Research date: August 2, 2026 Question: What is the best way to use multiple lead magnets to build an email list, and is the dropdown selector on boulderthingstodo.com the right presentation? Method: Six parallel research threads (content upgrades, choice architecture, placement mechanics, delivery and segmentation, competitor practice, contrarian case). Roughly 180 sources consulted, about 30 live competitor pages fetched directly. Overall confidence: MEDIUM, and deliberately uneven. Confidence is stated per claim below because it varies enormously.

---

The short answer

No, the dropdown is not the right presentation. But it is the fourth most important problem with the current setup, and fixing it alone would be polishing something while three larger things stay broken.

In order of what actually costs you subscribers:

1. You give the PDF away free directly above the form that asks for an email to get it.
2. The form sits in the lowest-attention part of the page, in the weakest-converting format there is.
3. The five guides select for tourists, and tourists are the low-value subscriber.
4. The dropdown hides your only sales tools (the guide titles) behind a closed control.

And underneath all four: the list has never been mailed, which means it is currently worth zero and decaying.

---

1. You are competing against yourself, and losing

This one needs no research, just a look at the page.

On all five guide pages, BaseLayout.astro:944 renders a bar reading "Printable one-page version, Download the PDF" as a plain ungated link. A few hundred pixels below it, the signup panel says "Get the printable version" and asks for an email to get the same file.

Verified live: https://boulderthingstodo.com/guides/trailhead-parking.pdf returns HTTP 200 and 62KB to anyone, no email required.

So on the exact pages where a reader is most motivated to want the printable, the printable is already free above the fold of the form. Nobody rational trades an email for a file they were just handed. No amount of dropdown tuning, placement tuning, or copy tuning fixes this.

Two honest ways out, and I recommend the second:

Option A: gate the PDF. Remove the free download bar. This makes the current offer coherent. But it contradicts the comment you wrote in that very file ("the facts are public either way, and hiding the download would be a worse experience for no gain"), and it costs you the SEO and AI-citation surface that an indexable, linkable PDF provides. For a traffic property, that is a real loss.

Option B: keep the PDF free and change what the email buys. Stop pretending the email purchases the download. It doesn't and shouldn't. Make the email buy something email is actually good at: knowing what is happening in Boulder before you get there. This is the option that also solves problem 3 below, which is why I prefer it.

Confidence: HIGH. This is a directly observed contradiction on your live site, not an inference from benchmarks.

---

2. Placement is costing you more than the dropdown is

The format. Omnisend's cross-account data puts embedded in-page forms at 1.28%, against 3.24% for flyouts, 3.77% for popups, and 6.47% for dedicated landing pages. Their device-level breakdown for embedded forms is harsher still (0.12% desktop, 0.17% mobile). Your indigo panel is an embedded form. It is sitting on the floor of the benchmark range, and that is structural, not a design failure.

The position. Nielsen Norman Group's eyetracking research found 57% of page-viewing time happens above the fold and 74% within the first two screenfuls, with the remaining 26% spread thinly down the rest of the page. A panel above the footer on a long Boulder listicle is competing for a slice of that last 26%. The problem is not that it converts poorly among people who see it. The problem is that few people ever see it.

The fix that respects your no-sticky rule. Move the primary offer into the article body, roughly a third of the way down, right after a section that delivered something genuinely useful. Keep it as a designed callout block in normal document flow. Nothing pinned, nothing floating, nothing overlaying. This is the single highest-leverage change available to you under your own constraints.

The honest cost of the no-popup rule. Giving up popups entirely costs roughly 1.5 to 3.5 percentage points of visitor-level opt-in versus a well-run popup program. On 10,000 monthly visitors that is on the order of 150 to 350 signups a month. I think it is a defensible trade for a site whose whole value proposition is being a pleasant place to plan a Boulder weekend, and moving the offer in-content recovers a meaningful share of it. But it is a real trade, and you should know the number rather than assume it is free.

