I’m building Glolly, which I think of as an operating system for small businesses in Ghana. We’re starting with a storefront product for sellers who currently run their operations through Instagram, TikTok, and WhatsApp DMs. The platform has been in production for about two weeks. We have one active seller who has done close to GHS 4,000 in sales on the platform in that window, plus a handful of stores in test mode that I don’t expect to convert. The thing I need to do next, urgently, is find more of the right kind of seller.
This post is about how I’m planning to do that. Specifically, the cold outreach playbook I’ve designed for the next four weeks. The playbook hasn’t run yet. I’m publishing it now, before I send the first message, on purpose. I’ll report back in four weeks with the actual results.
A note on why I’m writing this: the post is mostly aimed at other founders, builders, and people who follow what I’m doing. I’m sharing the playbook because I think the value in keeping it secret is approximately zero, and the value in publishing it (public accountability, feedback from people who’ve done this before) is real. If you have experience doing cold outreach in West African markets and you think parts of this are wrong, I want to hear from you.
At my previous startup, my approach to customer acquisition was brute force. I scraped contacts, cold-called with a pitch, and repeated. The conversion rate sat somewhere in the 1-5% range, which is terrible. We acquired customers, but only because we were calling so many people that some of them said yes despite the approach, not because of it.
I now think I know why that approach converted badly, and it’s not the obvious answer of “cold calling is bad.” It’s that someone receiving an unsolicited pitch in the middle of their day is in a defensive posture by default. They aren’t evaluating whether the product is useful to them. They’re evaluating whether they need to get off the phone. Any decision they make in that state is a defensive decision, not an informed one. Saying yes is a defensive decision sometimes. Saying no is one more often. Either way, you don’t learn anything about whether your product actually fits.
I built that earlier playbook out of ignorance, not strategy. The plan was “have a script, call a thousand people, do this every day.” I knew the conversion rate was low, but I didn’t know it could have been structurally different. I do now.
For Glolly, I want something different. I want prospects who are open to the conversation, eager even, who arrive at the call already thinking about the problem I solve rather than the fact that I’m interrupting their day. The plan is to have something like the same number of conversations I had at my previous startup, but with qualified, interested leads and watch the conversion rate move from 1-5% to something like 30%.
Whether that math actually works is the experiment.
There are five steps. I’ll walk through each one and explain why it’s designed the way it is.
flowchart TD
A[1. Identify] --> B[2. Engage]
B --> C[3. Message]
C --> D[4. Wait and follow up]
D --> E[5. Handle reply]
E --> F[Book the call]
Before any outreach happens, the prospect has to pass a written qualifier checklist. Based in Ghana. Actively posted products in the last seven days. Visible volume signals (sold-out announcements, restocking posts, customer reposts, multiple drops per week). Engagement on their posts that looks like buyers, not just lookers (“how much?”, “still available?”, “DM me”). No website, or a poor one. A follower-count floor of around 1,500-2,000, which I’m using only as a spam filter, not as the actual qualifier.
The reason the qualifier is volume-of-orders rather than follower count is that the pain we solve only kicks in at a certain volume. A seller doing five orders a week can manage DMs in their sleep. A seller doing ten or more orders a day cannot. The friction is constant and growing. That’s the inflection point where Glolly goes from “nice to have” to “I need this yesterday.”
When I add a seller to the pipeline, I screenshot their bio and one recent post. That material gets used later in the personalized message.
Two engagement bursts, twenty-four hours apart. The first burst is on Day 1 in the morning: follow the account, reply to their most recent Story, leave one substantive comment on a recent post, like and save a different post. The second burst is Day 2 in the evening: one substantive comment on a different post, plus a Story reply if a new one has been posted.
The research on cold DM outreach generally recommends three to five days of engagement before messaging. I’ve compressed this to two days. I know that may materially hurt the results (possibly meaningfully) but I want a faster cadence right now. I’d rather find out in four weeks that I compressed too much and adjust, than wait a week longer per prospect when I’m still finding my footing on the rest of the playbook. This is one of the deliberate tradeoffs I’ll be checking in the followup post.
