Pool

Rank/Filter my candidate pool by

Requirements vs. preferences *

Requirements determine the size of the candidate pool

Preferences highlight candidates within that pool

They matter to employers at different stages in the customer journey

23

Definition of terms

* Formerly known as must-have and nice-to-have requirements

24 of 48

24

High quality job post

Doing the work for employers

Methodology in order:

Extract from job description
Leverage past engagements
Indeed’s industry knowledge

Higher quality structured data that is in sync with job description content

Initial test designs in Figma - Joe Kang and Gittings Boyce

One of the first areas we want to test, is in how we collect requirements and preferences.

Recently, we saw success moving location requirements inline with where employers define the location of the job. So we already have encouraging evidence that integrating requirements into the funnel is a good direction. With this move, we also hope to keep the structured requirements and the requirements listed in the job description in sync with each other to reduce confusion for Job Seekers.

Job optimization is extracting requirements from the job description

Requirements team is moving their content inline with the job description

Method:

Extract from the job description
Leverage past engagements/past jobs to recommend techniques that have worked before
Suggest improvements based on Indeed’s industry knowledge

Preview candidate pool button is intended to help employers “right size” their requirements early in the process, rather than waiting to callibrate until after they’ve posted the job.

25 of 48

Layer 3

Competitive: Job posting is competitive & compelling in the market

Layer 1

Complete: Job includes all necessary information.

Word length
Precise location
Schedule/shift
Benefits
Necessary skills (if needed)
License/certifications (if needed)
Education (if needed)

Layer 2

Consistent: Job posting accurately reflects the role & the quality of candidates that it expects

The Layers of Quality

Lower

Higher

High quality job post

26 of 48

Place screenshot here

How do we evaluate candidates against requirements today?

Coarse questions are problematic:

Correct answer is obvious
Lack of detail or evidence
No room for “gray area”

Unreliable

26

Screening methods

27 of 48

Place screenshot here

How do we evolve our screening methods?

Introduce additional techniques:

Knowledge check (vision)
Evidence collection
Open-ended inputs

More reliable

27

Screening methods

Two stages in an employer’s screening process:

Criteria fit - recruiters checking the boxes to determine if they should hand this candidate over to the hiring manager

Evaluative - hiring manager doing a deep dive on skills, culture fit, inviting other interviewers into the discussion, inviting the person on site, practical tests/shadowing

Focus 80% on screening methods in the criteria fit stage - check as many of the boxes we can so employers can spend more of their time on the evaluative stage.

Requirements team is working on adding knowledge check questions to IQP

Screening team is pulling tests that worked well in Evaluate into the Core experience including driver’s license, commute, internet speed verification

Instead of asking…

Do you have a BLS certification?

Yes
No

Before administering cardiopulmonary resuscitation, which is the best place to check for a pulse?

Neck
Wrist
Elbow
Foot
Chest

Correct Answer: A

28 of 48

Place screenshot here

The real world is not black and white

Accurate data capture includes giving space for candidates to explain why they consider themselves qualified and communicating those nuances back to the employer to decide.

28

Screening methods

One way that we hope to support job seekers in more accurately representing themselves is making space for the “gray area” - this job seeker doesn’t currently have a bachelor’s degree, but they will get it soon, so we want to provide that level of detail to the employer so they can decide if that candidate meets their needs.

Carrie mentioned that someone who does not currently have their RBT license is not useful to her, but if she knows that they are already in the process of getting that certification, maybe that person is worth considering depending on the urgency of the hiring need and where the candidate is in their process.

We have a hypothesis that one reason job seeker’s appear to lie, is because they don’t currently have a way to accurately represent their situation. We have evidence from Indeed Spotlight that giving JS 2-3 lines to explain the “why” often yields useful information.

29 of 48

Precise knowledge check questions

“Before administering cardiopulmonary resuscitation, which is the best place to check for a pulse?”

Neck
Wrist
Elbow
Foot
Chest

Tests & Assessments

Real-time rating of an actual skill beyond a boolean

Structured question

“Select the shifts you can work”

Trusted Verification

(often from a third party)

Coarse questions

“Do you know CPR?”

