Measuring Employment Brand & Attraction

Now available: People Analytics For Dummies

This is deep dive into measurement of employment brand as it relates to "Attraction". Attraction is one of three "A"s (Attraction, Activation and Attrition) that form the primary mathematics of people in organizations - as defined in my philosophy of Lean People Analytics, which I have been doing quite of work on over the last few months. I have taken a minor detour from explaining the application of lean to people analytics to dive in on Talent Acquisition Analytics.

When I say brand, what comes to mind? You probably think about familiar companies like Apple, Google, Microsoft, Facebook, Amazon, Coca-Cola, Samsung, Disney, Toyota, McDonalds, GE, Intel, Louis Vuitton, IBM, Nike, BMW, Marlboro, Honda... Or other companies that you have a strong association to.

When I say employment brand what comes to mind? 

If your mind responds to this question like mine your thoughts here may be cloudy. Why is that?

For the companies I know well I can’t differentiate the employment brand much from the corporate brand. For most companies I can’t think of any employment brand at all.

I have some thoughts on a few of the outstanding employers on the list above. When I think of Google I think of bean bag chairs, free lunches and geeks having a good time. With GE, I think of Jack Welch and Performance. Disney: I think of Walt’s imaginations and Micky’s enthusiasm and fun. Apple: I think of Steve Jobs dedication to innovation and quality. IBM: I think of dark blue suits.

“What do you do?” or “Where do you work?” may be one of the most frequent questions we get when we are not working. When I worked at Google I was proud to answer this question. I would always get a surprised gaze and vigorous questions. I was somebody. After I left, not so much. For the first few years after I left I became depressed. Working at Google said something about me, working at those other places didn’t say the same things. I had to find myself again.

If you walk into a grocery store to buy laundry detergent (a basic commodity) you will likely buy a brand you recognize. All things equal, people prefer brands they know rather than brands they have not heard of. If we want to add some complexity, if you are cost sensitive you may be more likely to choose based on price more than on your recognition of the brand. If two brands are equal to you in your mind, then you might differentiate on price or the words that it says on the front of the container. Upon careful inspection the way we make decisions seems kind of silly. They put fresh on it. I’d rather have fresh than dingy. Why do we believe this? We seem to go with the information we have.

We progress through stages. We begin unaware. We become familiar. We form opinions, experiences and attitudes through repeated exposure. Finally, we form association, commitment and engagement.

This “journey” may be applied to employment brand also. Consider that the employment brand journey is likely started through the conversations recruiters have with candidates or from the emotional experiences candidates share with friends and loved ones after interviewing with a company.

Recruiters are on the front lines of employment brand. What is your brand? We are an “equal opportunity employer”? Not exactly differentiating. Furthermore hasn’t that been the law for 50 years? Hi, I'm Bob, I don't abuse children. If you feel the need to say it it sort of makes me not trust you.

Measuring Brand Awareness

You measure brand awareness using two approaches:

Unaided awareness: have a neutral party ask potential future candidates in key job families what they think the best places for them to work would be.

Aided awareness: have a neutral party ask potential candidates in key job families to rank the best potential places to work from a list that is provided.

You can start your aided awareness list with your closest product competitors in your industry, but keep in mind that many types of talent can drift between product classes and industries. Your actual talent competitors may be much broader than the companies that make the same things you make in your industry. If anything, the geography of corporate headquarters or key sites may be more likely to determine people's awareness and interest in working for a company than anything else.

You might begin by looking at top 5 prior employers among your employees and by looking at the top 5 employers where your employees go after exiting your company. For practicality and focus, I suggest you do this exercise among key job families and key employees.

Unaided Awareness

Have a neutral party ask a randomly selected set of potential future candidates in key job families what they think the best places for them to work would be. Unfortunately, because this is open-ended question this will result in a somewhat scrappy list of names. The scrappy list will need to be coded into to official names. With this you can calculate the frequency of response by official name.

