Friends,

You’ll already know that helping startups with compensation is a major passion of mine. But another is supporting the growth and success of the People profession in general.

So much so, last year I took a big bet, and created an event to platform the incredible work being done in this space. It was a bid to help those building People functions to not only see how others build excellence, but also form the relationships and networks that transcend the event, too.

That event was Startup People Summit.

And in it’s inaugural year, 800 people participated, sending a clear signal that this is a space people want to be in, learning from and building relationships in.

You see, while I know many of you reading are not from the APAC region, for us, we’ve never really had a place to call our own. Sure, we’ve had events for HR at big corporates, but anyone building inside a startup or scale-up knows the work is fundamentally different.

(And then rebuilding it 12 months later because the company has changed so much)

That’s the kind of content that Startup People Summit is serving up.

So if you’re building in this kind of world — this event is for you. And it’s happening on the 3rd of September.

I’ve been heads down with my team building this out, and the content and design this year is already giving me goosebumps. It’s that good.

You’ll hear snippets of the speakers and event from me here, of course, but your best place to get all the details will be over on the SPS insiders list. Hope to see you in there.

Enjoy ✌

Matt

In case you’re new here (or just missed it) here’s the past three editions of the FNDN Series:

Grant sizes by level, targeted versus broad-based split, refresh patterns at later stages, and how companies are handling option pool maths in the current market.

Built from data across thousands of growth-stage startups.

This is the report you need if your current program was designed in a different market, or if you need to know what good looks like in 2026.

Know a startup Head of People looking for answers 🙋 why not forward this to them for some instant karma? ✨

The fastest way to break a pay-for-performance system is to launch pay and performance at the same time. Most startups do exactly this. They build a rating scale linked to a merit matrix and roll the whole thing out in a single cycle. By the second cycle, half the org thinks the ratings are rigged and the other half is gaming the goals.

There is a sequence that works. It takes longer up front, but it pays off everywhere else.

Two operators who have rolled out pay-for-performance at scale recently landed on the same answer to the chicken-and-egg question. Amee Parekh, founder of Stello AI and former SVP of HR at Hims & Hers and Uber Eats, and Mark Lewis, CEO of Crewmojo, both put performance first. Pay comes later.

Here is what their answer looks like in practice if you’re a growing startup.

Mark's framing was pretty clear: if you can’t define performance, what are you actually paying for?

And it’s true. I see a lot of companies skip this step and go straight to a set of comp ranges that are linked to performance (which never really gets defined).

The exec team assumes "high performance" means the same thing to everyone, builds a comp model on top of that assumption, and finds out at the first calibration meeting it does not. 

In reality; your CFO is rating people on revenue impact, your CTO is rating on technical depth, and both think their definition is the obvious one.

When the exec team disagrees on the definition, that disagreement ripples through every manager conversation downstream. Ratings start to look more like reflections of which exec the manager reports up to rather than a measure of how the employee actually performed.

Amee made the same point. The exercise to get your exec team to agree on what "high performance" actually means in your business takes longer than people expect, and it needs to happen before any rating scale gets attached to a pay decision.

There’s four things to define, ideally by job family rather than role:

Once that is documented, publish it for managers and employees alike. Perfect comes later. Start by getting everyone working off the same definition before any money is on the table.

A common alternative is to scrap ratings entirely. Amee pointed to Expedia, which at one stage removed its rating system altogether. In year one, employees and managers loved it. By year two and three, the senior team wanted to know who their top talent was and who to invest in for succession, and there was no data available to answer either question.

Both experts landed on the same sequencing rule, and few startups follow it. Roll out performance first, and run it for one to two cycles before connecting it to pay. Amee was specific on the duration: a good year or two of performance only, so the system has time to iterate before the comp consequences land.

The first cycle of any performance system surfaces problems. Manager bias shows up. Definitions of "good" need refinement. Goals get set that are either too easy or unmeasurable, and employees push back on feedback they did not see coming. All of those are the cycle doing its job.

If pay is already attached, every one of those problems becomes a comp dispute.

Mark shared a hard lesson from a previous company, where performance bonuses were tied to organisation-wide results. The business had a standout year, the bonuses flowed automatically to every individual, and the company ended up overpaying because the model assumed everyone had contributed equally to the outsized result. They had not. The fix came later, when the team got more granular about who was actually driving the impact. Linking pay to a half-built definition of performance bought them an expensive lesson.

There is a smaller version of the same trap heaps of companies fall into: calling everyone a high performer while delivering a flat increase across the board. The mediocre raise tells employees the truth within one cycle, and Amee's observation kicks in. The strongest performers start to baseline their effort to the rest of the team.

Run the performance cycle without the pay link, and treat the manager pushback and the goal-setting failures as data. Use them to refine the definitions and fix the calibration process.

Then, once managers trust the inputs and employees believe the outputs, attach the pay.

Once you have run a cycle or two, you need two things in your back pocket before tying pay to performance.

The first is a published definition of high performance, broken down by job family. Engineering performance does not look the same as marketing performance. A single company-wide definition creates un-winnable calibration disputes. Job family definitions give your managers something concrete to point at.

The second is goal calibration data from at least one cycle. This is evidence that when goals were set across teams, the levels of challenge were roughly comparable. A salesperson with a stretch quota and a marketer with a deliverable that finishes in week two cannot sit on the same rating scale unless someone has done the work to compare them at goal-setting time.

Mark cited a calibration practice that does the heavy lifting upfront:

It costs little upfront and saves real time at calibration. The practice also closes a common manager failure mode: setting low goals that are easy to overdeliver, then claiming top ratings on the back of them.

Without those two artefacts (definition by job family, goal calibration data) any pay-for-performance formula will produce outcomes managers cannot defend and employees do not believe. In Mark's words, people spot unfairness fast, and disengagement follows almost as fast.

The sequence is what most startups stumble. Build the definition first, then run a performance-only cycle to surface what the system needs. Use that data to refine the definitions and tighten the goals. Connect pay last.

The hurdle is low for companies that want to adopt this today. They can start by writing down what "high performance" means in their business, by job family, and putting it in front of their exec team. If they cannot agree on the definition, the pay-for-performance system in place right now is incomplete.

It’s critical to get agreement first before building the rest.

If you enjoyed this post or know someone who may find it useful, please share it with them and encourage them to subscribe.

This piece is based on a session that ran at Startup People Summit 2025.

There are 40+ more sessions (just as practical as this one) in the On Demand library.

Everything from: Compensation frameworks, culture operating systems, high-signal engagement programs, change management that sticks, AI-powered people ops, and unusually honest conversations about what CEOs aren't telling their Head of People.

A roundup of the most interesting stuff from the week:

When you’re ready, here’s three ways I can help you:

1. Tools & resources
Resources and tools that give you what you need to build your own startup compensation practices.

2. Comp consulting
Building startup compensation practices that are clear, fair and competitive.

3. Startup People Summit
A 1-day annual event for People professionals in scaling companies. Creating the playbook for startup people practices. Grab recordings from past events, or subscribe to join the next summit.