About the Client
An Australian home goods DTC brand (in-house warehouse, no 3PL) with 3 physical stores in Sydney, Melbourne, Perth — 5–7 sales staff per store. 80% of online lead capture goes through Podium — a customer-comms platform popular in AU retail that bundles SMS, Webchat, Google Business, Instagram DMs.
Their Challenge
- Podium doesn't separate stores: all leads land in one inbox with no 'which store' signal. Podium's official multi-store option means a separate account per store at $400/mo each
- Reps fight over leads: 'I replied first' vs 'I contacted them last week' — 2–3 fights a day over ownership
- Slow response = lost lead: industry data — leads replied within 5 min convert 9× higher than 1-hour replies. Their average was 24 hrs because no one felt 'this is mine'
- Customer experience suffers: customer asks about Sydney stock, Melbourne rep replies 'we don't have that' — Sydney does have it
Why ManifoldX
The owner had quoted Podium's official multi-account option ($400/store × 3 = $14.4K/yr) and explored hiring a routing coordinator ($60K/yr). Our solution is an 'AI attribution layer' on top of existing Podium, $300/mo retainer — 75% cheaper than Podium's option, 95% cheaper than hiring.
The Solution
1. 13-feature attribution — beyond IP
IP alone isn't enough (AU ISPs share IPs across states). We built a 13-feature joint signal: phone area code, postcode, product SKU mentioned, past purchase address, which store address they ask about, which store IG account they follow, Google Business landing page, mentions of specific suburbs… AI returns a confidence-scored attribution.
2. Auto-routing — assigned within 5 min
After attribution, the lead routes immediately to the target store's 'on-shift sales rep'. Each store has its own rota (by hours), AI skips reps who are off-shift / already handling 5 active leads / offline >1 hr. If no reply in 5 min, escalates to store manager.
3. Manual fallback for ambiguous leads
Low-confidence leads (~5%) drop into a store-manager group for human decision. These decisions feed back to retrain the model, confidence threshold tightens month over month.
4. Owner dashboard — per-store KPIs visible
A simple web dashboard showing weekly per-store: lead count / avg response time / conversion rate / top performer.
Tech stack
Working with us
Week 1: client provided 3 months of historical Podium leads — 6,000+ entries, we labelled correct attribution as training set. Weeks 2–3: attribution engine + routing logic. Week 4: dashboard + escalation. Currently in a 4-week trial period, weekly accuracy reviews with the owner.
Sales used to fight over leads every day and I'd have to play referee. Now AI just routes them, escalates in 5 min if no one picks up — that whole internal tax just disappeared. — Client owner (week 2 of trial)
Impact
- 95% attribution accuracy: 5% goes to human review. Each store owner now has a trustworthy record of their lead volume by month-end
- Response time 24h → avg 3 min: based on industry conversion-curve data, projected lead → close conversion lift ~38%
- $14,400/yr saved vs Podium's official multi-store option
- $54,000/yr saved vs hiring a coordinator (at $60K junior ops)
What's next
Once the trial ends, we'll publish a formal attribution-accuracy + conversion-lift report. Next: auto-acknowledge SMS ('Got your enquiry, Eric from Sydney store will reach out in 5 min'), and language detection to route foreign-language leads to multilingual reps. Also being abstracted into a 'multi-store retail attribution template'.