If you run the same unaided awareness campaign again in 3 months or 6 months or 1 year you can see how the frequencies change. You will do this on a regular basis so when things happen (expected or unexpected) you can measure the impact that had. You also can measure the impact of planned company announcements and events. You can deliberately plan events and invite target future potential candidates and measure the impact of those events too. In fact, those events may be your ticket to have access to potential future candidates to begin asking and measuring brand and attraction.

When living in Austin, Texas I would attend software engineer meet-ups put on by the company Indeed, where to my amazement they would describe behind the scenes details on their matching algorithms or dev ops environment. My impression of the company improved over time - as I am sure it did for the hundreds of other Software Engineers that showed for the presentations, networking, pizza and beer on the Indeed campus. 

Circling back. The procedure is to ask your group of randomly selected potential future candidates in key job families what they think the best places for them to work would be. Count the number times each company has been mentioned and convert that total to a percentage. Then put confidence intervals around each percentage.

The confidence interval tells you if you were to poll a different group of people what is the likely range of values the average would fall in. The tighter the range the more confident you should be in the average. To improve your confidence in the average (tightening the range of the confidence interval) you need to poll more people. 

The beauty of mathematics is that you can determine what your level of confidence should be from the characteristics of data you have collected by itself! Isn’t that amazing? My math-oriented friends are rolling their eyes.

In a survey of employment brand awareness among 100 Software Engineers identified on LinkedIn that have some experience writing bios level code for hardware companies, when asked the generic question, “What are top five employers you would likely consider for your next career move?”.

You might learn:

• 70% say Google,
• 60% say Apple,
• 55% say Samsung,
• 20% say Dell, and
• 10% say Pure Storage.

! This is a totally fictitious example !

In the fictitious example, the 95% confidence interval for Dell may be between 14% and 28%. This means that if you ran the same survey again Dell is 95% likely to be somewhere between 14% and 28%, but not necessarily the same (20%).  There is a 5% chance you could get something higher or lower than that range.

Confidence intervals may show you that your categories overlap such that it is possible they change order. Be careful when using data to rank order things. Without careful analysis you should not assume the intervals are equal or that the precise order matters. Clearly in this example the difference between the top and the bottom of the list matters. The only way to increase confidence in the order is to survey more people until the ranges are tighter.

I will have to write another post on confidence intervals, but for many people this isn’t very difficult. Let me know if you need help. It’s no big deal.

Aided Awareness

With aided awareness you provide a list of potential employers to the candidates and ask them to identify the ones they would be likely to consider for their next job move. 

Here is a retail example: 

“What are top five retail employers you would likely consider for your next career move?” Check any that apply.

•  Amazon
• Costco
•  CVS
• Home Depot
• Kroger
• PetSmart
• Target
• Wal-Mart
•  Walgreens
• Whole Foods
• Other:

You should add a few employers that are wacky choices to see how many people select them. This will give you a relative point of comparison and/or provide a measurement of how much people are really paying attention.

Like you did with unaided recall, count the number of times each company is selected convert that total to a percentage, then put confidence intervals around each percentage.

The main distinction between unaided and aided awareness is that unaided measures top of mind brand and aided measures familiarity. These are slightly different concepts.

Measuring Brand Among Product Class and Job Family

General awareness of company name is one thing, having an appreciation for brand awareness among specialized product or service job families is much richer and useful. 

For example, at this stage Google has search, mobile operating system, enterprise (Gmail, etc.) and even devices. There is a unique world of people who do work somewhere in these categories and their appreciation for Google relative to other options may vary considerably. By targeting your surveys to individuals in specific job families and/or geographic location, you can develop a way of measuring how you are doing with employment brand among specific classes of jobs that are important to your company’s future success.

If you have selected a sample of potential future candidates for a data source like LinkedIn you may add to your dataset what you know about them (jobs, prior employers, education, gender, etc.) and/or you can add a series of quick questions to have them classify themselves.

