Insight scattered across tools
Rankings here, content there, competitors nowhere — the connections that drive decisions lived only in a strategist’s head.
Case StudyMedia & Communication
SearchWeft: A Search Intelligence Platform, MVP-First
SearchWeft is a search intelligence platform that weaves together the threads SEO teams usually juggle in separate tools — rankings, content inventory, and competitor movements — into one actionable view. We built it MVP-first: a focused product in weeks, validated with real users before deeper investment.
At a Glance
- Product
- Search intelligence SaaS
- Users
- SEO & content teams
- Engagement
- Product scoping → MVP → iteration
- Approach
- Fixed-scope MVP sprint
- Core view
- Rankings × content × competitors
Background
The Context That Made This Worth Building
SEO teams live in a patchwork: one tool for rank tracking, another for content audits, spreadsheets for competitor notes, and judgement stretched across all of them. The insight behind SearchWeft was in its name — the value isn’t another data source, it’s the weave: showing how your rankings, your content inventory, and competitor movements interact on one screen, so the next action becomes obvious.
As with every product bet, the question wasn’t whether the vision was appealing — it was whether real teams would use it weekly. That made it a textbook case for an MVP sprint: build the smallest weave that proves the habit, and let usage data argue for what comes next.
The Challenge
The Pain Points That Started It
Analytics products fail in a specific way: dashboards that impress in demos and go unopened by week three. The build had to attack that failure mode directly.
Dashboards without verdicts
Charts are cheap; conclusions are rare. The product had to say “this page is slipping and here’s the likely why”, not just plot lines.
A crowded, cynical market
SEO tooling is saturated. Surviving meant a sharp wedge — the woven view — shipped fast enough to test before the window of conviction closed.
Discovery
Understanding Before Building
Discovery centred on the weekly ritual of SEO teams: what do they actually look at on Monday morning, and what decision follows? The recurring answer — “which pages moved, and was it something we did or something a competitor did?” — defined the MVP’s single screen: movement, mapped to your content changes and competitors’ visible changes, with an AI-drafted explanation as the starting hypothesis.
Scope-cutting was ruthless. No backlink module, no site-audit crawler, no report builder — every incumbent has those. Version one earned its place with one thing done properly: the weave, updated automatically, readable in five minutes a week.
The Solution
What We Designed and Built
A focused product with three layers: the data spine, the woven view, and the AI narration that turns movement into hypotheses.
Data spine
Automated collection of ranking positions, content inventory snapshots, and competitor page changes — normalised into one timeline per topic.
The woven view
One screen answering Monday’s question: what moved, what changed on our side, what changed on theirs — filterable by topic cluster and page.
AI movement narration
Draft explanations for significant movements — correlating your edits and competitor changes — as reviewable hypotheses, not oracle pronouncements.
Scope Delivered
Everything That Shipped
- Multi-source data ingestion & normalisation
- Topic & page-level movement timeline
- Competitor change detection & snapshots
- AI-drafted movement explanations
- Weekly digest — the five-minute Monday read
- Team workspace with tracked topics
- Subscription billing & onboarding flow
- Marketing site for launch
Implementation
How the Build Ran
- 1
Wedge & scopeWeek 1
The Monday-morning ritual defined the single core screen; everything else cut or staged. Fixed scope, fixed price, launch date set.
- 2
Data spineWeeks 2–3
Ingestion, normalisation, and the movement timeline — the unglamorous layer everything readable depends on.
- 3
Weave & narrationWeeks 4–5
The core view and AI explanations built and tested against real site histories, tuned until hypotheses read as useful rather than confident nonsense.
- 4
Launch & learnWeek 6 onward
Live with early teams; the weekly digest emerged from usage data showing when users actually engaged — the first post-launch iteration that stuck.
Stack
- Next.js
- React
- Node.js
- PostgreSQL
- OpenAI
- Vercel
Results & ROI
What It Delivered
~6 wksScope to live MVP
3Data threads woven into one view
5 minThe Monday read it replaced hours with
1Screen that answers “what moved and why”
- The core hypothesis — teams want the weave, not another dashboard — validated with real weekly usage before deep investment
- AI narration positioned as reviewable hypothesis, earning trust instead of burning it
- A codebase built to scale past the MVP: new data threads plug into the existing spine
- Feature decisions after launch made by usage evidence, not roadmap debates
SearchWeft demonstrates what MVP discipline is actually for: not building less, but learning sooner. Six weeks bought a real answer about the product’s core habit loop — the woven Monday read — and every subsequent feature argument now has usage data to referee it. In a saturated market, that evidence-speed is the moat.
Takeaways
What This Project Teaches
Build the ritual, not the dashboard
Products survive on habits. Anchoring the MVP to an existing weekly ritual — the Monday review — gave it a slot in users’ lives instead of competing for a new one.
AI as hypothesis, never oracle
Framing AI explanations as drafts to review made occasional misses acceptable and frequent hits valuable. Overclaiming would have spent the product’s trust in a week.
Cut what incumbents already do
Every feature the market already has is a distraction at MVP stage. The wedge earns attention; parity features can follow revenue.
Before You Ask
What Readers Usually Want to Know
What happens if the MVP shows the idea does not work?
You have saved the far larger investment that would have followed, which is the entire point. An MVP sprint is priced and scoped so that a negative answer is affordable. In practice the result is rarely binary — usage data normally shows which part of the idea has a pull and which part nobody wanted, and that is a far better basis for the next decision than conviction.
Is MVP code throwaway? Will we rebuild in a year?
No — we build on the same production stack used for full platforms, so a validated MVP scales rather than restarts. What gets deliberately skipped is polish and secondary features, never foundations. Rebuilding from scratch after validation is a sign the first build cut the wrong corners.
How do you keep an AI feature useful rather than confidently wrong?
By framing it as a hypothesis the user reviews, not an answer they must trust. In SearchWeft the AI drafts an explanation for a ranking movement and the user confirms or dismisses it. That framing makes occasional misses acceptable and frequent hits valuable — overclaiming would have spent the product’s credibility in a week.
The Services Behind This Build
Need the Same Capabilities?
Want a Result Like This?
Tell us your pain points — we’ll walk you through how we’d scope it, what it would cost, and what a realistic outcome looks like. Honestly, both ways.