Quick answer
The ROI of a link campaign isn't calculated with new backlinks x DR value. Add up the full cost (asset creation, data, outreach, any placement, production, management, and maintenance), then measure a chain of outcomes: references obtained, survival, referrals, organic evolution of the targeted pages, conversions, and revenue. Keep these stages separate.
To better isolate impact, define before launch a cohort of targeted pages and comparable non targeted pages, a baseline period, the allowed changes, and a measurement window. Compare the evolution, not just before versus after. The result will often stay observational: seasonality, Google updates, content, and demand can all intervene. An honest report therefore provides a range, the assumptions, and alternative explanations, never a promise of position.
Key takeaways
- The full cost includes hours and assets, not just the placement invoice.
- An acquired link, a live link, a visit, a conversion, and revenue are five different units.
- Before versus after alone confounds the campaign with demand, the algorithm, and other changes.
- A comparable control cohort improves inference, but doesn't create perfect randomization.
- Measure the audience value of paid placements independently of any assumed SEO effect.
The six level outcome model
| Level | KPI | Formula | Question |
|---|---|---|---|
| Effort | Full cost | tools + placements + hours x hourly cost + maintenance | What did we actually invest? |
| Acquisition | Placement rate | unique placements / qualified prospects | Does our pitch interest the right publishers? |
| Quality | Qualified share | compliant references / placements | Do the links respect the grid? |
| Survival | 180 day rate | links still valid and contextual / links acquired | Does the asset stay present? |
| Behavior | Referrals and conversions | events per Analytics/CRM | Does an audience act? |
| Organic | Target vs control delta | target evolution minus control evolution |
Do targeted pages outperform a comparable group? |
Attributed revenue comes after these levels. If your sales cycle lasts six months, a four week window can't conclude on final ROI.
Calculating the full cost
Use this formula:
full cost = data + asset creation + outreach + placements
+ editorial production + legal/technical QA
+ tools + maintenance + internal time cost
A "free" study using 80 hours at 70 euros already costs 5,600 euros before design and outreach. A 2,000 euro platform can cost 3,500 euros if selection, rewriting, and monitoring add 25 hours at 60 euros. Keep the assumptions visible and update the actual hours.
Classic financial ROI is:
ROI = (attributed gross margin minus full cost) / full cost
Use margin, not revenue, if the goal is profitability. Write "attributed under rule X"; without a causal experiment, don't call it incremental revenue.
Eight step measurement procedure
1. Define the hypothesis
Example: "A Digital PR campaign built around our benchmark will, over 16 weeks, raise unbranded impressions and demo requests for our four methodology pages faster than four comparable, unpromoted pages."
2. Choose the units before the campaign
Select target and control pages by intent, position, traffic, age, and demand. Document why they're comparable. If you choose the controls after seeing performance, you introduce bias.
3. Freeze competing changes
Ideally, don't redesign titles, content, internal linking, and templates on the target cohort alone at the same time. If a change is necessary, log it and, if possible, apply it to both groups.
4. Capture the baseline
Keep at least eight to twelve weeks depending on volume and seasonality: impressions, clicks, position, conversions, demand, and incidents. Google recommends segmenting queries, pages, countries, and devices when analyzing variations (Google, debugging traffic drops).
5. Keep the reference log
For each mention or link: source, date, direct/syndicated type, target page, context, attribute, relationship, audience, status, referral, and cost. Deduplicate syndications and check at 30, 90, and 180 days.
6. Compare trajectories
For each group, calculate the evolution from baseline to the after period, then the difference:
observational effect = (target_after minus target_before)
minus (control_after minus control_before)
This simplified difference in differences assumes the trajectories would have stayed parallel without the campaign. Check the baseline period visually; if the curves already diverged, the comparison is weak.
7. Connect to conversions
Search Console observes impressions and clicks before the visit; Analytics observes sessions under its own rules. Google explains that canonical, time zone, attribution, and scope produce legitimate gaps (Search Console and Analytics). Reconcile the data without forcing them to match.
8. Publish the verdict and the alternatives
Write down: observed result, uncertainty, surviving links, cost, hypothesis supported or not, competing explanations, and the next decision. "Inconclusive" is a useful outcome if the test avoids a bigger expense.
Worked example
A campaign costs 12,000 euros. Four target pages go from 400,000 to 520,000 impressions over the comparable window, a 30% rise. Four controls go from 380,000 to 437,000, a 15% rise. The simple relative delta is 15 points, but the volumes aren't identical.