Measuring Past Exposure to Brand

Examples:

• Have you ever used the products and services of (insert company name?
• How familiar with (insert company name) are you?
• Have you ever been to an event sponsored by (insert company name)?
•  Do you know anyone who works at (insert company name)?
• Have you ever been approached by a recruiter at (insert company name)?
• Have you ever applied for a job at (insert company name)?

Measuring How People Learn About Employment Opportunities & Brand

Examples:

• How did you find out about your current job?
• What professional websites or blogs do you follow?
• What periodicals and magazines do you read on a regular basis?
• What professional associations or meet-up groups do you regularly participate in?
• What websites do you use to learn about or look for job opportunities?

If you have regular channels you use to reach potential candidates or build employment brand then you may provide an aided select all that apply list.

Measuring Brand Favorability

Measuring Overall Brand Favorability and Intent

Brand awareness measures just that, awareness, but you also want to measure the strength of opinion about that (positive or negative). One of the more powerful ways to measure favorability is with intent oriented questions.

Examples:

• On a scale of 0 to 10, how likely are you to seriously consider a new job opportunity in the next year (insert company)?
• On a scale of 0 to 10, how likely are you to consider a job opportunity at (insert company)?

Here is an example of how you can measure relative strength: (coincidental examples provided make up the auto manufacturing)

 “On a scale of 0 to 10, how likely are you to consider a job opportunity with following companies?

• Audi
• BMW
• Daimler
• Fiat Chrysler
• Ford
• GM
• Honda
• Nissan
• Tesla
• Toyota

Measuring Key Employment Features

Categorizing Your Employment Brand Concept

After getting a general sense of relative brand strength, the next deeper level of analysis is to develop an appreciation for words people associate to your employee brand. You may want to do this periodically among current employees as well as others who have been exposed to your company.

A concept analysis identifies the words people associate with your employment brand.

To measure what words people have associated with your employment brand, again you can use Unaided and Aided recall measurement devices.

Unaided

1.    Ask the survey taker what words come to mind when they think of your company. Have them list as many as they can.

2.    When all responses have been collected classify and recode like terms by category.

3.    Count the number of responses by category.

4.    Identify the most frequently provided categories.

Aided

1.    Construct a list of key company concepts, some positive and some negative and put them on a list.

2.    Make sure to include key company values and brand pillars, if you have them.

3.    Ask the survey taker to simply select the ones they think best describe your company.

For example:

·      Arrogant

·      Conservative

·      Creative

·      Diverse

·      Ethical

·      Friendly

·      Fun

·      Innovative

·      Intelligent

·      Intimidating

·      Performance

·      Professional

·      Quality

·      Successful

·      Snobby

·      Traditional

·      Trustworthy

·      Unethical

Measuring Key Brand Concepts

The next deeper level of analysis involves getting an appreciation for key brand concepts. While you may discover other important concepts, four generally central concepts are quality, value, emotional attraction and repulsion.

Quality: how successful the company is (or is likely to be), perceptions of the quality of the products & services, and perceptions of the quality of the company culture and work environment.

Value: perceptions about pay and/or about career opportunity. Keep in mind that the long-term value of a career is 10-100 times the dollar value of current annual pay, so current pay is not the only differentiator of value if you can produce an appreciation for your commitment to produce long-term value for the candidate through support of their long-term career objectives. This is an explicit strategy used in some typically low paying retail or restaurant job environments. See In-N-Outs People Strategy Breaks The Universe Opportunity for promotion created through business growth has value. However, it has to be expressed and believable – there must be demonstrable commitment.

Emotional Attraction or Repulsion: Pride and Shame. Love or Ick Factor. Perceptions of association to the vision, mission, values and culture.

Scale

I recommend you use a consistent scale to measure a range of topics you want to index together, compare to one another or associate with other measurable items or outcomes to identify “key drivers”. Using a 5 Point Likert Agreement Scale survey, provide a series of statements.

