Which Number Actually Matters Right Now?

Every founder we meet is tracking metrics. Very few are tracking the right metrics for the stage they’re actually in.

Here’s how the mistake usually looks. A founder raises a pre-seed on a beautiful number — 97% model accuracy, a treatment that works in a pilot, users who love the demo. Great. Then they spend the next eighteen months polishing that same number, walk into a seed or Series A pitch leading with it, and watch the room go politely quiet.

Nothing went wrong with the metric. It just stopped being the point.

The thing nobody tells you is that KPIs aren’t fixed targets you hit once and frame on the wall. They’re a moving conversation, and the question underneath them changes as you grow. Early on, every metric is quietly answering “does this thing actually work?” A little later, it shifts to “is there early proof this becomes a real business?” And by Series A, it’s the coldest, most expensive question of all: “does this make money in a repeatable way?”

We invest at seed, so we live in the middle of that arc — the stage where “it works” has to start becoming “it works as a business.” So let’s map all three stages, across the sectors we spend our lives in, so you can see the ground move before you’re standing on it.

Before the sector detail, internalize the shape of the whole journey:

• Pre-seed proves feasibility. Can the model perform? Do users engage? Does the treatment succeed? These are validation metrics. They answer whether you’ve built something real. Growth percentages here are mostly noise — going from $1K to $10K MRR is 900% growth, impressive as a number, irrelevant as a benchmark.

• Seed proves early economics and stickiness. This is our stage. The single most important thing at seed is proof that customers want your product and will keep using it — measured through retention, engagement, and willingness to pay. Investors also start scrutinizing capital efficiency hard: your burn multiple should be under 2.0x (every dollar you burn generating at least $0.50 in new ARR), and top-performing seed companies keep 18+ months of runway at all times.

• Series A proves repeatable, scalable economics. Series A readiness in 2026 means roughly $1–2M ARR, NRR above 110%, LTV:CAC of 3:1 or better, CAC payback under 12 months, and gross margins above 70% — and critically, evidence the growth engine is repeatable and not dependent on founder-led sales alone.

That’s the whole game. Notice the middle stage — seed — is where the story changes from “cool product” to “leaky bucket or compounding machine?” A company growing 15% month-over-month with 10% monthly churn is a leaking bucket, and seed is exactly when investors start checking whether you’re filling the bucket or bailing water.

Now let’s watch this play out where it actually bites.

If you’re building workplace AI or automation, your pre-seed story is technical. Your accuracy rate (or “hit rate”) and your Human-in-the-Loop ratio — how often a person has to step in and clean up after your model — are what matter. High accuracy paired with a falling HITL ratio says your system works and is getting smarter. Latency, uptime, and false-positive rates round out the “it functions in the real world” case.

At seed, the question becomes adoption and retention. Are people using it daily, or did it become shelfware? Time to Value (how fast a new customer feels the benefit) and real engagement metrics — DAU/WAU, NPS — become the signal that you’ve found a wedge into actual workflows. This is also where you start instrumenting willingness to pay and early cohort retention, because those are the seeds of the unit economics story you’ll need next.

By Series A, nobody’s impressed the product works — they assume it does. Now it’s the CAC:LTV ratio (roughly 3:1 is the “you have a real business” line), CAC payback under a year, and proof that efficiency gains repeat across many customers, not just your favorite three. That HITL ratio you tracked at pre-seed? It should now be on a sustained march downward, proving your automation matures as you scale instead of secretly requiring an army of humans behind the curtain.

The same arc governs upskilling platforms (pre-seed: people complete the training → seed: the training produces measurable competency and behavior change → Series A: provable ROI, retention, and revenue-per-employee impact) and hybrid-work tools (pre-seed: people adopt the tool → seed: the tool creates focus time and completed work → Series A: measurable, repeatable organizational productivity).

Healthcare punishes founders who confuse these stages, because the credibility bar is higher and the money is tied to outcomes.

In value-based care, your pre-seed proof point might be HCC recapture — evidence you’re documenting patient conditions accurately. At seed, you need to show sustained clinical engagement and the early signal that outcomes are moving — patients staying enrolled, adhering, showing measurable improvement. By Series A, investors want the Medical Loss Ratio trending down and hard evidence of fewer hospital readmissions and ER visits. Translation: you graduate from “the model is sound” to “we measurably reduced the total cost of care for real patients at scale.”