In absolute value: target +120,000; control +57,000. A cleaner method normalizes per page or uses a time series model. The campaign gets 28 pickups, including 12 direct, 16 syndicated, 18 links still present at 180 days, 620 referral sessions, and 14 demo requests. Six opportunities become customers for 18,000 euros of gross margin attributed per the CRM.
Attributed ROI is (18,000 minus 12,000) / 12,000 = 50%. The report adds: multi touch attribution, no randomization, a simultaneous product launch, and a possible brand effect. It doesn't say "the 28 backlinks caused 50% ROI." It says the campaign is associated with this performance under these assumptions.
Scenarios instead of a fictional precision
| Assumption | Attributed margin | ROI | Use |
|---|---|---|---|
| Conservative: 25% of the six customers | 4,500 euros | negative 62.5% | Weak influence or competing channels |
| Central: 60% | 10,800 euros | negative 10% | Majority assisted role |
| High: 100% | 18,000 euros | 50% | Full CRM attribution, non causal |
The scenario shows the decision depends on attribution. A next campaign can be justified by the asset's value and the organic trend even if the central scenario stays slightly negative; that decision must be explicit.
What the external evidence shows and doesn't show
Ahrefs disavowed 3,476 links to three of its articles for four weeks, then removed the file, and observed a drop and a recovery in its visibility estimates (Ahrefs, Do Links Still Matter?). Three pages, a short duration, a disavowal mechanism, proprietary metrics, and no control provide no ROI multiplier. The experiment illustrates recrawl delay and the difficulty of isolating the effect.
Its link rot study on links pointing at 2,062,173 sampled sites since 2013 classified at least 66.5% as lost nine years later (Ahrefs). This isn't a current universal rate, but it makes the survival rate essential: paying for 30 links and still counting 30 after disappearance overstates the return.
Google classifies purchases meant for ranking and excessive exchanges as link schemes (Spam policies). A positive short term ROI doesn't erase policy risk. Paid placements must be evaluated as audience and properly qualified.
Common mistakes
- Attributing all of the period's growth to links.
- Choosing control pages after the result.
- Counting syndications as independent editorial decisions.
- Using earned DR as revenue.
- Forgetting internal time, production, and maintenance.
- Tracking the link on day one but not at six months.
- Measuring a period shorter than the sales cycle.
- Changing all target pages at the same time.
- Presenting a correlation as an incremental effect.
How SEOryon fits in
SEOryon can support research, asset production, read only Search Console/Analytics tracking, and visibility analysis. Measuring a link campaign additionally requires a coverage log, the costs, the CRM, and experimental discipline. SEOryon shouldn't automatically attribute every rise to the content or links it helped create.
Measurable exercise
Take a past campaign. Reconstruct the full cost, the cohorts, eight weeks of baseline, competing changes, link survival, and attributed margin. Calculate three scenarios. Success: the report contains at least two alternative explanations, and a decision maker can approve or stop the campaign without looking at an authority score.
Recommended path
- Return to the framework on authority, backlinks, and Digital PR.
- Strengthen your inputs with the backlink quality, risk, and value matrix.
- Build an editorial cohort with original data in Digital PR.
- Present the scenarios in ROI and decision focused SEO reporting.
FAQ
How long should you wait before measuring?
Measure acquisition and referrals immediately, survival at 30/90/180 days, and organic impact over a window compatible with volume and recrawl. Revenue follows the sales cycle.
Can you causally prove a backlink's effect?
Rarely on a single page. Large scale randomization is stronger; a comparable cohort and a change log improve an observation without making it perfect.
Which KPI should you present to the client?
Present the chain: cost, qualified placements, survival, audience, target/control evolution, conversions, and revenue scenarios. Don't isolate the most favorable figure.
How do you value a nofollow link?
By audience, referral, credibility, partnership, and assisted influence. Don't attribute a guaranteed ranking effect to it.
What if the test is inconclusive?
Check power, duration, comparability, and competing changes. Reduce or redesign the campaign; don't mechanically double the budget.
Main sources
- Google: Search Console and Analytics together- Google: Debugging drops in Search traffic- Google: Spam policies- Google: Qualify outbound links- Ahrefs: Do links still matter?- Ahrefs: Link Rot Study- Reboot: Digital PR Statistics 2025
Method note
Method: SEOryon formulas for explicit observational measurement, not a certified causal model. Sources verified 16 July 2026, translated and edited 22 July 2026.