Here is a list statements that allow you to infer the survey takers perception about key aspects of their current employer’s brand:

Based on everything I know, I think that…

• (Company Name) is in a position to really succeed over the next three years.
•  (Company Name) products and services are generally as good as, or better than, competitors.
• “I am proud to work for (Company Name)”.
•  (Company Name) communicates a vision for the future that is inspiring.
• (Company Name) demonstrates a commitment to integrity in mission, values and actions.
•  “Overall, I think that my long-term career goals can be met at (Company Name)”
•  I would be thrilled to be working at (Company Name) 12 months from now.
• I would recommend (Company Name) to a friend as a great place to work.

After asking about the current employer then ask for their perception about YOUR company:

Based on everything I know, I think that…

• (YOUR Company Name) is in a position to really succeed over the next three years.
• (YOUR Company Name) products and services are generally as good as, or better than, competitors.
•  “I’d be proud to work for (YOUR Company Name).”
• (YOUR Company Name) communicates a vision for the future that is inspiring.
• (YOUR Company Name) demonstrates a commitment to integrity in mission, values and actions.
• “Overall, I think that my long-term career goals can be met at (YOUR Company Name)”
• I would be thrilled to be working at (YOUR Company Name) 12 months from now.
• I would recommend (YOUR Company Name) to a friend as a great place to work.

Measuring Key Employment Relationship Features

As you progress on the brand journey, it is useful to measure the current ideas, beliefs, and associations that potential candidates have toward their current employment experience, their ideal employment experience and what they may know or think about your company and current job opportunity.

Use a series of statements to have survey takers identify the features of their current employment experience. Again, you’ll use a 5 Point Likert Agreement Scale.

Please use the provided scale to rate your agreement with the following statements as they relate to their current (or last) employment experience:

• I believe strongly (Insert Company)'s mission & values.
• The work I'm doing at (Insert Company) has personal meaning to me.
• I have the opportunity to do what I do best in my work at (Insert Company).
• My objectives make an important contribution to the objectives of the broader team.
• I think my workload is reasonable for my role.
• I have access to the things I need to do my job well.
• My total compensation package is as good or better than with what I could get at most other companies.
•  I believe I am fairly compensated for the level of contribution I make.
• How pay is determined at (insert company) is fair.
• I know my possible career advancement paths at (Insert Company)
• The benefits and perks at (Insert Company) communicates that company leaders cares about employees.
• My current benefits package meets my and my family’s needs.
• I would recommend my manager to others.
•  I feel safe taking risks at work.
• I am treated with respect.
• I can speak up when I disagree.
• I have a close and trusting relationship with someone at work.
•  I fit well in the (insert company) employee culture.

Figure Brand-2: Use a series of statements to have survey takers identify the employee experience features that are important in considering a new opportunity ….

Please use the provided scale to rate the level of importance of each statement to you in considering a new employment opportunity:

• Company mission & values
• Personal meaning from the work I do.
• Opportunity to do what I do best in my work
• Making an important contribution to the objectives of the broader team
• Workload & work life balance
• Technical resources & support
• Total current compensation
• Reward for additional discretionary effort
• Relative fairness of compensation
• Commitment to your long-term career growth
• Benefits (health insurance, 401k)
• Perks (onsite amenities)
• Quality of manager
• Opportunity to take risks
• Respect from others
• Ability to speak up
• Friendships at work
• Employee Culture

Analyzing the Key Drivers of Attraction and Repulsion

The primary value in having survey takers rate a number of items on a survey is NOT to evaluate response to each item, but rather to evaluate the response to each item in association with some before or after action you care about. You don’t need a potentially infinite list of things to try to be or to try to get better at. What you need is a list of items that are important to focus on in order to drive an action or outcome you care about. 

Having collected the response to a number of employment features, next you should evaluate the correlation between employment features on overall brand favorability among target candidates in key jobs. Then, you should evaluate the correlation between employment features and likelihood to consider a new opportunity among target candidates in key jobs.

You need to understand what features have a disproportionately large impact on either attraction to a new opportunity and/or repulsion from a current employment situation.

Why is this Key Driver Analysis important and how it is useful?

