# Driving product adoption Youtrust · 2023 · Lead Product Designer · Case study Source : https://albancarmet.com/case-study/youtrust-adoption Reducing friction on a key feature and improving activation in self-serve SaaS. In one line : From 0 of 9 to 3 of 3 in testing. Team : PM, EM, 3 engineers and me · Duration : 2 months · AI : None ## At a glance - Problem : Forms, Youtrust’s document automation product, had launched but barely converted: fewer than 7% of free trials became paid subscriptions, and most users never finished creating a form. - What I did : Crossed four sources — funnel, survey, internal workshop, usability tests — to find where people got stuck, then redesigned the creation flow around those blockers. - Result : In testing, no one could create a form unaided before the redesign (0 of 9); after it, everyone could (3 of 3). Forms was then sold in self-serve, without a sales rep. - Limit : Round 2 had only three usable sessions, and adoption in production was never measured. ## The company Youtrust (formerly Yousign) is a European digital trust platform: electronic signature, identity verification and electronic seals. More than 30,000 companies across Europe use it, from Allianz and Fnac Darty to Payfit and Qonto. - Sector : B2B SaaS · e-signature - Customers : 30,000+ companies in Europe - My time there : 2021 – 2023 · Lead Product Designer ## The product : Forms Forms is Youtrust’s document automation product, built for HR and legal teams who send the same document over and over: a non-disclosure agreement, a promise to hire. Launched in 2022 as Workflows, it was Youtrust’s first product beyond e-signature. - Prepare the template : In Word or Google Docs, the creator puts what changes each time in brackets: [Company name], [Address]. - Get a form : Once the document is uploaded, Youtrust turns each bracketed variable into a question. - Fill in, then sign : Each answer to the form produces a completed copy of the document, sent straight to e-signature. Example: legal prepares the NDA template once. For each new partner, a sales rep fills in the form with the partner’s details, and the NDA goes out for signature, with no copy-pasting in Word. ## Process Each step fed the next. None of them used AI: this was 2023. - Diagnose : Funnel analysis with the PM, a survey of trial users, an internal workshop. (without AI) - Test : Round 1 on production with 9 users, to confirm where they got stuck. (without AI) - Redesign : The creation flow, rebuilt around the blockers with the design directors. (without AI) - Validate : Round 2 on a Figma prototype: three usable sessions. (without AI) - Ship : Shipped with an in-app onboarding; Forms goes on sale in self-serve. (without AI) ## The problem — Forms had launched, but almost nobody got through creating one Forms was moving to self-serve: customers would discover it, try it and subscribe on their own, with no sales demo to make up for friction. And the numbers said friction was everywhere. ## The approach — Find where people got stuck before redesigning anything Four sources, checked against each other, and they agreed: the value of Forms was not in question. The creation flow was, and one screen above all. ## The trade-offs — Three decisions, and what each one cost Each redesign choice came from a test finding, and each gave something up. The third one is about the research itself: knowing when to stop testing. ## What shipped — One kind of decision per screen The same principle carried through the rest of the flow: settings split by what they configure, and guided onboarding on the key screens. ## Constraints - Two other product teams in the same flows : Every change had to fit what they were building; the signer step, for instance, was theirs. - A light scope, on purpose : Ship and learn from real usage rather than perfect the flow in testing. - Editing happens outside Youtrust : Users leave the app mid-flow to prepare their document in Word or Google Docs, at the step with the worst drop-off. ## Signals and research Four sources, each checked against the others: product data to see where people dropped off, a survey and an internal workshop to understand why, then usability tests to confirm it with real users. - Funnel analysis — The upload step lost most users : Only 31% got past document upload, against 88–95% for every other step — and clearing it took over 7 hours on average: users left the app to prepare their document, and didn’t always come back. (31% : got past document upload, against 88–95% for every other step.) - Usability tests · round 1 — No one created a form unaided : All 9 testers got lost at least once, and 8 couldn’t add variables. Many thought they were filling in a form, not building a template. (0/9 : created a