Yes
No

More trust

More detail

(more friction, more cost)

The Direct Employer Journey Playbook (2023) - Nicole Kachelmeier

29

Screening methods evolution

30 of 48

30

Screening methods

31 of 48

Req 1

Req 2

Req 3

Preferences

Micro-content Q

Profile

Application Confirmation

*Updated*

Profile

Profile/Job Match

✅

🤷

Previous Apply

Job example

Req 1: Must have an RN license

Stacked validation example

Do you have an RN license? Y/N (required)

Enter license number _________ (optional)

Upload license photo ⬆️(optional)

Job seeker example

?

Application Qs

32 of 48

32

Candidate evidence

Now, we’re getting into the real bread and butter of Carrie’s needs - how do I know, with confidence, that this job seeker actually meets my requirements?

Indeed is now collecting and representing evidence back to the employer. In this case, we’ve extracted keywords from the candidate’s profile that indicate this person has surgical care and dialysis skills.

We’re thinking about this in terms of “showing our work”, providing transparency so employers don’t just trust our checkmarks that indicate a candidate has something, they can see how we came to that conclusion.

Future iterations might include giving employers a way to override or change our assessment and add their own notes. For example, if we don’t know that this candidate has surgical care, the employer could talk to the candidate, gather that information and then indicate in this profile what they learned.

33 of 48

Place screenshot here

Show evidence to support each application

Indeed offers the most transparent and equitable validation of requirements and preferences. We should display these details on the candidate profile and be transparent about where this information was obtained. (Employers are always asking for this)

33

Candidate evidence

Finally - Carrie’s ah - ha moment

She wants to know if candidates truly have their RBT license so she doesn’t have to spend a lot of time reaching out to each candidate and jumping through hoops to check that crucial checkbox in her hiring process.

So she’s defined that requirement in her job description, Indeed extracted the structured data to become a requirement in her job post, and she has confirmed that as well as any other requirements while she’s entering the job description. As part of Frontier she’s opted into a pricing tier that gives her access to comprehensive screening methods and high quality matches. During the apply flow, job seekers have provided evidence of their RBT license, or a rationale as to why they should be considered anyway.

Now when Carrie looks at her candidates she sees a shortlist of the candidates who have provided evidence of their certification, with access to the candidates who may not have it yet, but can see why they chose to apply anyway, and when she drills down on a particular candidate who has their license, she can see the license number and any other evidence that the job seeker has provided so Carrie can jump straight into the “evaluative” screening step, which is what she really enjoys.

34 of 48

Candidate Blocking & Auto-Reject (2024) - Patricia Nunez and Alison Berent-Spillson

34

Candidate evidence

This data and UX pattern is extended to the list, where it also drives more sophisticated bucketing and sorting

In coordination with the Frontier team and the sponsorship package that the employer selects, we plan to give varying levels of support to employers to interpret their candidates. In this image we’re showing the “Right fit” experience, which breaks candidates into three groups:

Candidates who meet all required qualification
Candidates who meet some
Candidates who meet none

There are valid reasons an employer might want to dip into one of the other “buckets” of candidates. For example, if they’re not getting enough candidates who meet all of their qualifications, or if they decide they’re willing to flex on some of their requirements, or if they realized that a preferred qualification is actually more important than they originally thought and now they want to promote that characteristic to required.

This experience could eventually evolve to include highlighting equivalencies/transferable skills - things that the employer didn’t specifically ask for but the job seeker offers.

35 of 48

35

Confidence in the reliability and accuracy of candidate’s application data

What does Carrie need?

High quality job post�Clearly communicate requirements and preferences for the role.

Screening methods evolution�Effective and equitable screening methods to evaluate job seeker qualifications.

Collect evidence�Demonstrate that candidates meet the job’s requirements.

01

02

03

What do Carrie and other employers who are trying to fill these difficult credentialed roles need?

Seems simple, but there it actually a bunch of work we need to do to deliver.