Longevity and women’s health follows the same arc: pre-seed is clinical robustness and credibility; seed is retention and durable engagement (are people staying, month after month?); Series A is the LTV:CAC ratio clearing that 3:1 bar. Specialty care moves from treatment success rates (pre-seed) to consistent outcomes plus early payer traction (seed) to the unglamorous but decisive economics — payer mix, reimbursement rates, and a falling cost per encounter as you scale (Series A).

The theme across all of Care: you graduate from proving the medicine works to proving the business around the medicine works — and seed is where that hand-off begins.

Frontier tech is where the gap between stages is most seductive and most dangerous, because the pre-seed demo is often genuinely dazzling.

If you’re building AI agents, pre-seed lives and dies on Success Rate per Task — high performers target north of 95%. At seed, the conversation adds workflow entrenchment and early cost discipline — is the agent embedded deeply enough that customers can’t easily rip it out, and are you watching your compute costs from day one? This matters more than ever now that AI-native companies face deeper diligence, with investors digging into compute economics, usage depth, and workflow entrenchment alongside a credible path to sustainable gross margins. By Series A, it’s a phrase every agent founder should tattoo somewhere visible: Inference Cost per Revenue Dollar. It’s not enough that the agent completes the task; the compute cost of completing it has to shrink relative to the revenue it generates, or you’ve built an expensive magic trick instead of a margin. Worth knowing: AI-native startups often show materially higher revenue per employee than traditional SaaS, and that efficiency signal is increasingly one investors want to see prominently.

Mobility and clean tech make the jump from utilization and uptime (pre-seed) to early proof of real-world reliability and unit-level margins (seed) to Asset Payback Period, where investors want capital back in 12–18 months (Series A). And in defense/RegTech, you move from clearing the compliance gate — SOC 2, FedRAMP, the price of admission (pre-seed) — to landing initial contracts and design partners (seed) to revenue visibility through contract backlog and Net Revenue Retention, where the strongest companies clear 120% (Series A).

One thing has changed the game across every sector, and seed founders especially need to hear it: the burn multiple has become, in the words of one 2026 benchmark report, the ultimate truth serum. In 2025, 56% of seed investors called burn multiple a critical metric in their evaluation process — a fundamental shift from 2021, when growth rate alone drove term sheets.

Translation for founders raising from us and firms like us: growth alone no longer clears the bar. A company growing 150% but burning $3 to earn $1 of ARR is less fundable than one growing 80% with a 12-month CAC payback. Efficient growth beats hypergrowth now, full stop. Start tracking this early, even before it’s flattering — you can’t show a clean burn multiple trend at your next raise if you only started measuring it the month before.

You don’t need to hit your Series A metrics at seed. That’s the trap in the other direction — founders who obsess over LTV:CAC before they’ve proven anyone will stick around, optimizing economics on a thing nobody’s committed to yet. Right metric, wrong stage.

The move is to know which conversation you’re in, and to see the next one coming:

• Pre-seed: prove it works. Nail the validation metric for your sector — accuracy, engagement, clinical success, task success rate — and be honest about what’s still directional.

• Seed (where we come in): prove they stay and you’re efficient. Retention, engagement, willingness to pay, and early capital discipline. Keep 18+ months of runway, keep your burn multiple under 2.0x, and show the bucket isn’t leaking.

• Series A: prove it scales profitably and repeatably. NRR above 110%, LTV:CAC past 3:1, CAC payback under a year — and evidence the growth engine doesn’t depend on the founder personally closing every deal.

• Watch the hand-off metrics. Some numbers — HITL ratio, cost per encounter, inference cost, burn multiple — show up at every stage, but the bar changes. Early on they just need to exist; later they need to be visibly improving. A flat line there is a quiet red flag.

• Lead with the right number for the room. The single most common self-inflicted wound we see is a later-stage pitch built on an early-stage metric. Read the stage, then pick the headline.

Metrics aren’t a report card. They’re a story about what you’ve earned the right to claim. Tell the story that matches where you actually are — and start quietly gathering the evidence for the story you’ll need to tell next.

High Street Insights is where we share what we learn backing early-stage founders across the Future of Work, Future of Care, and Emerging Technologies. If a founder in your life is staring at a dashboard wondering which number actually matters right now — send this their way.