How to Measure SEO ROI: A Board-Level Framework

Most organic programmes are reported in a language the business does not speak: positions, impressions, domain scores. The board nods and moves on to the paid-media slide, where the numbers are in pounds and euros. The result is that organic — usually the most efficient channel a company has — is under-funded because nobody can say what it returns.

This framework fixes the language problem. It measures organic search in four tiers, from visibility down to profit; it sets a baseline that survives scrutiny; it uses attribution honestly and incrementality where it can; and it produces a return-on-investment figure with a method note attached. It is the approach behind our Measurement & Intelligence work and the reason we insist on agreeing the measurement plan before any work starts. The thinking behind it is in From Rankings to Demand; this is the applied version.

Why ranking reports get ignored

A ranking report answers a question nobody on the board asked. Position three for a keyword is not a business outcome; it is a condition that may or may not produce one. Impressions are an audience estimate, not demand. A ‘domain authority’ score is a third party’s opinion of your link profile, with no unit.

Boards fund what they can evaluate. Paid media is evaluated easily because it comes with a cost, a click and a conversion in the same system. Organic has the click and the conversion, but its cost is spread across people, content, engineering and an agency, and its effect is delayed and compounding. The framework below makes those things explicit so that organic can be evaluated on the same terms — and, usually, wins.

The four tiers of organic measurement

Each tier answers a different question for a different audience. Practitioners live in tiers one and two; the board lives in tiers three and four. A good report shows all four and connects them, so that a movement at the top can be traced to a result at the bottom.

Build the baseline before you change anything

An ROI figure is only as defensible as the baseline it is measured against. The baseline is set before the programme starts, written down, and not revised afterwards. Five steps:

1. 01

   Take 13 months of history

   Thirteen months gives a full year plus the same month a year earlier, so seasonality can be seen. Use Search Console (16 months available) and analytics. Export it; do not rely on the interfaces, which change.

2. 02

   Split branded from non-branded

   Branded organic traffic is demand you already had; the programme’s job is mostly non-branded. Classify queries by brand terms and misspellings in Search Console, and use the split as the primary lens. Branded growth is reported separately as a signal of awareness.

3. 03

   Group pages by template and purpose

   Product, category, service, article, landing, support. Page groups are stable when individual URLs are not, and they map to the work you will do. Every number in the report is by page group.

4. 04

   Define conversions and their values

   Agree with finance what counts — a purchase, a qualified lead, a trial — and what each is worth on average. If lead-to-customer conversion and deal size are known, use them to value leads; if not, say so and use a stated assumption.

5. 05

   Write down the confounders

   Planned product launches, pricing changes, paid-media changes, seasonality, market events, tracking changes. When results move, this list is the first thing to check. It also stops the programme being credited for a launch it had nothing to do with.

Attribution you can defend

Attribution models allocate a conversion to the touchpoints that preceded it. None of them is true; each is useful for something. The practical position for organic is to report the conservative model the finance system uses, show the fuller picture beside it, and be explicit about what each one leaves out.

• Last-click (or last non-direct click) credits the final touch. It understates organic, which often opens a journey that paid or direct closes. Report it because it is what most finance systems and dashboards show, and because a number that agrees with finance is a number that gets believed.
• Data-driven or position-based models spread credit across touches. Report the organic share under one of these as the ‘fuller view’, with the model named.
• Assisted conversions — journeys where organic appeared but did not close — show the channel’s role in demand creation. Report the count and the ratio of assisted to last-click.
• Customer-level views — the share of new customers whose first known touch was organic, and how those customers retain — are the most persuasive for subscription and considered-purchase businesses.

Incrementality: the three tests worth running

Attribution describes who touched what. Incrementality asks the question the CFO actually has: what happened that would not have happened without the programme? Organic is harder to test than paid because you cannot switch it off, but three designs work in practice.

1. 01

   Pre/post with a control

   Compare the treated page group before and after the work against an untreated group with similar seasonality — another section of the site, or the same section in a market you did not touch. The control removes the market trend from the estimate.

2. 02

   Page-group holdout

   Deliberately leave a comparable set of pages untouched for a period while you work on the rest. The difference in growth between the two is the programme’s effect, with the usual caveats about spill-over through internal links.

3. 03

   Geo or market holdout

   For international businesses: roll the work out to some markets first and compare them with the markets that wait. This is the cleanest organic test available and is worth designing the roll-out around.

The ROI formula, with a worked example

Return on investment for an organic programme over a period is:

• Incremental means the part attributable to the programme after removing what would have happened anyway — from the incrementality estimate, or from the conservative attribution figure if no test was possible, labelled as such.
• Gross profit, not revenue. Revenue flatters every channel; the board thinks in margin.
• Programme cost includes the agency or consultancy fee, internal salaries for the time spent, content production, engineering time, tools and any paid promotion of content. Leaving out internal cost is the most common way to overstate organic ROI, and it is noticed.