Up front

What’s in flight, what’s being investigated vs. long term directional

How does the solution solve for particular problems - candidates “lying”/candidate control over how they’re presented to employers - hills

ALISON:

“Skills & Experience” strategic hill is broken into top employer concerns… which also seem to align to the problems addressed in this deck

Experience, likes, (assessments), licenses / certifications / education

From how you described the problems we heard about in the clips, here are the problems I think you address in this deck:

Value (do they get the candidate quality they expect for what they are paying?)
Efficiency / effort: can they easily determine which candidates meet their qualification criteria? How much effort / time does this require from the user
“Lying” - perceived & actual accuracy of candidate information - CONFIDENCE IN DATA ACCURACY

Related: opportunities to improve data accuracy (job quality?) -

Transparency into what info we are showing employers
Opportunity to correct / update info
ALSO: MORE NUANCED COLLECTION OF SCREENING DATA

Provides more accurate data
Reduces perceived inaccuracy
Allows JS to create context, further improving data quality & eliminating need to ‘round up’ on part of candidates

36 of 48

We think if we…

Improve job quality

Offer compelling, effective screening packages

Show evidence that candidates meet requirements

36

Hypothesis

“

I’m now having to go in and search and find are these people actually credentialed?

“

Ask, what is your certification number? Because they’re checking yes.

…we’ll see POVR go up, and Ask’em complaints go down

37 of 48

Frequently Asked Questions

37

38 of 48

How does this all work for Indexed jobs?

Big picture: The features should be a draw to operate on platform (Horizon)…

(“tuning” the screening plan has to happen on Indeed, screening interventions happen mid-apply)

…whereas some of the data could go to the ATS downstream

(e.g. we could, in the future, push an assessment score to an ATS field)

Practically speaking, we’re focused on Hosted for the foreseeable future

38

FAQ

39 of 48

How do we decide what’s free and what’s not, from the menu of “value tested” features?

We’ll weigh in, but Monetization makes the final call.

Example:

We [HI] think Assessments are philosophically suited to a high tier because they are meant to replace a step of the employer’s process.
On the other hand: it’s been available for free for a long time, and our current Assessment scoring isn’t particularly useful today
Net: Monetization can package today’s Assessments and future scoring/filtering improvements individually. We won’t launch new scoring without consulting Monetization.

39

FAQ

40 of 48

How does this map to the Screener Question Paygate experiment slated for launch soon (~May)

The SQ Paygate experiment “pre-tests” a broad hypothesis in a blunt way:

Is “set up filters to see fewer irrelevant candidates” a pay-worthy concept?

…and a specific articulation of that hypothesis in a precise way:�“Are basic screener questions valuable enough to monetize?

Both answers inform the future.

40

FAQ

41 of 48

How does crowtag fit in?

Crowtag has a few major components:

A different taxonomy from what Indeed uses today
A large library of questions (in Japanese) tied to that taxonomy
Mechanisms to accelerate the generation of that library by converting unstructured questions into new structured ones
JS-facing onboarding experiences to enrich profiles with answers upfront
Matching based on the question and answer data collected under this taxonomy

Our team is building the necessary infrastructure to enable (2-5). Along the way, we’re also testing a homegrown version of (2).

Again, we’re flexible about how to package the wins from our building and testing. We’ll advise Monetization’s decisions here.

41

FAQ

42 of 48

Execution strategy

42

43 of 48

43

Execution strategy

We must provide value to customers

44 of 48

44

Execution strategy

Three primary tracks of work

Build and test in product

We have enough evidence to start building quite a few ideas.

Product Roadmaps: requirements, job optimization, screening, explainability

Generative and evaluative research

Some ideas require more confidence before we decide on a particular approach.

Some ideas require more refinement before they are ready for development.

The goal is to be just ahead of development so when resources become available we are confident in the execution.

Prototype (WIP)

Collaborate with frontier

As frontier continues to define the tiers and test assumptions, we are supporting their work through ideation, development, and advising.

45 of 48

Evidence is the key�Employers breeze through the “criteria fit” and spend more time in the “evaluation”

Screening packages collect evidence�Employers accept Indeed’s methods and job seekers provide evidence

JD and requirements need to be in sync�Showing requirements in context with the job description improves job quality

Frontier collaboration

01

02

De-risk these core assumptions through testing

03

Quantitative

In-product testing

Qualitative

Generative research

Prototyping

04

46 of 48

Iterating in product

46

Execution strategy

HMW get more employer engagement at the point of requirement collection?

One of several concepts, from idea to QA in ~2d

47 of 48

47

HI E2E Experience Planning Sheet | Product Roadmaps: requirements, job optimization, screening, explainability |Prototype (WIP)

UX Design: Product roadmap

UX Research

DT Prototyping

Team milestones

UX Design: Concepts

Key

48 of 48

Discussion

48