Your client asks the reasonable question: we already put their first name in the subject line, so why are we paying for personalized images on top?
The tempting answer is to reach for a statistic. There are plenty circulating - impressive percentages about engagement lifts from personalized visuals, usually with no methodology attached and often traced back to a study about something else entirely.
Don't. There are two problems with it, and the second is worse than the first.
The numbers are unreliable. Results in this category vary enormously depending on the audience, the industry, what data is being personalized, and how visible the personalization is. A figure from a retail campaign to consumers tells you very little about a B2B nurture sequence. Anyone quoting a single confident number across all of that is not measuring the same thing you are.
And your client will check. If the statistic is thin, someone in procurement will find that out, and at that point you are not discussing personalization any more - you are defending your credibility. A weak number is worse than no number.
The better approach is to build the case from your client's own figures. It takes twenty minutes, it cannot be contradicted by a search, and it produces a much more interesting conversation.
Invert the question: what lift would it take to break even?
Rather than promising a result, calculate the result required to justify the cost. Then let the client decide whether that looks achievable.
Here is the shape of it, with illustrative numbers - replace them with your client's actual figures.
Their current position:
| Emails sent per month | 50,000 |
| Current click-through rate | 2.5% |
| Clicks per month | 1,250 |
| Click-to-conversion rate | 3% |
| Average order value | £80 |
| Revenue attributable per click | £2.40 |
That last line is the one that matters, and it is just 3% × £80.
The cost of adding personalized images, say £150 a month all in - tool plus the setup time amortized over the year.
So the break-even is £150 ÷ £2.40 = 63 additional clicks a month.
Sixty-three extra clicks out of 50,000 sends is 0.125 percentage points of click rate. Going from 2.5% to 2.625%.
In relative terms, that is a 5% improvement in click-through rate to break even.
Now you have a genuinely useful conversation. You are not claiming personalized images will transform their results. You are asking whether a 5% relative improvement in click rate seems plausible, and that is a question a marketer can answer from experience rather than take on faith. Most will say yes without much hesitation, because 5% relative is a modest number - and crucially, they have reached that conclusion themselves rather than being told it.
Everything above break-even is upside, and you have not promised any of it.
The argument that doesn't need a statistic at all
There is a second case, and for some clients it is the stronger one, because it is a cost saving rather than a projected gain.
Ask how personalized visuals are produced today. If the answer involves a designer making variants by hand, you can price it directly.
A worked example: twelve segment variants a month, twenty minutes each, is four hours of design time. At an agency rate that is a real line item, and it recurs every month. It also carries costs that don't show up on the invoice - the turnaround delay before a campaign can go out, and the fact that nobody asks for a thirteenth variant because it isn't worth the hassle.
Automated generation replaces that with one template and a data source. The setup is a one-off; the marginal cost of the next variant is approximately nothing.
This argument is verifiable. The client can check the hours against their own records. It doesn't depend on anyone's engagement statistics, and it lands particularly well with finance, who tend to find cost removal more persuasive than revenue projection.
There is a strategic version of the same point. When each variant costs twenty minutes, you produce twelve. When it costs nothing, you can produce one per recipient - and that is a different marketing capability, not a cheaper version of the existing one. Segment-level personalization and individual personalization are not the same product.
Why "we already use their first name" isn't the same thing
This is usually the actual objection underneath the question, so it's worth answering directly.
Text personalization has been commoditized. Every email a person receives opens with their first name. It stopped being a signal of effort roughly a decade ago, and recipients have learned to skip past it. It is now table stakes - noticeable when absent, invisible when present.
Images are still processed differently. People scan visuals before they read body copy, so personalization inside an image is seen by recipients who never get as far as the second paragraph. That is a structural difference in how the two are consumed, not a claim about a percentage.
And images can carry things text can't gracefully hold. A chart of the client's own usage. A map. A product they looked at. Their name rendered on a ticket or certificate rather than announced in a greeting. Some of this is possible in text and simply reads worse.
The honest framing: text personalization signals that you have their record in a database. Image personalization can show them something specific to them. Those are different messages, and the second is harder to ignore.
Structure a pilot that generates your own evidence
The most durable answer to "does this work?" is your own data from your own client's list. Propose a pilot designed to produce it.
Pick one campaign with enough volume to produce a readable result. A recurring send is better than a one-off, because you get a repeat.
Change one thing. The personalized image, nothing else. If you also rewrite the subject line, you will learn nothing attributable.
Agree the success metric before you start, in writing. This matters more than people expect, because it prevents the post-hoc argument about which number counted.
