A KPI is not simply any number a dashboard happens to produce. It is a specific, measurable indicator chosen in advance, before a campaign runs, because it would prove or disprove progress toward a stated objective — not selected afterwards from whichever numbers look best. Parmenter (2007) argues that a genuine KPI is tied to the organisation's critical success factors, measured frequently, and understood well enough that everyone can act on it, which rules out most of what a channel's own reporting dashboard surfaces by default. Parmenter himself reserves "KPI" for those frequent operational measures and calls financial outcomes such as revenue "key result indicators"; this module uses KPI in the broader sense common in marketing planning, for the indicator that proves an objective.

The discipline is to work backwards from the objective, not forwards from whatever data is easiest to pull. Module 1 give two distinct things to prove: that revenue is growing toward its target, and that distribution is expanding toward its target. Each needs its own KPI, measured on its own terms, rather than a single blended "marketing performance" number that can hide one problem behind one success. It also helps to pair a lagging indicator, which confirms an outcome after the fact (revenue, active listings), with a leading indicator, which moves first and can still be acted on (the conversion rate of meetings with supermarket buyers).

A KPI framework built this way also resists the pressure to add metrics simply because they are available. Site traffic, email open rates and social engagement can all be useful diagnostic numbers, but none of them is the KPI for a revenue or distribution objective unless a credible chain of evidence connects the two. Where that chain doesn't exist, the number belongs in a diagnostics appendix, not the headline report.

A vanity metric is a number that looks good in a report but doesn't connect to a business outcome — impressions, followers or page views counted on their own, with no line drawn to revenue, margin or distribution. Ries (2011) popularised the term, contrasting vanity metrics with actionable ones that show cause and effect clearly enough to guide a decision; the same logic applies to a marketing report, which should be built around the numbers that connect to economic value even when the activity numbers are larger and more flattering.

The pressure to report vanity metrics anyway is real and rarely comes from dishonesty alone. A campaign that moved impressions by a large multiple but revenue barely at all is an uncomfortable thing to report plainly, and a dashboard full of growing follower counts is an easy way to fill a slide while avoiding that discomfort. The discipline this module asks for is a simple audit question, applied to every metric before it goes in a report: would this number still look good if revenue had fallen? A vanity metric usually survives that question unchanged; a meaningful one moves with the business outcome it claims to represent.

This matters most exactly where the temptation is strongest — in the channels generating the most visible but least accountable activity. A large paid social or earned PR campaign can produce enormous reach and engagement numbers that say nothing on their own about whether a single viewer became a paying customer, which is precisely why the next lesson's attribution models matter: they are what connects that activity back to results the business actually needs.

Attribution models exist because a customer rarely converts after a single interaction with a brand, and something has to decide which of the several touchpoints along the way gets credit for the sale. A first-touch model gives all the credit to the very first interaction a customer had; a last-touch model gives all the credit to the final interaction immediately before conversion; multi-touch models split the credit across every touchpoint in the journey, either evenly (linear) or weighted toward the touchpoints closest to conversion (time-decay). Dalessandro et al. (2012) and Berman (2018) both show, from very different data sets, that the model chosen is not a neutral technical detail — it systematically changes which channel appears to be doing the work.

Last-touch is the easiest model to implement and the most common default, which is exactly why it is dangerous on its own: it structurally over-credits the channels that tend to sit right before a conversion, such as paid search and retargeting, because those are often the last thing a customer clicks before buying. It structurally under-credits the channels that start a journey but rarely finish it, such as earned PR coverage or a paid social post that a customer later searches for by brand name rather than clicking through directly. A multi-touch view run on the same underlying data can tell a strikingly different story, revealing that a large share of "paid search" conversions actually began with an earned or paid upper-funnel touchpoint that last-touch had assigned zero credit to.

The honest conclusion is not that one model is "right" and the others wrong, but that any single model is a simplification, and the choice of simplification changes the recommended budget. Comparing more than one model on the same journey data, rather than reporting whichever one the ad platform surfaces by default, is what turns attribution from a reporting convenience into a genuine input to next year's channel decisions.

A performance report earns trust by stating results against the targets set before the plan ran, not against a comparison invented afterwards to make the number look better. Rust, Lemon and Zeithaml (2004) make the underlying case directly: marketing spend deserves the same financial accountability as any other investment, measured against the return it was projected to deliver, not against a softer benchmark chosen once the actual return is known. That discipline applies as much to a shortfall as to a success — a report that reframes a missed target rather than stating it plainly has stopped being a measurement and started being a sales pitch to the board.

Honest reporting also means holding every objective to the same standard, not just the one that happened to go well. A plan with two objectives can show real but uneven progress — one measure roughly on pace against its target, the other some way behind — and the report's job is to say exactly that, in those terms, rather than letting the stronger number carry the weaker one. Partial progress, stated plainly, is more useful to a board making a year-two decision than a single blended figure that obscures which half of the plan still needs work.

The attribution finding from the previous lesson belongs in this report too, not as a footnote but as a recommendation: if multi-touch analysis shows an upper-funnel channel was quietly doing more than last-touch credited it for, that is evidence for shifting next year's budget, not a curiosity to mention and move past. A credible report connects the measured shortfall, the attribution finding, and a specific recommended change in the following year's spend — building directly on Module 7 this analytics work is measuring.