How We Made Salaries Transparent
It took us a year to roll out salary transparency. We didn’t intend for it to take a year, but the deeper we dug into what we thought should be involved, the more considerations we uncovered that should’ve been on our radar. Since this was about salaries, a deeply personal topic for most people, and we didn’t lose anything by taking a measured approach, we took our time to unwind the issue and architect how to roll out this change thoughtfully. Here’s a look at our motivations, what we did, and what we discovered along the way.
Phase I: Questioning the premise
The initial motivation was drawn from market data which seemed to indicate that compensation among people from different demographics was not equal for equal work performed. Perpetuating this inequality didn’t sit well with us, and we knew that if we didn’t make a move like this early on in our company, it was going to be more difficult to roll out later.
We saw a compensation structure which was designed to be more fair from its inception, with accountability and transparency built-in, as one of the core components of a workplace structured to support diversity and inclusion. We also sought transparent practices in general as a key component of a high-trust workplace which data show has efficiency gains for the business overall. All-around, it seemed like a no-brainer to make compensation transparent.
We did further research on what the dynamics, problems, and solutions of pay fairness in the workplace were in practice. We discovered a few things.
The economic data to support the fairness/unfairness claims were really unclear. The commonly-cited .70 on the dollar statistic of women’s salaries with respect to men’s was not actually for equal work performed — it was for the American job market as a whole. This captured variables such as job-choice, and full-time/part-time distinctions all in the same dataset. So if women on average worked in lower paid professions or fewer hours per week, this was counted as an inequality in pay. This is a real issue, however it was not the issue we were trying to address. We wanted to know - how big was the gap? How large was the problem of inequality of pay for equal work performed?
Studies which captured data about salaries in our field were difficult to find in a form that could boast conclusive results.
We did find evidence that making pay transparent or performing compensation audits made pay more equal across demographics by about 5-10%, though it was unclear whether this represented a correction of “equal pay for equal work”, or something else.
Ultimately the most complete economic analysis we found (at the time) on fairness of pay in our industry concluded that the problems that existed were less in the category of equal work not being rewarded with equal pay, but rather in the category of the structures in the workplace making opportunities to perform equal work in the first place more difficult for certain demographics. For instance - flexibility in work schedule, location, and sick days make any job significantly more (or less) manageable for individuals who are primary caregivers. Being flexible in these ways makes opportunities to do interesting work more equal for the caregiver demographic. The economic analysis indicated that these kinds of lifestyle benefits were the key component to ultimately balancing pay inequality, rather than simply trying to make sure you have “equal pay for equal work.” You first must ensure that equal work is even possible.
After-the-fact, we did discover better data on pay within our industry, comparing apples to apples. This did in fact show that pay was chronically unequal for equal work performed in our industry, however we did not know about this at the time we rolled pay transparency out.
Given the above revelations, we questioned ourselves — if the original premise that pay was not equal for equal work was false (or at least not provably true in our industry by the data we found at the time), then did we really have a reason to make pay transparent?
We presented these findings to our employees (18 at the time), and took a survey. Simply put — at the end of the day, our employees overwhelmingly wanted to have pay transparency. Regardless of whether we were able to prove there were problems with pay fairness, there was a perception that it might not be fair if it wasn’t transparent. That, in combination with our values as a company, pointed us toward taking the dive and going fully transparent.
Phase II: Architecting the change
With the decision to dive into pay transparency, by hook or by crook, we had to come up with a rollout plan. We considered just publishing it and saying “whoomp, there it is!”, then thought again of how personal and emotional compensation can be, and decided to come up with a better plan.
Here’s the crib sheet for how we ended up making this happen in practice:
Created a team to work on pay transparency rollout.
Wrote a draft rollout plan and asked for feedback from the company.
Wrote a leveling matrix (we did not yet have a leveling system at the time).
Had transparency team self-evaluate against the leveling matrix, looking for areas to edit and refine.
Published the leveling matrix to the company and asked for feedback.
Asked employees to self-evaluate against the leveling matrix.
Held 1:1s with each employee to discuss their self-evaluation, any discrepancies with their manager’s evaluation.
We took this opportunity to discuss where they should focus efforts to fill skill gaps.
Edited matrix again, based on feedback from this process, to clarify ambiguities and add missing information.
Wrote a draft of salary bands based on market data.
Compared existing employee levels and salaries to the salary bands, tweaking salary bands if needed.
Performed salary audit. Decided on any salary adjustments that were immediately appropriate.
Confirmed pay adjustments with founding team.
Ran financial impact analysis on projected adjustments.
Held 1:1s to communicate adjustments.
Made pay adjustments in payroll tool (Gusto).
Communicated updated status and schedule of final rollout of transparent salaries to company.
Created spreadsheet with names, levels, and salaries.
Published spreadsheet internally!
Had managers ask reports in 1:1s if they had thoughts or feedback on salary data.
Scheduled an internal survey to evaluate results.
There’s a lot packed in that list. Here’s a breakdown of what actually took the real introspection and work.
Determining what factors should affect salary
We experimented with using an equation-based approach to define salary a la Buffer (who has published their salary equations, and also practice compensation transparency). We found that this was more complex than we needed for our business stage and came up with a simplified version of the equation that better fit our scenario.
When compared with how salary decisions had already been made, we realized that the equation we came up with was essentially the same as the more informal process we had already been using - that is, grossly bucket individuals by skill level, then fine-tune within that bucket. Then make sure we are competitive with market rates, and pay a living wage.
