Reading a Shopify CRO Proposal: Separating Method From Marketing

There is a proposal open on your desk, and page one already names a number. A 20% lift, a 30% lift, some confident percentage attached to a store the agency has not yet been given access to. It reads like ambition. It is worth pausing on, because a figure promised before anyone has seen your analytics is not a forecast. It is a sales device, and the rest of the document deserves to be read with that in mind.

Most guides on choosing a CRO partner hand you a list of questions to ask on a call. Useful, but the proposal itself is the more honest artifact. It is written down, it commits the agency to specifics, and it can be read line by line without anyone talking over the gaps. Nearly every line in it does one of three jobs: it describes the method, it does the marketing, or it sets the terms. Learn to sort the lines into those three piles and the document stops being persuasive and starts being informative.

What a CRO proposal is actually supposed to do

A CRO proposal exists to scope work, not to promise an outcome. Its job is to describe how an agency will find where your store loses revenue, how it will decide what to change, how it will measure whether the change worked, and what you are committing to in time, money, and access. A proposal that leads with a guaranteed result and stays vague on the method has inverted its own purpose: it is selling the destination while staying quiet about the road.

That inversion is the single most useful thing to watch for, and it splits the whole document cleanly. Read each section and ask which of three jobs the sentence is doing. Method language describes the work: the research phase, how hypotheses get prioritized, how tests are measured, what happens to tests that lose. Marketing language describes the reward: lifts, revenue figures, client logos, proprietary-sounding frameworks with no mechanism attached. Terms language describes the deal: pricing, contract length, who does the work, who owns the data, how you leave. Every proposal contains all three. What separates a scoping document from a sales deck is the ratio between them and whether the method section can actually stand up on its own.

None of this requires you to be a CRO specialist. It requires reading the proposal the way you would read any contract for expert work: assume the substance lives in how they describe the doing, not in how they describe the winning. The next three sections take the piles one at a time, starting with the one that should carry the most weight and usually carries the least ink.

Bucket A — Method: how the work actually gets done

The method section is where a real CRO engagement lives, and it should be the longest, most specific part of the proposal. Read it for four things: a research phase that comes before any test list, a stated way of prioritizing what to test, a definition of how a result gets measured, and an honest account of what happens when a test loses. If those four are present and concrete, you are reading a scoping document. If they are thin or missing, the confident numbers elsewhere have nothing underneath them.

Start with the order of operations, because it is the tell that is hardest to fake. A methodologically sound proposal describes research before it describes testing: analytics review, funnel diagnosis, session data, customer input, and only then a backlog of hypotheses ranked against where the money is actually leaking. Shopify’s own guide to A/B testing puts the split bluntly, calling the work roughly 80% research and 20% testing. A proposal that opens with a test list (“we will test your hero, your CTA, your product page”) and no diagnostic phase has skipped the 80% and kept the part that looks like activity. The discipline of locating the leak before building a backlog is the same one that separates a diagnosis from a guess, and it is worth reading our own note on the Shopify conversion leak diagnostic alongside any proposal to see what that phase should contain.

Prioritization is the second thing to find. Any agency can generate a hundred ideas. The method question is how they choose the order, and a serious proposal names a framework: hypotheses ranked by expected impact against the diagnosed leak, not pulled from a generic checklist. If the document lists tactics without a word on sequencing, you are being sold a queue, not a strategy, and the sequencing is where most of the return actually comes from.

Measurement is the third, and it is where method quietly separates from marketing. Look for the metric the proposal commits to reporting on. Revenue per visitor, or contribution margin, ties the work to your business. Conversion rate alone does not, because a discount popup can lift conversion rate while destroying margin. A proposal that measures success on revenue per visitor is describing a method. One that measures it on conversion rate in isolation is describing a metric that flatters the report.

The fourth is the one honest proposals include and sales decks omit: what happens when a test does not win. Most tests do not produce a winner, and a proposal that treats every experiment as a future success is either inexperienced or selling. Watch for language about statistical significance (a real winner clears a confidence threshold, it is not called after a few good days) and about extracting a learning from a losing test rather than quietly moving on. An agency that plans for losing tests is an agency that has run enough of them to know the odds.

