Bronze, a MENA fashion brand, grew its Instagram DM sales 13x in a single month after putting ReplAi in front of its inbox, with zero new support hires. It is the strongest result we have measured, a top outlier: the median merchant in the same Q1 2026 cohort saw +150%. Here is what actually changed.
What was Bronze's situation before automation?
Bronze is exactly the profile of brand for which DMs are the storefront: a fashion label whose customers discover products through Instagram posts, Reels, and stories, then message to ask about prices, sizes, fabrics, and delivery before buying. The inbox was not a support side-channel; it was the top of the sales funnel.
That funnel had the standard shape we see across social-first fashion brands:
-
High message volume clustered in evenings and late nights, when the team was smallest or asleep.
-
A large share of repetitive questions: availability, sizing, delivery times, and "where is my order?" (WISMO questions are commonly 25 to 50 percent of ecommerce ticket volume, per industry reporting).
-
Customers writing in Arabic dialect and Franco-Arabic as often as English.
-
Cash on delivery as a dominant payment expectation, meaning orders had to be taken conversationally, not just linked out.
What was the actual problem?
Not effort. The team answered messages; they just answered them late, because humans process a chat backlog serially and sleep at night. And in DM commerce, late is lost: the buying impulse that triggered the message decays within minutes, and customers message competing stores in parallel. For the mechanics of why, see why response time decides the sale.
The result was a quiet leak rather than a visible failure. Measured across our merchants, the baseline DM-to-order close rate without automation is about 3 percent. The other 97 percent of conversations are not refusals; most are questions that were answered too late or never. Bronze's inbox was full of money evaporating on a timer.
The conventional fix, hiring more agents, has an obvious ceiling: coverage across 24 hours, multiple dialects, and volume spikes scales payroll linearly forever, and still cannot answer in seconds.
What changed when ReplAi went in?
The deployment itself was unremarkable, which is rather the point: one-click install from the Shopify App Store, Instagram connected through Meta's official OAuth flow, tone and escalation rules set in the visual configuration, live in about 10 minutes. From then on:
-
Every message got an answer in seconds, at 3pm and 3am alike. Measured across our merchants, replies average under 14 seconds.
-
Answers came from live store data. Product and catalogue search, real stock levels, real order status for WISMO, returns started in the thread.
-
Orders closed inside the conversation. Including COD orders placed directly in the chat, which matched how Bronze's customers actually wanted to pay.
-
Language stopped being a bottleneck. The agent answers in Egyptian, Gulf, and Levantine Arabic, Franco-Arabic, and English, with gender-aware addressing, so no thread waited for the right speaker.
-
Humans kept the judgment calls. The team took over sensitive conversations from the same inbox, with AI toggled off per thread. Across our merchants, about 80 percent of messages are fully automated; the rest route to people.
Nothing about Bronze's marketing changed in that month. The change was that the demand their content was already generating stopped leaking away between message and reply.
What were the results, honestly framed?
| Metric | Result |
|---|---|
| Bronze DM sales, one month | 13x (top outlier) |
| New support hires at Bronze | Zero |
| Q1 2026 cohort, top quartile DM revenue lift | +300% |
| Q1 2026 cohort, median lift | +150% |
| Q1 2026 cohort, bottom quartile lift | +75% |
The owner's words: "Our DM sales grew 13x in a single month. ReplAi just kept closing while we slept."
Now the honest framing, because we would rather you plan on real numbers. Bronze is the top outlier of the Q1 2026 pilot cohort. The 13x combines an unusually favourable setup: very high inbound DM volume relative to what was being converted, a nocturnal customer base the team could not cover, dialect-heavy conversations, and COD buyers who needed in-chat ordering. The more leak you have, the more automation recovers. A typical merchant should anchor on the median, +150%, meaning DM revenue roughly two and a half times its previous level, with the bottom quartile still at +75%.
What should other merchants take from this?
Three transferable lessons:
-
Audit your leak before buying anything. Count last week's DMs, then count how many became orders. If you are near the ~3 percent baseline with meaningful volume, the upside is sitting in your inbox already.
-
Speed plus data beats speed alone. An instant "we'll get back to you" recovers nothing. The wins came from instant answers containing stock, price, and order status, and from closing COD in the thread; the full flow is described in how to take orders in Instagram DMs.
-
Keep humans in the loop by design. Bronze did not fire anyone into a bot; the team moved up the stack to complaints, VIPs, and edge cases while automation absorbed the repetitive 80 percent.
If your inbox looks like Bronze's did, high volume, late-night traffic, dialects, COD, you can test the same setup on the free plan (100 replies a month, details on the pricing page), or book a demo and we will run it live against your own catalogue and real customer questions.
Frequently asked questions
Is the 13x result typical?
No, and we say so deliberately: it is the top outlier of our Q1 2026 pilot cohort. The honest planning numbers are the cohort quartiles: +75% at the bottom, +150% at the median, +300% at the top. Merchants with big unanswered volume and off-hours traffic sit toward the high end.
Did Bronze reduce its support team?
No, and that was never the goal. The result was zero new hires despite handling far more conversations, with the existing team focused on the roughly 20 percent of messages that genuinely need human judgment, taken over from the same inbox with AI off for those threads.
How long did it take to see results?
The measurement window here is one month from going live, and setup itself took about 10 minutes. Practically, expect the first week to include reviewing conversations and tightening tone and escalation rules; the revenue effect follows the coverage, which begins immediately.
Would this work for a store without Arabic-speaking customers?
Yes; the core mechanism, instant answers with live Shopify data and in-chat closing, is language-independent. Arabic dialect support mattered enormously for Bronze's audience specifically, but English-market merchants leak sales through the same gaps: slow replies, night traffic, and unanswered WISMO.