Two refinements are worth adding once the basic figure is accepted. First, payback period: the month in which cumulative incremental gross profit passes cumulative cost — organic typically pays back later than paid but keeps paying. Second, customer lifetime value for businesses with retention: organic-acquired customers often retain differently from paid-acquired ones, and the difference belongs in the calculation.

Organic is an asset, not a campaign

A paid campaign stops producing the day it stops being funded. An organic programme builds pages, structure and authority that keep producing, with maintenance, for years. Period ROI misses this, which is why organic looks weaker than it is in the first year and stronger than expected in the third.

Two secondary lenses capture it without abandoning the primary figure:

• Replacement cost. What the same non-branded clicks would cost if bought as paid search, using your own paid-search cost-per-click by query group. It is not profit and should not be presented as such, but it shows the scale of what the asset delivers and what it would cost to replace.
• Trend. Incremental gross profit by quarter, plotted with cumulative cost. The shape of that chart — cost roughly flat, return rising — is the argument for organic that the single-period ROI cannot make.

What to report, to whom, how often

• Every figure carries a source and a method note. ‘Organic revenue (last non-direct click, analytics, page groups A–D)’ — not ‘organic revenue’.
• Report in the units the business already uses: its currency, its fiscal periods, its conversion names. Do not make the board learn yours.
• Show baseline, actual and plan on the same chart. A number without a comparison is a fact without a meaning.
• End every board report with a decision. Reports that ask for nothing are filed; reports that ask for a decision are read.

When the number is bad

Sometimes the framework produces an ROI that is negative or unimpressive. That is the framework working. The response is diagnosis, not a different model. Work down the tiers: if visibility rose but demand did not, the query set is wrong or the intent is not commercial; if demand rose but revenue did not, the conversion path or the page groups are wrong; if revenue rose but ROI is poor, the cost side needs examination or the timeframe is too short for a compounding channel.

Stating a bad number plainly, with its cause and the proposed correction, does more for the programme’s credibility than any amount of favourable attribution. It is also the point at which a company should be willing to hear that organic is not the right investment for it at this time — which is a conclusion we have reached with prospective clients, and say so in a first conversation.

The bottom line

• Measure organic in four tiers and lead with efficiency and revenue; rankings explain, they do not justify.
• Set a 13-month, brand-split, page-grouped baseline with defined conversions and written confounders before the work starts.
• Report the conservative attribution figure alongside the fuller view, and run an incrementality test whenever the site or the markets allow one.
• ROI = (incremental gross profit − full programme cost) ÷ programme cost, with every line sourced and annotated.
• Show trend and replacement cost as context for a compounding asset — after the real number, never instead of it.

Frequently asked questions

What is a good ROI for SEO?

There is no universal benchmark, because margin, sales cycle and starting point vary so much. A more useful question is whether organic ROI beats the company’s next-best use of the same money — usually paid acquisition — over a two- to three-year horizon, and whether it is improving quarter on quarter. In most businesses with real demand for what they sell, it does and it is.

How long before SEO ROI can be measured?

Tiers one and two move within one to three months of the work. Revenue effects large enough to measure against a baseline usually take two to four quarters. Report from the first month, but do not draw an ROI conclusion before the third quarter unless the movement is unambiguous.

How do I measure SEO ROI for B2B with long sales cycles?

Measure pipeline as well as closed revenue, with stage-weighted values agreed with sales. Track organic-sourced opportunities through the CRM by first touch and by last touch, and report both. Use cohort views — deals opened in a quarter and their eventual outcome — rather than period revenue alone.

Should branded search traffic count towards SEO ROI?

Report it separately. Branded demand reflects the whole company’s marketing and reputation; crediting it to the organic programme invites the challenge that it would have come anyway. The exception is where the programme demonstrably created the brand awareness — for example through citations on answer surfaces — and even then, label it.

What if our analytics data is unreliable?

Fix the measurement before starting the programme; it is part of the work. Verify conversion tracking, remove internal traffic, align analytics and CRM definitions, and document what cannot be trusted. Reporting a precise ROI from data known to be wrong is worse than reporting a range from data known to be right.

Can an agency’s reporting be trusted to measure its own ROI?

Only if the baseline, the definitions and the method were agreed in advance and the data lives in your systems, not theirs. That is the standard to hold any agency to — including us — and it is one of the questions in our guide to choosing an SEO agency.

How this guide was put together

• The framework is the one Lucidens uses in its Prove stage. The worked example uses invented round figures to show the structure of the calculation; it is not drawn from a client engagement.
• Incrementality designs are described in general terms; their precision depends on sample size, seasonality and the availability of a genuine control.

Sources

1. Google Analytics Help — Attribution models overview
2. Google Search Console Help — Performance report (Search results)
3. Lucidens — From Rankings to Demand
4. Lucidens — The Lucidens Method, stage four: Prove
5. Lucidens — Measurement & Intelligence service