Bucket B — Marketing: the language that sells instead of scopes

Marketing language is not a flaw in a proposal. Every agency has to show why it is worth hiring, and confidence is reasonable. The problem is only when the marketing sits where the method should be, doing the persuading that the method could not. Five patterns recur, and each has a plain question that deflates it.

The first and most important is the promised lift before access. A proposal that commits to a specific percentage improvement before the agency has seen your analytics is selling, not scoping, because the honest version of that number cannot exist yet. Nobody can size the recoverable revenue in your funnel without reading your funnel. The tell is sequence: a number that arrives before the diagnostic phase is a marketing figure wearing a forecast’s clothes. The question that deflates it is simple. On what data was this based? If the answer is industry averages or past clients, it is not about your store.

The second is the vanity metric dressed as a result. Case studies that lead with conversion-rate percentages in isolation, click-through improvements, or engagement lifts are reporting numbers that may not touch revenue at all. A 40% lift on a step nobody struggled with moves nothing. Read every cited result for whether it connects to revenue per visitor or order value, and treat a proposal that never denominates its wins in money as one that is choosing the flattering number over the meaningful one.

The third is the logo wall without a mechanism. Client logos and testimonials prove someone paid the invoice, not that the work moved the number. A strong proposal shows what was tested, what the hypothesis was, what the result was, and how it was measured, ideally with the client’s economics attached. Names and five-star quotes in place of that are social proof standing in for evidence. Ask to see one worked example, redacted if necessary, of a research process turning into a hypothesis turning into a measured outcome.

The fourth is the proprietary black box. A “proprietary optimization framework” or a trademarked-sounding methodology is fine when the proposal then explains how it works. It is a red flag when the name is the whole explanation, because a method you cannot inspect is a method you cannot evaluate, and opacity at the proposal stage rarely improves once the contract is signed. The question is direct. Walk me through what this framework does in the first thirty days.

The fifth is manufactured urgency. Limited slots, a discount that expires this week, pressure to sign before the analysis has happened: none of these are features of a scoping relationship, and all of them are designed to shorten the exact reading you are doing right now. A CRO engagement is a months-long partnership. Nothing about evaluating one carefully should cost you the deal, and an agency that implies otherwise is telling you something about how it operates.

Bucket C — Terms: the fine print that decides the relationship

Method tells you whether the work is sound. Marketing tells you how hard you are being sold. Terms tell you what you are actually signing, and it is the pile most proposals keep thinnest, because specifics here are where an agency’s flexibility gets pinned down. Six items are worth locating before a signature, and a proposal that leaves several of them vague is asking you to resolve them after you have lost your leverage.

Pricing is the first, and the question is not the number but whether it fits your scale. Flat-rate pricing that ignores your traffic and revenue is a red flag in both directions: it can mean a store doing €300k a month subsidizing the agency’s smaller clients, or a store doing €30k a month paying for a program its traffic cannot support statistically. A proposal should connect its fee to the work and the scale, and a productized audit priced the same for everyone is fine as a diagnostic but tells you little about the ongoing economics.

Contract length and lock-in are the second. Look for the commitment term and what happens inside it. A multi-year plan locked at signing, before a single test has run, removes the one piece of leverage that keeps an agency accountable: your ability to leave if the method does not deliver. Quarterly or evolving terms that adjust to findings signal an agency that expects to earn the next quarter. A long lock-in signals one that would rather not have to.

Who does the work is the third, and it is the question the proposal is least likely to answer unprompted. The people in the pitch are often not the people on the account. Ask, in writing, who runs your research, who designs the tests, who ships the code into your Shopify theme, and whether that last part needs your developers or theirs. On Shopify Plus especially, an agency that cannot implement inside the platform without your dev team has a bottleneck the proposal will not mention.

Data and account ownership is the fourth, and the one that hurts most when it is skipped. When the engagement ends, who keeps the test results, the research, the analytics configuration, and the accounts everything runs on? A proposal should leave you owning your own store’s history. Some do not, and you find out at the exit.

Reporting cadence is the fifth. The proposal should state how often you will hear from the agency, in what form, and connected to which metric. An agency that commits to a monthly report tying tests to revenue is describing accountability. One that goes quiet between invoices is describing a different kind of relationship, and the cadence written into the proposal is the version you can hold them to.

