Discussion starter: an agency question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Agency
Performance-based SEO pricing: have you seen it work without creating incentives to chase easy, irrelevant traffic?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
Yes, it can work, but only when the thing you pay for sits close to revenue, is limited to traffic the business actually wants, and is measured against a baseline both sides agreed before work started. Pay on sessions, rankings or impressions and you will usually get the behaviour you are worried about, because those are the easiest numbers to move with content nobody buys from.
The harder trade-off is not “performance versus retainer”. It is alignment versus control. The closer a metric is to revenue, the less of it the agency controls: pricing, stock, the sales team’s follow-up speed and the client’s development backlog all sit between a ranking and a sale. Every performance deal is really a negotiation over where on that line the payment sits, and who carries the risk for the parts neither side controls.
Public evidence on how performance-priced SEO engagements turn out is thin. No transparent dataset comparing outcomes across pricing models exists that I can point to, so what follows is the mechanism, the structures that tend to hold up, and the warning signs that a deal is rewarding the wrong traffic.
Why paying for traffic buys the wrong traffic
This is Goodhart’s law in a very pure form: once a measure becomes the target, people optimise the measure rather than the thing it was meant to represent. Organic sessions were only ever a proxy for “more of the right people finding us”. Pay per session and the cheapest sessions win, and the cheapest sessions are rarely the valuable ones.
A useful comparison is paying a real estate agent per open-home visitor instead of on the sale. Visitor numbers are easy to inflate: advertise in the wrong suburb, offer free coffee, list at an unrealistic price. None of that sells the house. Even a commission on the sale price is imperfect, because the agent’s share of an extra $20,000 is small compared with the extra weeks of work, so a quick sale can suit the agent better than the vendor. SEO performance contracts can have both problems: the proxy problem if you pay on traffic, and the split-incentive problem even when you pay on outcomes.
In SEO the “easy” traffic is predictable:
- Broad informational queries next to the niche (glossaries, “what is” posts, loosely related how-tos) with no path to a purchase or enquiry.
- Branded queries that were growing anyway because of TV, PR or paid social.
- Visitors from countries or regions the business cannot serve.
- Image, meme or news-cycle traffic that spikes and converts at close to zero.
- Low-difficulty keywords chosen because they are rankable, not because anyone searching them buys.
Watch for the risk asymmetry:
Under a performance deal the agency captures most of the short-term upside, while the client carries the long-term downside if tactics breach Google’s spam policies (bought links, scaled low-value content). Google’s guide to hiring an SEO states plainly that “No one can guarantee a #1 ranking on Google”. A contract that pays heavily for rankings quietly rewards anyone willing to act as though they could. Write prohibited tactics into the contract and require link acquisition to be itemised.
How each payment metric holds up
Metric paid on How easily it is gamed How much the agency controls it Measurement traps Keyword rankings Very easy: pick low-competition terms High Tracker methodology changes. Google’s removal of the &num=100parameter in September 2025 disrupted rank trackers.Impressions or average position Easy High The same change cut reported desktop impressions and improved average position for many sites, while clicks were affected far less, according to Brodie Clark’s analysis. A contract on impressions would have paid out or penalised for a reporting change. All organic clicks or sessions Easy Medium to high Brand demand, seasonality and core updates. AI Mode data sits inside Search Console’s Web totals, and practitioners report AI Mode clicks arriving in GA4 as google / organic, so the channel is changing underneath the contract. Non-brand clicks to an agreed page set Moderate Medium Needs a written brand definition. Search Console’s branded queries filter helps, but check it against your own brand term list. Leads or form fills Moderate: spam, low-intent forms, free-tool signups Medium Duplicates, junk and bot submissions Qualified leads (CRM stage) or new customers Hard Low to medium Sales speed and CRM hygiene affect the number; attribution window disputes Revenue or gross margin from organic Hard Low Pricing, stock, promotions, conversion rate work and attribution model; long sales cycles delay payment The pattern is clear enough to act on. The top of the table is easy to game; the bottom is hard for the agency to influence. Workable deals usually sit in the middle: non-brand, scoped, quality-gated outcomes.
Structures that tend to hold up
Base plus capped bonus
A retainer covers delivery cost. The bonus is paid on a quality-gated outcome above baseline, with a cap. The base removes the pressure to chase anything that moves; the cap limits the client’s exposure to a windfall the agency did not cause, such as a competitor closing or a favourable core update.
