How to prove LinkedIn ROI to the board in three slides

The board wants three slides, not your dashboard

Three slides, in a fixed order: what it cost and what it produced, three independent readings of the same period, and the list of claims you are deliberately not making. That sequence works because it answers questions in the order a board actually asks them, and because it puts the weakest evidence last rather than hiding it in a footnote.

The failure mode is bringing the analytics dashboard into the room. A dashboard is built for the person who tunes the channel every week. A board meets for two hours to allocate capital across everything the company does, and nobody in that room is going to convert impressions into a funding decision on your behalf. If you present a chart, you have handed the interpretation to people whose default interpretation is scepticism.

Label every number with how much you actually know

Put a confidence rung next to every figure on the slide, as a printed label rather than a verbal caveat. Boards do not distrust small numbers, they distrust numbers of unknown provenance, and one unlabelled figure makes the audience discount the labelled ones sitting next to it.

The rule that makes the ladder worth having is a spending rule. Rungs one, two and three can support a request for budget. Rung four supports a request for a test with a stated end date and a kill condition, and nothing more. A finance director who watches you refuse to over-claim on your own weakest evidence will extend more trust to your strongest evidence than any chart would have earned.

Slide one: money in, pipeline out, nothing else

Six lines maximum, every one of them Counted except the credited opportunity line. The discipline is subtractive: if a line cannot be traced back to a billing export, a payroll allocation or a CRM stage timestamp, it does not go on slide one. Move it to slide two where its uncertainty can be labelled properly.

Illustrative only. The figures are assumptions for the worked arithmetic below and should be replaced with your own exports before anyone sees this slide.

Now do the arithmetic in front of them, and do it twice. Total programme cost is 48,000 plus 18,000, which is 66,000. Divided across all 23 credited opportunities that is roughly 2,870 per opportunity. Divided across only the 14 that carry a traced click it is roughly 4,714. Revenue of 210,000 against 66,000 of cost is a 3.2 times return on the generous reading.

Present both numbers. The generous reading is the one you believe, the conservative reading is the one that survives a hostile question, and volunteering the second before anyone asks for it is the single highest return move in the whole deck. If your cost per lead sits outside every figure you have read online, that is expected, and why published cost per lead figures disagree is worth understanding before you let anyone benchmark you against a blog post.

Slide two: three readings of the same quarter

Show the same period measured three different ways, and show where the three disagree. Agreement between three imperfect methods is the strongest honest claim available to a company without an attribution platform, and it is far more persuasive than one confident number from one system.

Three readings of one quarter. Each one is wrong in a different direction, which is exactly why you run all three.

The instinct is to reconcile the three into one number before the meeting. Resist it. A board that sees three readings converge draws the conclusion itself, and a conclusion the audience reaches is stickier than one you assert. Where the readings diverge, say which one you trust and why, in one sentence, then move on.

Building the traced and self-reported readings is a Monday morning job rather than a software purchase, and the method for it is written up in full in attributing pipeline to organic posts, including the arbitration rule for the case where the two readings name different channels for the same deal.

Slide three: the things you are not claiming

Write the disclaimers as a list on a slide, not as a sentence you mumble. This is the slide nobody else brings, and it is the reason the previous two slides get believed. A person who states the limits of their own evidence is read as competent, and a person who states none is read as selling.

Finish the slide with the decision. One sentence naming what you want, one sentence naming what would make you recommend stopping. A budget request without a stated kill condition reads as an open ended commitment, and open ended commitments are the thing boards are institutionally built to refuse.

Script the answer to the question about causation

The CFO will ask how you know LinkedIn caused any of it, and the correct answer starts by conceding that you do not. Every other answer sounds like evasion, and the moment a finance function decides you are evading, the meeting is about your credibility rather than about the channel.

We cannot prove causation, and I will not pretend otherwise. Twenty three of the sixty two opportunities we created carry a LinkedIn touch. Fourteen of those carry a click we can trace to a specific record. The other nine told us in their own words on the discovery call. Pipeline moved in the same direction as posting volume in seven of the last twelve months, which is a correlation and I am labelling it as one. I am not claiming LinkedIn created this demand. I am claiming it is currently the cheapest place we find it, and I have designed a test that would change my mind if it comes back flat.The script, adapt the figures to your own exports

Three things are happening in that script. You concede the strong claim before it is taken from you. You give the numbers in descending order of confidence so the weakest one arrives after trust is established. You end on a test, which converts the conversation from a verdict on the past into a decision about the next quarter, and a decision is something a board knows how to make.

The holdout test that upgrades an inferred number

One test moves a number from Inferred to something defensible: withhold the activity from a matched slice of your market and compare. It is the only method available to a company without an attribution platform that produces evidence of causation rather than evidence of coincidence, and it costs nothing except the discipline to leave one segment alone.

A holdout takes a full sales cycle to read, which for most B2B companies means a quarter at minimum. That is the honest cost of the method. Anyone offering you causal certainty faster than your own sales cycle is selling software, and the software has the same problem, it simply hides it behind a confident interface.

When the board still says no, and what that actually means

A no after this deck is usually a no about something other than LinkedIn, and it is worth finding out which of three things it was before you rebuild the deck. Rebuilding the presentation when the objection was about the sales cycle wastes a quarter.

• The evidence was accepted but the constraint is elsewhere. If the sales team cannot work the pipeline it already has, more pipeline is not the ask, and the board is right. Come back with a conversion argument instead of a volume argument.
• The time horizon is wrong for the company. A business that needs revenue this quarter should not be funding a channel that reads on a two quarter lag, and saying so yourself is more credible than arguing.
• The numbers were believed and judged too small. This is the only one of the three that is genuinely about LinkedIn, and the response is a cost argument rather than a volume argument: what would the same 66,000 have produced anywhere else you can actually buy.

One further check before you assume the programme is underperforming. If your reach fell during the reporting period, find out whether the platform fell with it, because a decline that matches a platform-wide decline is not a performance problem and reporting it as one costs you a budget you should have kept. The method for that comparison is in how to normalise an impressions drop, and it takes about twenty minutes.