Measure clicks, not opens. Apple Mail Privacy Protection prefetches images whether or not a human opened the email, and Gmail and Outlook each handle images differently again. An open-rate comparison partly measures which mail clients the audience uses rather than whether the image worked. Click-through, and ideally what happens after the click, is the honest measure. Our guide to measuring whether personalized images actually work goes into this properly.
Agree in advance what a null result means. A pilot you would spin regardless of outcome is not a pilot. Saying "if it doesn't beat the control over two sends, we drop it" costs you very little and buys a lot of credibility.
Then you own a number that is about their audience, and the industry statistics stop mattering.
Objections you'll get, and honest answers
"Isn't this just a gimmick?" It can be. Personalization that is decorative - a name dropped into a banner where nobody looks - does very little, and the client is right to be sceptical of it. What works is personalization carrying information the recipient actually wants: their data, their status, their item. Concede the first part; it makes the second more credible.
"Our list is small." Then the response-lift argument is weak, because you cannot measure a small effect on a small list, and you should say so. The production-cost argument may still hold. And for genuinely small lists, the honest answer is sometimes that this isn't the right investment yet.
"Will it slow the emails down?" Images render when the email is opened and are cached afterwards, so there is a generation step measured in fractions of a second. In practice it is not something recipients notice.
"What if the data is wrong?" Then a mistake appears in a picture, which is worse than a mistake in text because it is more prominent. This is a real risk and worth naming rather than waving away. Fallback values for missing fields and a test send against real data handle most of it. The related risk - long names breaking a layout - is a design question, solved by building the template around the longest plausible value rather than a short sample.
"What happens when we want to change it after sending?" With open-time rendering, you can. The image is generated when opened, so correcting a live design updates it for anyone who opens afterwards. That is a genuine advantage over pre-rendered images, and it is worth mentioning because most clients assume email is unchangeable once sent.
What not to promise
A short list, because over-promising here is how agencies lose accounts.
Don't promise a specific percentage. You do not know it, and you will be held to it.
Don't promise it fixes a deliverability or list-quality problem. If emails are landing in spam, or the list is stale, better images will not help and you will have spent the client's money on the wrong thing.
Don't promise it works equally everywhere. Some audiences respond strongly, some barely register it. Which is precisely why the pilot is the right structure.
Don't imply it replaces good copy or a good offer. It amplifies a campaign that already works. It does not rescue one that doesn't.
Frequently asked questions
How do I justify personalized image software to a client without quoting a statistic? Calculate the break-even instead. Work out the revenue attributable to a click, divide the monthly cost by it to get the additional clicks needed, and express that as a percentage improvement in click rate. Then ask whether that improvement seems achievable. The client reaches the conclusion from their own numbers rather than taking a claim on trust.
Why not just use published statistics on personalization? Because results vary enormously by audience, industry and implementation, and most circulating figures have no methodology attached or come from a different context entirely. If a client checks a weak number, the conversation stops being about personalization and starts being about your credibility.
What's the difference between text personalization and personalized images? Text personalization is universal and largely invisible to recipients now. Images are scanned before body copy is read, and can carry information text holds awkwardly - charts, maps, products, documents with the recipient's details rendered into them.
How do I prove it worked? Run a single-variable pilot on one campaign, agree the success metric in advance, and measure click-through rather than opens. Open rates are distorted by Apple Mail Privacy Protection prefetching and by how different mail clients handle images.
What if my client's list is too small to test? Then say so. On a small list the response-lift case is hard to demonstrate and you should not pretend otherwise. The production-cost argument may still apply if they are currently making variants by hand.
Is the cost saving argument stronger than the engagement argument? Often, yes - particularly with finance stakeholders. Removing four hours of recurring design work a month is verifiable against the client's own records, where a projected engagement lift is not.
What's a realistic break-even? It depends entirely on their list size, click rate and order value, which is the point of running the calculation. In the illustrative example above it works out at a 5% relative improvement in click-through rate - low enough that most marketers consider it achievable.
Putting it in front of a client
The strongest version of this conversation isn't a deck, it's a spreadsheet with their numbers in it and a pilot proposal attached. Twenty minutes of arithmetic beats any case study.
If you want to build the pilot, OKZest generates personalized images from a single template and a data source, with no per-variant design work - which is what makes a one-campaign test cheap enough to run before committing anyone to anything.
Create a free account, no credit card required, or see pricing if you need a figure for the break-even calculation.
Related reading: do personalized images actually work, and how to measure it, A/B testing personalized images, personalized image software for agencies.