A warning about the benchmarks above. Every quantitative source in this area sells email-capture software and measures only its own customers, who are self-selected optimizers. Worse, they do not agree on what "conversion rate" means. Popupsmart headlines a "3.49% average popup conversion rate," but their disclosed method is leads divided by interactions. Their raw dataset is 105,277,811 displays and 720,189 conversions, which is 0.68% at the visitor level, five times lower than their own headline. Treat every number in this section as a prior for what to build first, not as a forecast.

Confidence: MEDIUM-HIGH on direction (embedded forms are weakest, and attention concentrates at the top: both replicate across independent sources, and NN/g has nothing to sell). LOW on any specific percentage.

---

3. The deepest problem: your magnets select for the wrong audience

This is the finding I would keep if I could only keep one, and it is not about presentation at all.

The only proven business model for a local email list is 6AM City: roughly 450,000 subscribers, about $9M projected revenue, 85% of it from local advertising, at an ARPU target of $10 per subscriber per year in mature markets. That works because their subscribers are residents, who local advertisers want to reach every single week.

Your five guides are, all five of them, tourist offers. "Trailhead parking cheat sheet," "48 hours in Boulder," "what a day costs": these are perfectly matched to a visitor with a one-week problem horizon. They plan, they visit, they leave, and they never open an email from you again. That subscriber is close to worthless to a Boulder advertiser and close to worthless to you.

"What's on in Boulder this weekend" is a resident offer. So is anything about seasonal closures, new openings, or trail conditions. You already have a /this-weekend/ page, which means you are already producing the resident-shaped content. You are just not asking anyone to subscribe to it.

My arithmetic on what the current list is worth, derived from published CPM ranges rather than taken from a single source: 1,000 local subscribers, weekly send, one sponsor slot at a $20 CPM is about $1,040 a year gross at 100% fill. At a realistic 40% to 60% fill rate, that is $400 to $600 a year, or roughly $0.40 to $0.60 per subscriber per year, and only if you actually publish weekly. A tourist list earns a fraction of that, because the audience has left town.

Confidence: MEDIUM. The 6AM City figures come from press interviews rather than audited filings, but they are specific, internally consistent, and the closest available comparable. The audience-mismatch reasoning that follows from them is mine, and it is an argument rather than a measurement.

---

4. The dropdown: wrong, but not for the reason you'd expect

First, kill the argument you were probably about to make. The "paradox of choice" does not apply here and you should not use it. Two meta-analyses settle this: Scheibehenne, Greifeneder and Todd (2010) pooled 63 conditions across 50 experiments (N = 5,036) and found a summary effect of D = 0.02, statistically indistinguishable from zero. Chernev, Böckenholt and Goodman (2015), 99 observations (N = 7,202), likewise found no significant effect without moderators. The famous jam study has never cleanly replicated; the same author's larger follow-up returned an effect of 0.02 against the original's 0.77. And every study in that literature compares roughly 6 options against 24 to 30. Nothing in it speaks to 5 versus 1.

Worse for the overload argument: Chernev found that one of the strongest suppressors of choice overload is the presence of a dominant option, which is exactly what your pre-selected default provides. On a trailhead parking page, the parking guide is the dominant option. The precondition for overload is not met.

Now the reasons the dropdown is genuinely wrong, which are better evidenced:

• Nielsen Norman Group's threshold is explicit. For five or fewer mutually exclusive options, use radio buttons, not a dropdown. Baymard Institute, drawing on 200,000+ hours of usability testing, says dropdowns are "generally a poor choice for offering fewer than 5 or more than 10 options."
• A default does not stop people interacting with it. Baymard's most on-point measurement: 55% of test subjects opened an optional dropdown just to see what was inside, then closed it without selecting, even when it was clearly marked optional. Your default suppresses the deliberation but not the inspection. A label reading "Which guide?" is an interrogative, it sits above the email field, and it gets read first.
• It halves discoverability of the other four guides. NN/g's 179-participant study of hidden versus visible navigation found options behind a collapsed control were used in 27% of desktop cases versus 48% to 50% when visible. You are paying the interaction cost of a chooser and getting none of the merchandising benefit of a menu.
• It hides your only sales tools. A closed <select> shows neither a cover image nor a specific title until the visitor works for it. Publishers sell downloads with exactly those two things.
• Nobody credible does this. Across roughly 30 live pages fetched (travel blogs, state tourism boards, HubSpot, Backlinko, Smart Blogger, Smart Passive Income, Morning Brew), zero used a dropdown to select a lead magnet. The only <select> elements found in any opt-in form were country and state fields on postal-mail request forms.

