A metric is anything you can measure. A KPI is a metric somebody promised to move, with a target attached and a decision waiting on it. Most marketing dashboards fail by treating every metric as a KPI, which is how teams end up reporting forty numbers and steering by none.
This is the sorting guide: the distinction that tidies everything, the money layer, the channel layer with its caveats, the vanity trap and Goodhart’s law, and how to build a KPI set that actually drives decisions.
What is the difference between Metrics & KPIs?
Metrics describe. Sessions, clicks, opens, followers: all true, all cheap to collect, and mostly idle. A thousand followers means a thousand people who might see a post. It is a fact, not a decision. At the beginning of digital era, metrics and intuition were enough for decision making. Since we live in multi-channel ecosystem and each channel / platform offers different metrics, it is difficult to grow a business with just reporting and observing those metrics.
A KPI carries a target, an owner, and a decision that changes when it misses. If nobody is setting targets and acting on misses, the honest description is that the team has metrics, not KPIs, however the dashboard is labeled. Therefore, businesses have to advance from just reporting metrics to setting up KPIs.
The promotion test is one question: if this number halved tomorrow, who does what differently by Friday? A real answer, budget moves, a campaign stops, a page gets rebuilt, earns the number KPI status. No answer means it stays a metric, consulted when diagnosing, ignored otherwise.
Digital metrics are cheap to collect and expensive to stare at. The discipline of marketing analytics is not gathering more of them; it is choosing the few that steer.
The money layer
The metrics closest to the till outrank everything else, and each carries a caveat worth knowing.
Revenue and contribution from marketing. The top line marketing influenced, and the honest version after costs. Caveat: platform-reported revenue flatters, since every platform claims generously under its own attribution rules. Trust the till, treat platform revenue as claims.
Incremental contribution per channel. The causal version of credit: what each channel added beyond what would have happened anyway. This is the number budget decisions actually need, and it comes from models and experiments, not from platform dashboards.
Customer acquisition cost, honestly loaded. All the costs of acquiring, media, content, tools, people, not just the ad spend. Blended CAC hides channel truth, so split it by channel before deciding anything.
Lifetime value against CAC. What a customer is worth over the relationship, against what they cost to get: the ratio that reprices growth. Caveat: LTV is an estimate wearing a precise number’s clothes. Use ranges and update them as cohorts mature.
Payback period. How long until a cohort’s revenue covers its acquisition cost. The cash-flow reality check on any growth story, and the number finance will ask about first.
The channel layer, with caveats attached
Search. Cost per conversion, split brand versus generic. Brand harvests demand that largely existed; generic creates it. Blending them flatters the blend.
Paid social. Incremental conversions over platform-reported ROAS, with frequency watched, how often the same person is being hit, since rising frequency with flat conversions is money meeting fatigue.
Email. Revenue per send and list health, with opens treated as directional since Apple’s Mail Privacy Protection began firing opens no human made.
Organic and content. Assisted conversions and search-volume capture, on content timelines: months, not weeks. Patience is part of the metric.
Site. Conversion rate by intent segment, never site-wide: averaging checkout visitors with blog readers produces a number that describes nobody.
The pattern: every channel’s platform metric flatters, and the honest version usually needs your own neutral layer. That reconciliation discipline is covered in campaign and advertising analytics.
The vanity trap, and Goodhart’s law
Engagement metrics, impressions, likes, followers, CTR, are the easiest numbers to move and the easiest to be fooled by. They can all rise while the business falls, and a clever team can inflate any of them on demand, which is precisely what makes them dangerous as targets.
The general principle is Goodhart’s law: when a measure becomes a target, it stops measuring. Teams optimize the number, not the thing the number was meant to represent. The symptoms are everywhere once you look: clickbait lifting CTR while conversions sag, discount-heavy revenue hitting target while margin bleeds out, follower campaigns growing an audience that never buys.
The antidote is guardrails: pair every KPI with a metric that catches the gaming. CTR with conversion rate. Revenue with margin. Lead volume with lead quality. The guardrail asks the harder question, are we actually winning, or making a number look good, and its whole job is to be the metric nobody thought to game.
Building a KPI set that drives decisions
Setting up the right KPIs is not an easy task. Furthermore, KPIs that drive decisions change overtime. You should aim for relevance, transparency within your organization and of course adaptability. Following some rules on how to build the right KPIs:
One KPI per recurring decision. Work backward from the decisions that actually recur, weekly budget calls, monthly reviews, and give each one number. Decisions without a KPI run on vibes; KPIs without a decision are decoration.
Targets with owners. A number nobody owns is a spectator sport. Every KPI gets a name attached.
Thresholds that trigger action. Define in advance what reading forces what response, so the dashboard produces moves rather than commentary.
A weekly cadence where someone acts. Marketing runs in weeks; the review should too, and its output is decisions, not attendance.
Annual pruning. KPI sets only ever grow. Once a year, delete the ones no decision consumed.
Then the deeper layer: KPIs report position, models explain cause. Marketing mix modeling and MarSci’s Causal-Driven Attribution, peer-reviewed and accepted by the Journal of Marketing Analytics, turn the KPI conversation from “what moved” to “what moved it,” which is where budget answers live: see MarSci’s solutions. To wire your KPI set to the causal layer, book a call with MarSci.
Frequently Asked Questions
A metric measures something; a KPI commits to something: a target, an owner, and a decision that changes when the number misses. All KPIs are metrics, few metrics deserve to be KPIs. The promotion test: if this number halved tomorrow, who does what differently by Friday?
The money layer first: revenue and contribution from marketing, incremental contribution per channel, honestly loaded CAC split by channel against LTV, and payback period. Channel KPIs support those with caveats attached, and engagement metrics rarely deserve KPI status at all, however easily they move.
One per recurring decision, which for most teams means a handful, not forty. Everything else stays a metric, consulted when diagnosing rather than reported weekly. Prune annually, because KPI sets only ever grow, and every added number dilutes the signal of the ones that matter.
When a measure becomes a target, it stops measuring: teams optimize the number rather than the thing it represented. Clickbait lifts CTR while conversions sag; discounting hits revenue targets while margin bleeds. The fix is pairing every KPI with a guardrail metric that catches the gaming.
