How Compa levels and matches thousands of offers from the best tech companies

Comp folks, you know the drill. Every year, as fall approaches, teams get ready for annual benchmarking. A standard practice for most companies, benchmarking aims to match internal jobs to compensation data from survey providers.

You dig up, knead, and analyze tons of data — an intensive process that takes months. The submission results, which come out in the spring, play a major role in shaping your compensation and recruiting strategies.

What makes this data so valuable? It’s apples-to-apples market data. Teams work diligently to map their own jobs to a standard market rubric. As a result, companies see how their myriad pay decisions stack up against each other. It’s powerful data and the kind that Compa is excited to reimagine using offers.

Compa’s approach to job matching and leveling provides companies with the same assurance, data quality, and value their compensation teams deliver through benchmarking. Since it’s offers-based, our advanced matching provides new insights your comp teams have never seen, but certainly need for the next era of pay transparency.

Here’s how Compa matches and levels your jobs using real-time, offers-based data.

Why offers-based data

Offers-based data is inherently different from traditional employee-based data from surveys — no better, no worse, but different. It’s also largely overlooked by compensation teams.

For one, survey data inevitably lags the market since companies typically update pay once or twice a year. By the time survey data becomes available, it could be upwards of a year old.

Offers, on the other hand, are fundamentally real-time: an upstream pay signal and live buyer-seller transaction reflective of true market value. Rich with metadata like volume and acceptance rate, offers also help comp executives navigate shifting market trends.

How do we level offers-based data? Similar to traditional surveys, Compa uses a standard job architecture to map jobs, levels, and locations. Since we’re only concerned with what happens at time-of-offer — five or six pay elements — we’re able to match jobs successfully in a few weeks, and without any heavy lifting from your team.

Sound unrealistic? We get it. Offers-based data is new, so there’s a learning curve. We’re also expediting a historically lengthy and painstaking process, behind the scenes, for you. Let’s break it down.

How Compa matches levels and jobs

Three key ingredients make up Compa’s apples-to-apples market data: job architecture (consisting of job, level, career stream, and location), careful data validation, and decades of collective compensation experience.

Job architecture and onboarding

Compa has a canonical career architecture that helps us establish a baseline framework of jobs.

A job architecture, at its most basic, organizes jobs and levels; defines job families, subfamilies, titles, and codes; outlines HRIS data requirements; anchors HR processes; and creates career paths and opportunity insights for employees.

Most organizations have a standard job architecture, and survey houses certainly have one. We use ours as a starting point to reveal where there’s alignment and where there are gaps. With a basic questionnaire and a little more qualitative information — usually a survey key — we’re able to line up Compa’s job architecture to yours (Figure 01).

Sometimes everything matches nicely, and sometimes it doesn’t. If the latter is the case, we know it’s time to dig for clues. And a survey key is a good place to start. It reveals the kind of logic (or clues) behind a company’s levels and how they’re matched to a standard rubric.

If that’s not enough, we keep digging. Carefully, we piece together fragments of data that may be random scraps to you but are gems to us — sample offer letters and other internal documents — to ultimately shape a precisely leveled and matched data mosaic.

All of this is part of Compa’s onboarding process. The lift for your compensation teams is next to none compared to traditional benchmarking. No annual submissions, no poring over countless job descriptions, and no waiting.

Now let’s take a look behind the curtain.

Data validation

Compa thinks of data validation as a big, systematic clue hunt. But before any work is done we hear a few common concerns:

• We don’t have a formal job architecture. How can we participate?
• Our data is ugly. How do you make it usable?
• There’s not enough data.

Stubborn data is somewhat of the norm in compensation.

When a company lacks sufficient data or a standard job architecture, we turn to the applicant tracking system (ATS). This is where things can get messy without knowing what to look for. After all, the ATS stands alone, always fed data from recruiters and never from comp teams.

Rather than looking for ready-made data, we use your ATS as a map. We closely study job and pay information (e.g. new hire grants) and offer data (which point to job families and levels), ultimately working with you to understand which pieces in the ATS are viable clues (Figure 02).

Some degree of collapsing — merging multiple levels into one — is necessary to preserve the data. This is a standard practice in job matching. For example, Compa may discern that two levels lack sufficient differentiation, so we collapse them into one. Or sometimes one level may be a single incumbent job and indicates an outlier.

Compa, and our team of compensation vets, makes these decisions from a perspective of experience. We know benchmarking all too well, and the filter simply isn’t worth the degradation of the data.

After validation, our team meets with you to ask simply: “Did we get it right?” And truthfully, we don’t always get it right. This is a critical point in our matching. At once, you gain a bird's-eye view of our entire process and can inform where we veered off. 

For instance, maybe we’re able to map one-to-one save for a sneaky E8 role. Your survey key shows a mapping to E2/E3, which are two separate levels. We then ask if you consider your E8s a VP or SVP, and the answer to that question ultimately determines how we map to Compa’s job architecture (Figure 03).

All of this information feeds Compa’s part-AI, part-human job matching engine. The more data we process, the more intelligence our engine gains to improve mappings. This is core to our give-to-get model with Compa Index. 

And what exactly do you get? All new data stories like acceptance rates, location variance, offer volume, prevalence, and average spend immediately become available (Figure 04). Like, did you know there’s a top three US city where P2 and P3 engineers accept jobs 15% more often at a third of the cost? Certainly not the kind of stuff you get with survey data.

We always make decisions based on what’s best for the data, and when we let our engine run, we make sure it lines up with your expectations. This is why we’re able to onboard customers in as little as one month.

This is why we believe in offers.

Collective compensation experience

Before founding Compa, Charlie Franklin spent a decade in HR, mostly in compensation. In-house stints at Mercer, Juniper, and Workday helped him understand how things really work.

It’s this fundamental understanding — specifically of offers and employee pay — and a genuine care for the future of compensation that allows Compa to foster a deeply experienced, star-studded team. 

Collectively, the team at Compa has decades of experience navigating compensation deep and wide. This experience makes all the difference in our approach to job matching and leveling. Without it, we couldn’t deliver the same confidence, accountability, ownership, and security that our customers gain with our product.

What’s to come

Whether we’re talking about pay equity issues or market rates, offers represent the source of our biggest problems and our most powerful solutions.

That’s why Compa’s approach to job matching and leveling is only concerned with what happens at the time an offer is made. We essentially cut the noise by narrowing our focus on a handful of pay elements, rather than compensation as a whole.

At the same time, we’re steadily committed to data wellness. If you jumped into the product today, you’d see the dataset only covers Professional and Management levels (and excludes Support and Executive). Not because it’s not available, but because it tells the best story.

Remember when some surveys only had three levels? As we continue to grow, we expect to make more decisions like this for the sake of the data. (Think AI-led confidence scores and advanced matching algorithms.)

Maya Angelou famously said, “You can't really know where you're going until you know where you have been.” For that, we thank survey data for its service and welcome offers to take us forward.