Building a Simple Scoring Model for Publisher Selection
A repeatable formula beats gut-feeling comparisons once you're vetting more than a handful of sites a month.

Comparing publishers one metric at a time works fine for the occasional one-off placement. Once you are vetting more than a handful of sites regularly, a simple scoring model — a formula that turns several metrics into one comparable number — makes decisions faster and more consistent, especially across a team.

Why a single blended score beats gut feel
Without a structured approach, it is easy to let one impressive number, usually DR or DA, dominate a decision even when traffic, spam risk, or relevance tell a different story. A scoring model forces you to define upfront how much each factor should matter, which produces more consistent outcomes than judging each prospect afresh based on whichever number happens to stand out.
Choosing your inputs
- Authority — DR or DA, see https://successtoroad.com/blog/domain-rating-explained/ and https://successtoroad.com/blog/domain-authority-explained/.
- Traffic — estimated organic traffic and its trend, see https://successtoroad.com/blog/organic-traffic-estimates-why-tools-disagree/ and https://successtoroad.com/blog/organic-traffic-estimates-why-tools-disagree/.
- Referring domain diversity — see https://successtoroad.com/blog/referring-domains-vs-backlinks-explained/.
- Spam or risk score — see https://successtoroad.com/blog/spam-score-explained/.
- Relevance — niche and, where applicable, country fit; see https://successtoroad.com/blog/traffic-by-country-for-placements/.
A basic weighting example
| Factor | Suggested weight | Rationale |
|---|---|---|
| Traffic (with positive trend) | 30% | Real audience is the ultimate point of a placement |
| Authority score | 25% | Still a meaningful proxy for link value |
| Relevance (niche/country) | 25% | An irrelevant audience limits value regardless of size |
| Spam score (inverted) | 20% | Risk reduction matters even on otherwise strong sites |
These weights are a starting point, not a fixed rule — adjust them to match what actually matters for your typical campaigns. A brand focused heavily on geographic targeting might weight relevance higher; a buyer mainly building general authority might weight the authority score higher.
Normalising each factor
Since DR, traffic, and spam score sit on very different scales, convert each into a comparable range, such as 0-10, before applying weights. A simple approach is ranking prospects within your own list on each factor and scoring relative to that range, rather than trying to apply a universal formula across totally different niches and site sizes.
Where manual review still fits in
A scoring model speeds up ranking a list, but it should not fully replace the manual checks covered in https://successtoroad.com/blog/how-to-read-a-link-profile/ and https://successtoroad.com/blog/vet-a-site-in-ten-minutes/. Use the score to prioritise which sites get that manual attention first, and treat a borderline score as a prompt for a closer look rather than an automatic decision either way.
Applying the model at scale
Once you have a working formula, apply it consistently to bulk-checked lists (https://successtoroad.com/blog/bulk-domain-check-for-prospect-lists/) or data pulled via https://successtoroad.com/blog/api-for-pulling-metrics-into-spreadsheet/, so every prospect across every campaign is judged the same way. This consistency is especially valuable when more than one person on a team is making buying decisions, since it reduces variance between individual judgment calls.
Revisiting the model over time
Review your weights periodically against outcomes — which placements actually performed well, which underdelivered despite strong scores — and adjust accordingly. A scoring model is a working tool, not a one-time setup, and should evolve as you learn what predicts good placements for your specific niche and budget.
Handling missing or unavailable data
Not every metric will be available for every prospect, particularly for smaller or non-English sites where some tools have thinner coverage. Decide in advance how your model handles a missing input, rather than letting a blank field silently distort the final score. A common approach is treating a missing metric as neutral, using the average or median of the rest of your list, rather than as zero, which would unfairly penalise a site simply because one particular tool lacks data on it.
Guarding against over-optimisation
It is possible to build a scoring model so elaborate that it becomes difficult to maintain or explain, with dozens of weighted inputs that obscure rather than clarify the underlying decision. A simpler model with four or five clearly justified inputs, reviewed and adjusted periodically based on real outcomes, tends to serve most teams better than an intricate formula that nobody can fully account for when a stakeholder asks why a particular site scored the way it did.
Sharing the model transparently with publishers
Some buyers choose to share a simplified version of their scoring criteria with publishers directly, particularly for an ongoing relationship. This can encourage a publisher to maintain or improve exactly the qualities that matter most to you, such as content freshness or a clean outbound link pattern, since they understand concretely what is being evaluated rather than guessing at an opaque internal process.
Starting simple and adding complexity only when needed
It is tempting to build an elaborate model on day one, anticipating every edge case in advance. A more practical approach is starting with three or four clearly weighted inputs, using it for a few real campaigns, and only adding further complexity, such as a new-domain adjustment or a country-relevance multiplier, once you have concrete evidence that the simpler version is missing something that actually matters for your decisions.
Where to go next
For definitions of every input metric, see https://successtoroad.com/blog/seo-metrics-explained/. Apply your model to live prospects using https://successtoroad.com/marketplace/ or https://successtoroad.com/bulk-check/, and set budget-appropriate thresholds with https://successtoroad.com/blog/metric-thresholds-by-budget/.
Frequently asked questions
Do I need special software to build a scoring model?
No, a spreadsheet is usually sufficient. Pull metrics via bulk check or API export, normalise each column, apply your weights, and sort by the resulting score.
How often should I update my weighting formula?
Review it periodically, such as quarterly, or whenever you notice a pattern of placements underperforming or overperforming relative to what the score predicted.
Should every campaign use the same scoring model?
Not necessarily. Different campaign goals, such as geographic targeting versus general authority building, may justify different weightings, though a shared baseline model is a reasonable starting point.
What if two prospects end up with the same score?
Use a manual tie-breaker — read both link profiles and recent content, and consider practical factors like turnaround time or communication quality.
Can a scoring model replace manual vetting entirely?
No. It prioritises which prospects deserve manual attention first; it should not fully replace a final manual check before committing budget to a placement.
Put it into practice
Filter 54,483 websites by DR, traffic, niche and price, and order with escrow — you pay only after the link is live.
Browse the marketplace