What real publishers do instead, both patterns attested:

• The bundle (Nomadic Matt). Dozens of guides, but the sitewide form offers one thing: the "Free Travel Starter Kit." Email only, no choice offered. Zero decision cost, zero segmentation signal.
• The visible checkbox grid (Travel Iowa, FreeTravelGuides). Iowa's form reads "Step 1: Select Your Guides (must select at least one)" over cover-image checkboxes, then collects the email. Multi-select, every cover visible. This is the only attested pattern where multiple named guides genuinely share one form, and it is used by organizations with real budgets.

Confidence: HIGH that the dropdown is the wrong control, because NN/g, Baymard, and the competitor survey converge independently. LOW on how much you would gain by changing it. The only head-to-head test found (MECLABS, roughly 15% in favor of radio buttons) explicitly failed to reach significance, and no controlled A/B of a multi-magnet dropdown versus alternatives exists anywhere in public. Anyone telling you to rip it out is reasoning from analogy, including me.

---

5. The list is currently a liability, not an asset

As built: addresses in a D1 table, no email ever sent, no unsubscribe path, no authenticated sending domain, no bounce handling, no verification, decaying at roughly 23% a year.

The specific risk. The thing most likely to push you over Gmail's spam-complaint threshold is not volume. It is mailing a two-year-old list of people who forgot they downloaded a PDF, from a domain with no sending reputation. That risk is created by the current design, not by growth.

What actually applies to you. The Google and Yahoo rules everyone panics about are tiered. One-click unsubscribe and DMARC only bite at 5,000+ messages per day to personal accounts, which you will not reach for years. What applies at any volume is SPF or DKIM, valid forward and reverse DNS, TLS, and a spam rate below 0.3% (Google's monitoring guidance actually asks for below 0.10%). Any reputable ESP handles the first three automatically at signup. Add a DMARC record at p=none yourself while there is no sending history to break.

Your consent promise is narrower than you may realize. "We keep your address to let you know when these get updated. Nothing else, and no third parties" does not support a newsletter. CAN-SPAM would not catch it, because US law does not require opt-in, but the FTC's deception authority under Section 5 does reach broken privacy promises to consumers, and "nothing else" is about as unambiguous as promises get. If you want a newsletter, the clean path is a separate explicitly-labeled opt-in going forward, not a reinterpretation of this one. Treat that sentence as a strategic choice with a cost, not as neutral copy.

On segmentation by magnet: store the tag, don't act on it. Writing which guide was requested into the row is one column and ten minutes, and it is impossible to recover later. But the case for branching sequences on it is weak: five guides for one small city will produce heavily overlapping audiences, and Mailchimp's segmentation data (the only real dataset, from 2017) compares segmented against broadcast sends, not tagged-by-source against untagged. The genuinely useful thing the tag buys you is product intelligence about which guide people actually want, which is worth more at this stage than a 14% open-rate difference.

Confidence: HIGH on the regulatory mechanics (primary sources, quoted verbatim). LOW to MEDIUM on the list-decay percentages, which come from companies selling email validation and are probably overstated for a small local list.

---

6. The contrarian case, which deserves a hearing

The strongest argument against this entire piece of work is about sequencing, and I think it is partly right.

You have 2 subscribers, and both are tests (your own address and mine, both from July 28). You have no analytics installed at all, so you cannot compute an opt-in rate even if signups start arriving. At a recently launched site's traffic level, doubling the opt-in rate yields a handful of extra subscribers per day. The same hours spent shipping indexed pages compound.

