Ad creatives for SaaS when you don't have a designer

Most makers stall at the same place. The campaign is built, the conversion event fires, the budget is set — and then the ad wants an image, and you don’t have one. You’re not a designer. You open Canva, lose an afternoon to font pairings, and upload something that looks fine and performs like nothing.

Here’s the reassuring part: for software, the polish is rarely what’s holding you back. The thing that decides whether an ad works is what it says and what it shows — and you already own both.

The polish is not the variable

An ad has to do three jobs in about a second: interrupt the scroll, make the problem recognisable, and make the next step obvious. Design contributes to the first one. It contributes almost nothing to the second and third.

This matters because it tells you what to spend your limited effort on. A beautifully typeset ad that says “Streamline your workflow” will lose to a plain screenshot captioned “Your invoices are in three places and none of them agree.” One of those describes a product category. The other describes a Tuesday.

So the honest priority order for a maker with no design skills:

1. The hook — the first line of text and the first frame. This is the variable with real range.
2. The proof — showing the product doing the thing.
3. The format — right ratio, right length, readable on a phone.
4. The aesthetics — last, and mostly a matter of not looking broken.

Most people work that list backwards.

Screenshot-first: the patterns that carry their weight

Your product is the best asset you have, and it costs nothing to photograph. Four patterns that consistently do a job:

The single clean screen. One screen of your app, cropped to the part that matters, on a plain background with generous padding. Not the whole dashboard — the one panel that represents the outcome. If your product’s value is “you finally see all your ad spend in one place,” show that one view, not the settings page.

The annotated screenshot. Same thing, plus one arrow or one highlighted box and four or five words. You’re directing attention to the specific pixel that carries the value. This is the single highest-leverage upgrade over a bare screenshot, and it takes two minutes.

The before/after. Left: the mess (a spreadsheet, six browser tabs, a chaotic inbox). Right: your product. This works because the left panel is the reader’s actual life, and recognising yourself is what stops the scroll. You can shoot the “before” in thirty seconds because you almost certainly still have the mess somewhere.

The screen recording. A 15–25 second capture of the core loop, no voiceover, captions burned in. For anything with an interaction worth seeing, this beats every static option — and “record your screen while doing the thing” is a skill you already have.

What to avoid: stock photography of people at laptops, abstract gradients with a slogan, and your logo as the focal point. None of them tell a stranger what your software does.

The specs you actually need

You don’t need a spec sheet. You need three exports and a few character counts. Meta’s own ad guide lists these as the recommendations for the main placements:

What	Ratio	Resolution	Notes
Feed image	4:5	1440 × 1800	Takes more vertical space than 1:1 on a phone
Square (fallback)	1:1	1080 × 1080	Safe everywhere, wins nowhere
Reels / Stories	9:16	1080 × 1920	Keep text away from top and bottom — UI covers it
Primary text	—	—	Meta recommends roughly 50–150 characters
Headline	—	—	About 27 characters before truncation

Two practical consequences. First, export the 4:5 and the 9:16 yourself rather than uploading one square image and letting automatic cropping decide — auto-crop doesn’t know which part of your screenshot matters. Second, that 27-character headline is brutally short. “Ad reporting that matches your Stripe” is 36 and already too long. Write the headline as a constraint, not an afterthought.

Text on images: the rule is gone, the restraint isn’t

Meta dropped its old 20%-text-coverage rule back in 2020, so an image dense with text won’t be rejected for that reason any more. But “allowed” and “effective” are different questions, and Meta’s guidance still leans towards less text on the image itself.

The reasonable middle: one short line burned onto the image — the hook — and let the primary text field carry the explanation. On Reels and Stories, captions are the exception; a lot of that inventory is watched muted, so burned-in captions on a screen recording aren’t decoration, they’re the audio track.

Meta will edit your creative unless you stop it

Worth knowing before you upload: Advantage+ creative enhancements are applied by default. Depending on what’s enabled, Meta may brighten your image, expand its background to fill a placement, generate text variants, or add overlays — then serve whichever version it thinks will perform better per impression.

For a lot of advertisers that’s a free lift. For a product screenshot it can be actively bad: background expansion invents pixels around a UI, and image “touch-ups” can shift the colours of your own interface. Open the enhancements panel, preview the samples, and switch off the ones that mangle a screenshot. Keep the ones that only rearrange your assets.

Test the hook, and test fewer things

The instinct is to make eight creatives and let the algorithm sort them out. On a small budget that produces noise, not answers. Some arithmetic — say your cost per trial signup is €30 and you’re running €20/day:

Variants	Days	Total spend	Signups per variant
6	14	€280	~1.5
3	14	€280	~3
2	21	€420	~7

None of those are statistically comfortable, but only the last one is even in the neighbourhood of a readable result. Two variants, run long enough, beats six variants run until you get impatient. And when you do vary something, vary one thing — ideally the hook, since that’s where the range is. Two ads with different headlines teach you something. Two ads that differ in headline, image, colour and call-to-action teach you nothing except which bundle won.

The same budget logic applies to the campaign around it: if you’re not sure how much runway a test needs, how much a small SaaS should actually spend on Facebook ads works through the minimum-data problem in more depth, and your first week of ads covers what to leave alone while a test is running.

A one-hour production routine

Once a month, block an hour:

1. Screen-record your core loop once, cleanly (10 min). One take is fine.
2. Cut two clips from it — a 15-second and a 25-second — and burn in captions (20 min).
3. Screenshot three key screens, crop tight, add one annotation each (15 min).
4. Export each still at 4:5 and 9:16 (10 min).
5. Write six hooks. Keep the two you’d actually stop scrolling for (5 min).

That’s a month of creative from one hour, produced entirely with tools you already have. It won’t win a design award. It will out-perform the gradient-and-slogan ad you were about to spend an afternoon on.

Where Flowjat fits

Creative testing only means something if you can see which version produced revenue rather than which one produced clicks — and a good-looking CTR on a creative that brings in refund-heavy signups is a trap. Flowjat ties each ad back to the money that actually arrived, so a hook test is judged on ROAS against your break-even line instead of on engagement metrics that grade their own homework.