form unaided, and 8 of them couldn’t add variables.) - Survey — A weak signal, kept as such : Only six answers: too few to decide on, so they could confirm a finding but never lead one. Lack of time came first, and 4 of 6 had never seen a demo. (6 : answers from trial users: too few to decide on. They could confirm a finding, never lead one.) - Internal workshop — Too much on one screen : Product, design and marketing role-played HR and legal users on fake documents. Their top complaint matched the funnel: the upload page asked for too much at once. (3 teams : role-played HR and legal users: the upload page asked for too much at once.) ## Decisions and trade-offs - Prepare the document step by step, not on one dense page. Decided with : The design directors. Rejected : One upload page — Every instruction on one text-heavy screen: mark variables in brackets, save as .docx, drop the file. Kept : Three guided steps — An overview of the process, then how to insert variables in Word or Google Docs, then which format to save in. Why : The funnel and round 1 pointed at the same screen: 69% of users stopped there, and 8 of 9 testers couldn’t add variables. People skimmed the instructions, so the fix was fewer per screen, not more. What it cost : A longer path: three screens to go through before uploading anything. A deliberate detour, so the instructions actually land. In hindsight : In round 2, all three testers marked variables and saved in .docx without help. The main blocker was gone. - One customization step instead of three back-and-forth screens. Rejected : Three screens — Write the questions, place signature fields on the document, then come back to edit the form — with empty form fields visible from the start. Kept : One step, edited inline — Variables, questions, success page and signers in one view, edited inline, with wording that says you are building a template. Why : Testers lost track in the back-and-forth, and seeing empty form fields made them think they had to fill the form in themselves. What it cost : One sub-task per screen: the merged step is denser. In hindsight : All round 2 testers understood the step, and how to turn variables into questions. - Move forward after three test sessions rather than wait for a bigger panel. Rejected : A complete round 2 — Recruit more testers and make every planned design change before moving on. Why : Time was short, and the scope had been kept deliberately light: we had little to lose by shipping and learning from real usage. What it cost : Certainty: three sessions show a direction, they don’t prove it. And some of the planned changes didn’t make it in. What makes up for it : The two hypotheses that failed went to the right owners: placeholder signers (signers whose details are collected by the form) to the team in charge of that step, using a new form to an in-app onboarding. ## What shipped - Settings split by what they configure (Form settings and signature request settings, each in its own step.) : problem : Form and signature request settings shared one long page, which blurred what users were actually building. Solution : I split them into two steps, on the same principle as the rest of the redesign: one kind of decision per screen. - Guided onboarding in the product (Step-by-step tooltips on the key screens of form creation.) : problem : Even after the redesign, defining who signs the generated document stayed the least understood concept. Solution : We added contextual tooltips on the key screens, starting with the signer step, so users know what each stage expects before they act. ## Results Before the redesign, no tester could create a form on their own. After it, all of them could. (Usability tests · round 1 in production (9 users) → round 2 on prototype (3 users)) - 0/9 → 3/3 : Testers who created a form unaided - 6/8 : Hypotheses validated in round 2 - 1/9 → 3/3 : Knew how to insert variables — the step where 69% of users dropped off - 3/3 : Understood the customization step, the most confusing one in round 1 Three sessions show a direction, not proof. What worked: - Forms went on sale in self-serve. What didn't work: - Placeholder signers stayed unclear to 2 testers out of 3, and nobody knew what to do with a form once created. - Forms, meant as a standalone product to grow ARR, became a premium add-on with a cap on documents. - Launching multi-respondent forms before fixing adoption cost us the time we needed for the rest. ## What I took from it I never closed the loop. Neither the funnel nor adoption in production was measured again after shipping: the team moved on to other priorities, and I left Youtrust. Today I’d track activation (forms actually put online) from day one, and come back to it even once the team has moved on.