First of all, if you understand the most important features, you can go to work on being better at these things to beat your competition. Second, if you understand these features you can instruct recruiters on how to carefully identify them and use them to actively sell the differentiating characteristics of opportunity at your company. Finally, these frame the differentiating employee stories you need to tell through advertising channels, event speakers, blogs and so forth.

How do you conduct Key Driver Analysis?

You use a statistical technique called “Multiple Regression”.

Multiple regression is used to explain the relationship between one continuous dependent outcome variable (y) and two or more independent variables (x1, x2, x3…) 

In our brand example here our (y) variable could be the numerical response of any of the following:

On a scale of 0 to 10, how likely are you to seriously consider a new job opportunity in the next year (insert company)? (YI)

On a scale of 0 to 10, how likely are you to consider a job opportunity at (insert company)? (Y2)

Theoretically you can use any item on the survey, an index of items, or any other measurable outcome you can associate to an individual survey taker as your (y), however you need to consider what you are trying to achieve, the logical relationship between measures and practicalities.

In our brand example here, Y1 above, your (x) variables could be the numerical responses to the questions about their current employment situation.

In the brand example here, Y2 above, your (x) variables could be the numerical responses to the questions about your company

What does Multiple Regression Do?

Multiple regression is employed to understand how multiple independent variables (x) are statistically related to the dependent variable (y) and to what degree. The statistical analysis is really doing three things at once:

1.    It helps mathematically describe the form of the relationships between multiple variables (x1, x2, x3) and (y),

2.    it helps to mathematically determine how good the overall model (inclusion of all variables) is at describing or predicting the behavior of (y)

3.    and it helps to mathematically isolate the independent contribution of each (x) variable to the total variance in (y). 

A unique feature of the resulting multiple regression equation is that it can be used to predict things. After running a multiple regression analysis you can use the output in a standard formula to predict how much a change to each independent variable (x) will likely affect the dependent one (y). 

What does Multiple Regression look like?

The regression formula looks like this: y = a + bx + c. Where variable (y) is the dependent variable and (x) is the independent variable and (b) is the “correlation coefficient”, which represents the magnitude of impact of a change in (x). You will have multiple (b)(x)’s - one for each (x) variable in your model. A comparison of (b)s for each (x) variable can tell you which has the largest impact on (y). If you plug in the numerical values you want or expect for each (x) at a point in the future into the regression equation, do the math, it will output a single number that represents the predicted future (y).

Coefficient of Determination (r^2)

The coefficient of determination (denoted by R^2, pronounced "R squared") is a calculated value that is a standard output from regression analysis. It is interpreted as the proportion of the variance in the dependent variable that is predictable from the independent variable. A higher r^2 indicates the multivariate model does a good job predicting the outcome. A lower r^2 implies the outcome is unpredictable or some variables relevant to prediction are missing.

A low R^2 likely indicates you have missed important variables in your analysis so it would suggest you need to add new data or new survey items.

While a high R^2 doesn’t necessarily imply you have measured everything that matters, it can tell you if you have done a reasonable enough job to make a decent prediction.

R^2 used in the context of statistical models whose main purpose is either the prediction of future outcomes or the testing of hypotheses, on the basis of other related information. It provides a measure of how well observed outcomes are replicated by the model, based on the proportion of total variation of outcomes explained by the model. The equation takes predicted scores in a data set and compare them to the actual scores. The coefficient of determinations is the square of these two.

How To Actually Do a Multiple Regression:

1.    Identify the variables theory would suggest you include in your model for whatever it is that you want to understand or predict. In our example we want to understand the relationship between perception of key employment features and overall employment brand, as well likelihood to consider an offer.

2.    Use instruments (such as system data and/or surveys) to collect data as quantitative measurements – e.g. scaled items.

3.    Bring the data into a statistical program (R, SPSS, SAS, Stata, STATISTICA, ...) and run the appropriate regression procedure. The program will yield an R^2, a level of significance of your model and the statistical significance and magnitude of contribution of each of the included variables.