Scope by page set or query list
Only conversions or non-brand clicks landing on agreed commercial URLs, or on an agreed query list, count. New content outside scope earns nothing unless both parties add it. That removes the glossary-farm incentive at the source.
Quality gates and clawbacks
Count leads only when they reach a CRM stage such as sales-accepted. Net out refunds and cancellations, and hold a clawback window. If discounting is common, pay on margin rather than revenue.
Controlled measurement on large sites
On sites with many similar templated pages, changing a random half of comparable pages and holding the rest as a control isolates uplift far better than before-and-after comparisons. It needs enough pages and traffic, which rules out most small-business engagements.
The baseline is where most disputes start
A performance fee is only as fair as its counterfactual: what would have happened without the work. A simple before-and-after comparison credits the agency with seasonality, brand campaigns, competitor exits and algorithm changes. Google completed both a March 2026 and a May 2026 core update, so any contract running through this year has at least two events neither party controlled. Agree these points in writing before kickoff:
- The baseline period. Use the same months of the prior year plus the trailing trend, not just the 30 days before the start date.
- How brand is defined and excluded, and who maintains the term list.
- Which markets, page sets and conversion types count.
- The attribution rule (for example, an organic landing session within 30 days, confirmed in the CRM) and who owns the data.
- What happens after a measurement change. GA4 added an “AI Assistant” default channel group in May 2026, which reclassified some referral sessions. Channel definitions shift under long contracts, so specify how you will restate the baseline.
- A control segment where possible: comparable pages, product lines or regions the agency is not touching.
Hypothetical example:
A B2B software company pays an agency per organic free-trial signup. Within six months signups double, mostly through a new “free templates” hub. Paid conversions from organic stay flat, because the templates attract students and freelancers outside the ideal customer profile. Nobody broke the contract; the contract described the wrong outcome. A restructured deal pays a smaller base plus a bonus on sales-qualified trials whose first session landed on an agreed list of product, integration and comparison pages, with trials from unserviceable markets excluded.
Warning signs that a deal is rewarding the wrong traffic
- Organic sessions rising while conversion rate per organic session falls for several consecutive months.
- A growing share of landing pages that are informational and have no internal path to a product or service page.
- Growth concentrated in countries or regions outside the service area.
- Non-brand growth that mostly disappears when you remove one or two content hubs from the numbers.
- New content topics the client never approved.
- Link acquisition the agency will not itemise.
None of these proves bad faith. Several of them together usually mean the contract is paying for the wrong thing, and it is better to fix the metric than to argue about the invoice.
When performance pricing is workable and when it is not
Condition Performance-weighted pricing more workable Retainer or project pricing more sensible Conversion tracking Clean, CRM-connected, agreed definitions Patchy, disputed or recently migrated Sales cycle Days to weeks Many months with several decision-makers Implementation control Agency can ship changes or has committed developer time Changes queue behind a product roadmap Demand mix Mostly non-brand and commercial Dominated by brand demand driven by other channels Volume Enough monthly conversions to separate signal from noise A handful of conversions a month Risk profile Low compliance and brand risk YMYL, regulated or brand-sensitive, where aggressive tactics carry real cost There is also a selection effect worth naming. Performance deals disproportionately attract clients with low baselines (easy early gains) or clients who cannot afford a retainer (often with weak tracking and little development capacity). Agencies that do well on these deals tend to be selective about which clients they offer them to, which is another reason to be cautious about generalising from anyone’s success story.
A fast way to test the structure before committing
Run a six-month pilot on a base-plus-bonus contract scoped to an agreed commercial page set. The hypothesis is that it grows qualified outcomes without shifting effort towards off-target traffic.
- Primary metric: qualified conversions from in-scope pages against the trend-adjusted baseline.
- Secondary metrics: share of organic sessions landing outside scope, conversion rate per organic session, and the brand versus non-brand split.
- Confounders to log: core updates, pricing changes, paid media bursts, site releases and tracking changes, each with a date.
- Decision rule: if qualified conversions rise and the off-scope share stays flat, extend. If sessions rise but qualified conversions do not, the metric is wrong, not necessarily the agency.
My practical position
Never pay on traffic alone. Pay on qualified outcomes from a pre-agreed scope, above a baseline that accounts for trend and season, with a base fee large enough that the agency has no reason to chase anything that moves. If the client cannot yet measure qualified outcomes reliably, fix the measurement first and price the engagement conventionally until it is fixed. A performance contract built on bad tracking just moves the argument from “is the SEO working?” to “whose number is right?”.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.