There are two skip conditions. If the seller has a public WhatsApp number listed on their profile, I skip the Instagram engagement bursts entirely and go straight to messaging on WhatsApp. If at any point during the engagement bursts the seller reciprocates (likes my comment, replies to my Story, follows back), I skip the remaining wait and message immediately. Reciprocation is a real signal that they’ve noticed.
The message has four parts: a one-sentence intro (who I am, what Glolly is), a two-sentence specific observation about their business that proves I’ve actually engaged with their content, a one-sentence token framing plus the delivery, and a single question at the end designed to get them talking about their business.
Three things about this. The specific observation has to be something only someone who actually engaged with their content would know. Not “love your products.” Something tied to a specific post, product, or pattern. Without this, the message reads as templated, and the rest of the work was wasted.
The token is one of two options, which I’m A/B testing. Token A is a Ghost Store recording — a short screen recording of a pre-populated storefront I’ve built on staging, using the seller’s actual products, brand colors, and styling. It has higher emotional impact and much higher effort to produce. Token B is a checkout audit — a 90-second walkthrough of the seller’s current ordering flow as a customer would experience it, calling out two or three specific friction points. It is lower effort, but the tone is harder. Critique can land as insult if it isn’t framed carefully. I’m only running two tokens because those are the only two I’ve come up with that feel substantive. If I could think of four, I’d test four. The point is to find which one actually books more calls per hour of effort invested, and kill the other.
The closing question is not a request for a call. That’s deliberate. Asking for a call in the first message may convert badly because the prospect hasn’t invested anything in the conversation yet. The question is something easy to answer that opens a thread — roughly how many orders are you handling a week right now? The call ask comes later, in message two or three, after they’ve responded.
I wait four to five days for a reply before sending follow-up one. Maximum of two follow-ups, each spaced four to five days apart. Every follow-up has to add new value or a new angle — no “just bumping this.” The final follow-up explicitly closes the loop with something like “I’ll stop bugging you. If anything changes, here’s my number.” That removes pressure and leaves the door open for a graceful re-entry weeks or months later.
After two follow-ups with no reply, the prospect is marked cold and I stop messaging. They can come back into the pipeline in 60+ days if circumstances change.
Replies fall into four patterns, and each gets a different response.
A polite acknowledgment (“Cool, thanks”) is not a call-ready reply. Asking for a call here gets a soft no. Instead, I send the question that opens up their business — quick question while I have you, roughly how many orders are you handling a week right now? The call ask comes in message three or four, after they’ve invested in the conversation.
A question reply (“Tell me more / how much / how does this work?”) is call-ready. The trap here is answering in DMs. If I start explaining pricing or features over text, the conversation has no natural endpoint and the call never happens. The redirect is something like “Good question — the honest answer depends on your specific setup. Easier to walk through on a quick call. Tuesday 4pm or Wednesday 11am work?”
A substantive reply (“Yeah, I do struggle with this”) goes straight to the call ask.
A soft no (“Not right now / maybe later”) gets a warm parking message: “Got it. I’ll check back in two weeks. Anything useful I can send you in the meantime?”
In four weeks, I’ll publish a followup post with the actual numbers:
I’ll publish that post regardless of whether the results are flattering. Honest “this didn’t work” posts are more useful than waiting indefinitely for a story with a clean ending.
If you’ve done cold outreach in West African markets, or for non-technical buyers in any emerging market, and you think parts of this are wrong — please tell me. I’d rather find out the playbook has a hole in it before I run it on 30 sellers than after.
If you’re building something for a similar market and want to compare notes, same thing. I think there’s an under-served conversation about how customer development actually works when your buyers don’t read SaaS blogs and don’t live on LinkedIn, and I’d value being part of it.
See you in four weeks with results.