4.    Use these factors and their weighting to construct models, perceptual maps and other conceptual tools useful to communicate the findings to others and plan actions to influence intended outcomes.

The design of the initial data set, the decisions surrounding the type of regression analysis you use and how to interpret the output is usually done by a data scientist, behavioral science or mathematics professional. If you are not one of those, you probably should consult one.

Logistic Regression

The logistic regression version of regression used to estimate the probability of a categorical dependent variable based on one or more independent variables, allowing measurement of factors that increases the odds of a given categorical outcome.

A logistic regression is just like a regular Multiple Regression except the (y) variable is categorical – generally it is a condition that is binary – that is either on (1) or off (0).

An example of a binary dependent variable where the outcome is one of two possibilities is employee exit. Over 12 months of company tenure did each employee exit the company or not? If an employee exited the company in that 12 months, then it is 1 and if they stay it is 0. The options are binary, meaning for all relevant employees y can only be a 1 or a 0. 

You are applying this in reverse in your brand analysis. You collect a bunch of data, which you regress against a binary outcome: did they proceed with the opportunity (1) or not (0)?

With this you can measure the overall degree to which you understand what drives the behavior to proceed with the x variables as well as the relative contribution of each x variable, you know what doesn’t matter and given a set of x variables you can make a prediction about how likely someone is to proceed.

[ Pause there and think about how you could use that to your advantage in recruiting ]

The binary logistic regression is a probability model, so it returns a y that is between 0 and 1, where 1 represents 100% chance that the person will be interested in the opportunity you present given certain x inputs and 0 represents that given x inputs you have no idea what will happen. Generally, you are going to get something in the middle. Your job is to figure out the most important x’s. These are your employment brand and job features.

A logistic regression can also be applied to rank order thousands of candidate’s likelihood to accept a recruiter phone call so that your sourcers can prioritize who to deliver the message to and how.

This is done by taking each individual x value and plugging it into the multiple regression equation and calculating the individual y. Then you can rank order each individual by the individual y output. 

Grouped into quartiles you have a method of triage for sourcers and recruiters - triage meaning: prioritized focused action. You want to act first for those with greatest likelihood of producing a successful outcome.

Employment Brand Index

Here are 10 questions you may want to use on your annual employee survey, with candidates, or that you could poll a random sample of employees about on some periodic basis.

• I have a clear understanding of (Insert Company)'s brand identity and products
• I have a clear understanding of what differentiates (Insert Company) products from competitors
• (Insert Company) products and services are generally as good as, or better than, competitors
• (Insert Company) is in a position to really succeed over the next three years
• (Insert Company) has meaningful mission, values & products
• (Insert Company) is highly regarded by the general public
• (Insert Company) compares favorably with competitors as an attractive place to work
• (Insert Company) has a reputation for treating people equally regardless of ethnicity, gender, age, sexual orientation, religion, or disability
• I would recommend other people I know apply at (insert company)
• I have confidence in the leaders at (Insert Company)
• OPEN COMMENT QUESTION : When you think of our company brand what words comes to mind? (List as many as you can)
• OPEN COMMENT QUESTION : What concerns, if any, do you have about (Insert Company) and its future prospects?

Talent Acquisition Analytics

• “Measuring Candidate Experience”, July 17, 2018, LinkedIn.
•  “Measuring Employment Brand and Attraction”, July 14, 2018, LinkedIn.
• “Creating Competitive Advantage with Talent Acquisition Analytics”, July 12, 2018, LinkedIn.

More

• Find more of my writing here: Index of my writing on people analytics at PeopleAnalyst
• Connect with me on LinkedIn here: https://www.dhirubhai.net/in/michaelcwest
• Check out the People Analytics Community here: https://www.dhirubhai.net/groups/6663060
• Buy my book on Amazon here: People Analytics For Dummies , directly from the publisher (Wiley) here: People Analytics for Dummies , or from other places where books are sold.