Add the zero-click picture and it gets sharper. US searches ending without a click reached 68% in early 2026 (SparkToro/Similarweb). Pew found that when an AI Overview is present, clicks to traditional results fall from 15% to 8%. 72% of local queries and 74% of informational queries are now zero-click. Semrush found AI Overviews concentrate on low-volume informational keywords, which is precisely the long-tail "things to do in Boulder in November" territory a new traffic property is built on.

So the honest reading cuts both ways. Email is worth more per visitor because search traffic is less reliable, and simultaneously there are fewer visitors to convert, with the shortfall hitting informational content hardest.

What I take from that: do the cheap fixes below because they take a few hours, then go back to content. Do not turn list architecture into a project. And note that the zero-click data argues for keeping your PDFs ungated and indexable, which supports Option B in section 1.

One more steelman for the current design, which is stronger than I expected. No controlled test of a multi-magnet dropdown versus alternatives exists in either direction. The self-selection data it captures is genuinely valuable and free, and as third-party cookies degrade, declared first-party interest is a real asset: "312 people who chose the hiking guide" is sellable to a Boulder outfitter in a way "1,000 subscribers" is not. And one component with one endpoint is honest engineering value across a four-site portfolio.

---

Recommended actions, in order

This week, roughly three hours total:

1. Resolve the contradiction on the guide pages. Keep the PDFs free and indexable (Option B). Stop selling the email as the price of the download.
2. Change what the email buys to something email is uniquely good at and that selects for residents: knowing what is on in Boulder before it happens. You already publish /this-weekend/. Point the offer at that.
3. Install Cloudflare Web Analytics. You cannot optimize what you cannot measure, and you already have this pattern working on Bone Voyage.

This month, roughly a day:

4. Replace the <select> with one visible, named, contextual offer per page, keeping the existing per-page matching logic that already works. Put the other four guides on a real /guides/ index, linked in plain text below the form.
5. Move the offer block into the article body, about a third of the way down, in normal document flow.
6. Send the PDF by email as well as showing it inline. Do not remove the inline download. This verifies the address at capture, is squarely transactional under your existing promise, and establishes your from-address in the inbox before you ever need it. Use a real reply-to address, not no-reply.
7. Store the magnet slug and handle bounces. One column, plus marking bounced rows dead.

Before you ever send a broadcast:

8. Publish a DMARC record at p=none on the domain.
9. Decide the cadence and actually keep it, even twice a year. A list mailed twice a year stays warm and permissioned. A list first mailed in 2028 is a cold-list reactivation, which is the highest-risk send in email.

Explicitly do not do:

• Do not password-protect a resource library. Search engines cannot see inside it, which is disqualifying for a traffic property.
• Do not build branching email sequences by magnet. The audiences overlap too much to pay for the complexity.
• Do not add a name field. Sleeknote's data across 26,270 campaigns puts one field at 4.41% and two at 2.90%.
• Do not cite the "785% lift" content-upgrade case study as evidence. It is a before/after with no control that changed relevance, placement, and mechanic simultaneously.

---

What I could not determine

• No controlled A/B test of a multi-magnet dropdown versus a single or contextual offer exists in public. The case against the dropdown is built on usability research and competitor practice, not on a conversion test of this exact pattern.
• No published opt-in data exists for consumer local-information or travel-guide sites. Essentially every benchmark cited here is e-commerce or B2B.
• The relevance-versus-placement question is confounded in every content-upgrade case study found, including the foundational one. Nobody has isolated relevance.
• Several widely circulated statistics could not be traced to primary sources and should not be used: the "60 to 70% of PDFs are never opened" figure, the "gated content gets 82% less organic traffic and 94% fewer AI citations" claim attributed to BrightEdge, the "760% more revenue from segmented campaigns" figure, and "landing pages with multiple offers get 266% fewer leads" (which is also arithmetically impossible).

Two weeks of your own per-page opt-in data, once analytics are installed, will be worth more than every benchmark in this report.