Leading, coincident, and lagging KPIs are the three ways a business metric can relate to time — a number that moves before the market has priced something in, one that moves at the same moment everyone else sees it, or one that only confirms what has already happened. For a thesis-tracking system to actually warn you before a stock re-rates, at least one KPI behind every thesis needs to be leading — not just relevant. Most portfolio trackers are built entirely out of the lagging kind, which is why the KPI usually agrees with the price only after it's too late to act on it.
The first pass I ever took at tracking CDSL looked like everyone else's. Investor deck slides. Quarterly revenue growth. PAT growth. Operating margin. The same narrative every other analyst covering the stock was reading off the same three pages.
It took a full rebuild to get past that — down to the actual toll-booth economics. ₹11 per folio once a company crosses its fixed billing slab. ₹3.50 per debit transaction, flat, regardless of trade size. Line items sitting in the Notes to the Financial Statements that almost nobody opens.
And even after that rebuild, the matrix I ended up with had a quiet flaw in it. Every number that validated the thesis could only be computed after the quarter had already closed.
The wrong question is whether a KPI is accurate. The right question is whether it arrives before the price does.
That's the whole idea this piece is built on. A KPI doesn't earn a place in your thesis tracker because it's relevant to the business. It earns a place based on when it reaches you — relative to when the market finds out.
What's the difference between leading, coincident, and lagging KPIs in stock investing?
Think of it the way you'd think about a fire. Smoke is a leading indicator — it shows up before the flames do, and if you catch it early enough, there's still something to do about it. The fire alarm is coincident — it goes off exactly when everyone in the building notices, at the same moment you do. No edge either way. The insurance assessor's damage report, filed a week later, is lagging — thorough, accurate, and completely useless for putting the fire out, because the building is already gone.
Most thesis-tracking sheets are full of damage reports.
Lagging — the numbers everyone reaches for first
Revenue, PAT, margins, and in CDSL's case, the two metrics I originally treated as my core validators: Revenue per Issuer and Transaction Velocity. Both are backward-looking by construction. Revenue per Issuer needs the audited annual report's Notes to the Financial Statements. Transaction Velocity needs the SEBI Bulletin's depository statistics table, summed across three months and divided by average active demat accounts — a number that doesn't exist until the quarter is already over. By the time either one moves, the concall has happened, management has already answered the volume question, and the stock has already re-rated on whatever was said. These aren't bad numbers. They're validators. Their job is to confirm the thesis is intact or broken — never to warn you first.
Coincident — numbers that move with the news, not ahead of it
A large order-win filing on the exchange. A line in the concall transcript. Data that lands in front of you at the same moment it lands in front of every other analyst holding the stock. There's no lag here, but there's no edge either. It keeps your tracking honest in real time — it just doesn't buy you a head start.
Leading — the numbers almost nobody bothers to check
This is the tier that actually earns its place in an exit-conditions trigger, and it's also the tier people give up on fastest, because the ideal leading number is usually the hardest one to get cleanly.
| KPI type | Timing relative to the market | CDSL example | What it actually gives you |
|---|---|---|---|
| Lagging | Confirms after the market has already reacted | Revenue per Issuer, Transaction Velocity | Proof the thesis held or broke — after the fact |
| Coincident | Arrives the same moment everyone else sees it | Order-win filings, concall commentary | Honest real-time tracking, no head start |
| Leading | Arrives before the market has fully priced it in | No. of equity trades (txn income proxy), demat account growth (annuity proxy) | An early, directional prompt to go dig deeper |
What do you do when the ideal leading indicator isn't available?
For CDSL's transaction income, the cleanest leading proxy turned out to be something almost embarrassingly simple: the number of equity trades, published monthly rather than quarterly. It moves before Transaction Velocity is even computable, because Transaction Velocity needs the full quarter's SEBI Bulletin data averaged against active accounts. If monthly trade counts are decelerating two months running, that's a signal to go dig — well before the quarterly numbers get around to confirming it.
The annuity side is where the honest limitation shows up. The theoretically ideal leading indicator for the folio-driven, ₹11-per-folio income stream is folio growth itself. On paper it sits in the same monthly depository statistics data. In practice, getting a clean, consistently updated read on it every single month is harder than the framework makes it look — so I stopped treating it as the primary trigger.
The next-best proxy is demat account growth. It's not a precise translation into revenue — plenty of new demat accounts never touch an issuer that's close to crossing its fee slab, so the read is directional, not exact. But it updates monthly, it's public, and it tells you whether the funnel feeding the annuity stream is still widening or has quietly started to stall.
That imprecision is the trade you're making on purpose. A directionally-useful number that reaches you a full quarter early is worth more than a perfectly accurate one that reaches you after the price has already voted. The leading KPI's job was never to be the verdict — it's to buy you time before the market prices in the answer. What you do with that time will vary. A sharper question on the next concall instead of a passive listen. A call to someone closer to the ground. A second look at management's last commentary, read more skeptically this time. There's no single move that always works, and pretending otherwise would be its own kind of false precision. The point of a leading indicator isn't to tell you what to do next. It's to make sure you still have time to figure it out.
How do you build a leading indicator into your thesis-tracking system?
Start with whatever lagging validators you already have — you probably already do, they're the easiest ones to find. For each one, ask a single question: is there any proxy, even an imperfect one, that moves before this number does?
If the honest answer is yes, that's your leading KPI. If the honest answer is no, that's not a reason to stop looking — it's a reason to write down the imperfect proxy anyway, and treat it as a trigger to investigate, not a trigger to sell. Precision was never the requirement. Timing was.
This is also the reframe worth sitting with: a thesis isn't why you bought a stock — it's what has to stay true for you to keep holding it. If nothing in your tracking sheet can tell you that truth is cracking until the market has already re-priced the stock around it, you don't have a tracking system. You have a confirmation system. It'll agree with you right after it's too late to matter.
The Thesis That Survived Everything covers the step before this one — how to name the mechanism behind a thesis so the KPIs (leading ones included) derive from it directly, instead of being whatever number was easiest to find.
When is chasing a leading KPI not worth it?
This doesn't apply evenly across every position.
- Long-execution-cycle businesses — where order-to-revenue recognition already spans multiple years — often don't need a monthly leading proxy. The lag is already known and priced into how you read the lagging numbers, so a quarterly validator may be enough.
- Tracking-sized positions — small, exploratory holdings you're not yet fully convicted on — rarely justify the effort of building a proxy KPI. That rigor is worth reserving for positions where the cost of finding out late is actually high.
- Businesses that already disclose fast — some managements publish genuinely granular monthly updates on their own. When the coincident data is already fast, chasing a separate leading proxy adds effort without adding real lead time.
- Any proxy you've started treating as certain rather than directional has stopped doing its job. A leading KPI that gets read as a verdict recreates the exact overconfidence this framework is trying to remove.