So we decided to keep this process, and make it explicit:
Map a person to a level using the Dreyfus Model of Skill Acquisition
Figure out where that person landed within their skill level by comparing to other people who shared similar skills
Look at market data for other people at that skill level to draw up an offer
Defining dimensions and creating rubrics to measure them
Once we decided to keep our original process of basing salary on levels from the Dreyfus Model of Skill Acquisition, then cross-referencing with market data, we realized we needed to clarify how we measure these dimensions.
We needed a way to describe skills and behaviors as concretely and simply as possible, while still being comprehensive, to create levels with somewhat consistent interpretation. A solid leveling rubric was a key component of this. Starting with our baseline model, we tried to define each level in a way that was meaningful to our company.
First, we tried describing each level by referencing our company values. This turned out to be too abstract and subject to interpretation. Next, we tried going bottom-up — describing concrete behaviors that would collectively describe a skill level. This proved to be a much more effective approach.
The list of behaviors for each level turned out to be relatively short—seven to 15 items—and covered a range of core capabilities that we wanted to reinforce, from skill to collaboration and leadership capabilities.
To test the rubric, the salary transparency team sat down and tried evaluating themselves and each other within the team, sharing our results with each other. We filled gaps, clarified unclear wording, and made some tweaks. At the end of this exercise, we had a version we liked. We published the document to the company and asked for feedback.
The founding team had done ad hoc market research to get reference points to define salary offers for our early employees. However, we had not been exhaustive, nor had we documented our findings anywhere canonical.
The salary transparency team gathered cost-of-living data to get a baseline for salary calculations, then gathered market-rate salary data points for our location specifically (San Francisco). Though we had some distributed employees at the time, we explicitly decided not to factor in locality at the time, since it added more complexity and didn’t significantly change the numbers for us (the employees at the time were in Seattle and New York).
We then drew up a salary bands document roughly describing what salary ranges could be expected for each given level.
Measuring employees against rubrics
We initiated an employee leveling process. Each individual rated themselves against our leveling rubric, then scheduled a time to meet with their manager to share their conclusions and resolve differences of opinion.
This process ended up being much more fruitful than we anticipated. Several people radically adjusted their idea of what was expected in their role, leading to increased alignment between managers and reports. Although this wasn’t a goal of compensation transparency, it was a huge asset.
Evaluating existing salaries and making corrections
Once everyone agreed on their levels, we did a pass on existing salaries and made salary corrections.
We thought we had been fair and consistent in our salary offers to early employees. We certainly had tried to be. To our surprise, just a year after we had hired our first employee, we found several salaries in need of adjustment. We realized that due to differences between demonstrated skill, differences in negotiation approach, and skill changes over time, nearly a third of our employees required adjustments for us to feel that the numbers were fair.
This on its own was proof positive to us that exposing this data was necessary. Even though we considered ourselves extremely even-handed when giving salary offers, when we subjected ourselves to review, we found there were corrections to be made.
We suspect that this kind of drift is where many salary discrepancies begin to arise. Even if no one intends to be unfair, it’s work to do a comprehensive review of salaries, and there’s little immediate payoff to those who would perform the review if they are the senior members of the organization. It’s precisely those with the least power that benefit most from this process — at least in the short-term. Long-term, arguably, the whole company is better off.
We published salaries, levels, and name in a simple spreadsheet made internally visible to the entire company.
We waited to see what disaster would ensue. And... there wasn’t one. After all the work that went into laying the groundwork for salary transparency, we discovered that there weren’t any surprises left. We had shared our market-rate research, leveling document, and salary rollout plan as we worked. When the rollout finally happened, the reaction was, “oh, sure, that makes sense.”
Phase III - Retrospective
Two years in, here are some of the effects we’ve noticed on the business:
It makes hiring easier
Candidates can see we care about fairness and honesty. Whether they came in caring about transparency or not, this is compelling. If trust and transparency matter a lot to a candidate, salary transparency often tips them in our favor.
We attract more principled candidates
We attract more employees that care about trust and transparency, and they can see our money is where our mouth is regarding culture. We want principled employees, so this creates a virtuous cycle.
We have a higher level of trust from employees out of the gate
We share our salary and leveling data at the time we make an offer. Candidates can see clearly where they stand. This creates a higher degree of trust even before their first day on the job.
We get less haggling on offers
Because our salary and leveling data is out there, if haggling occurs it’s usually around, “Hmm, I think I might be a level 4” rather than, “Give me $10k more please.” We didn’t intend for this to happen, but it’s an interesting result.
Employees celebrate others’ raises
When we give raises, everyone can see who got what adjustment. We make adjustments to keep salaries in line with fairness and professional progress. People can look at those adjustments, and see in black and white how they make things more equitable. Seeing changes which feel just is inherently satisfying. During raise rounds, we see people celebrating others’ progress far more than mourning their own lack of raise.
It clarifies gaps in our pay and leveling systems
When our people grow faster than our salaries or our leveling rubric, this is astonishingly clear to everyone and incentivizes us to fix it.
It improves employee retention
If people leave, it is generally not due to salary. Folks know the score regarding their current salary, and the transparency helps them be patient when adjustments need to be made and haven’t yet been rolled out.
The most significant effect of the salary transparency was how it created structural proof that Truss cares about fairness, trust, and agency. This alteration of “everyday normal” is the principal result of this work, and it is exactly what we had hoped for.