The sixth is the exit, including rollback. What happens to the changes when you part ways? A clean engagement leaves your store stable and documented, with test code that can be rolled back rather than left as sediment in your theme. A proposal that describes how it ends is written by an agency that has ended engagements well before.

Decision rules: the scoring lens to run before you sign

Once every line is sorted into method, marketing, or terms, the decision comes down to three reads, not a gut feeling about the agency’s polish. Run them in order, because the first one can end the evaluation on its own.

First, the disqualifier. Does the proposal commit to a specific lift before the agency has seen your data? If yes, treat it as a hard stop worth raising directly, because a number promised before a diagnosis is the one signal that a proposal is selling a result it has no way to size. Everything else is a matter of degree. This one is binary, and it is the fastest way to shorten a stack of proposals.

Second, the ratio. Weigh the method pile against the marketing pile. A scoping document spends most of its specificity on how the work gets done: the research phase, the prioritization logic, the measurement metric, the plan for losing tests. A sales deck spends it on lifts, logos, and framework names. You are not looking for zero marketing. You are looking for a method section that could stand on its own if you deleted every promised number, and a marketing section that would collapse without them. When the confident language is load-bearing, the proposal is thinner than it reads.

Third, the completeness check on terms. Are pricing-to-scale, contract length, who-does-the-work, data ownership, reporting cadence, and exit all answered in writing? Each blank is a negotiation you will have later from a weaker position. A proposal that answers all six is not necessarily the cheapest, but it is the one that respects you as a counterparty rather than a signature.

Those three reads are portable. They work on any CRO proposal, from any agency, including ours. If you are weighing a Flatline proposal, run the same lens on it: check that the method section outweighs the marketing, that no lift is promised before we have seen your analytics, and that the terms are answered in full. As the Shopify Plus agency whose own proposals get judged by this same lens, we would rather you hold us to it than take our word for it. We keep a scoping checklist that mirrors these reads and are glad to share it on request, so you can pressure-test whatever lands on your desk before a signature, wherever it came from.

Frequently asked questions

Should a CRO proposal guarantee a conversion lift? 

No, not before the agency has seen your analytics. A specific percentage promised at the proposal stage is a marketing figure, because the recoverable revenue in your funnel cannot be sized without reading your funnel first. A grounded proposal commits to a method and a measurement, then estimates only after a diagnostic phase. Treat a guaranteed number attached to a store nobody has analyzed as a selling device, not a forecast.

How long before a CRO engagement should show results? 

Longer than a sales deck implies. A sound program spends its early weeks on research and diagnosis before running tests, and each test needs enough traffic and time (often one to two full business cycles) to reach statistical significance rather than being called on a few good days. Ask the proposal to describe its first thirty days as a research and prioritization plan, not a promised win by week two.

What’s the difference between a CRO audit and a CRO proposal? 

An audit is a diagnostic deliverable: a one-time analysis of where your store loses revenue, ideally sold as a standalone engagement. A proposal is the scope for an ongoing relationship built on that diagnosis. Many strong agencies offer a paid audit first, which lets you evaluate their thinking before committing to a retainer. A proposal that skips any diagnostic step and jumps straight to a test list has left out the audit’s work.

Is a cheap flat-rate CRO proposal a red flag? 

Not automatically, but the flat rate is worth reading against your scale. A fixed price that ignores traffic and revenue can mean a larger store subsidizing smaller clients, or a smaller store paying for a testing program its traffic cannot support statistically. A productized audit at a fixed price is fine as a diagnostic. For an ongoing engagement, the fee should connect to the work and the scale.

Key takeaways

• A CRO proposal exists to scope work, not to promise an outcome. Read every line as method, marketing, or terms.
• Method is where the real engagement lives: research before test lists, prioritization logic, revenue-based measurement, and a plan for tests that lose.
• The single hardest stop is a specific lift promised before the agency has seen your data. That number cannot honestly exist yet.
• Marketing language is fine until it sits where the method should be. Vanity metrics, logo walls, black-box frameworks, and urgency are the recurring substitutes.
• Terms decide the relationship: pricing-to-scale, contract length, who does the work, data ownership, reporting cadence, and a clean exit. Every blank is a later negotiation from a weaker position.
• The lens is portable. Run the disqualifier, the method-to-marketing ratio, and the terms completeness check on any proposal, including ours.