Experiences from eight recent engagements
We commissioned the demand forecasting engagement primarily to help with Q4 procurement. The model picked up a pattern in our sales data tied to the school holiday cycle that we had never formally captured — it was obvious in hindsight. The uncertainty ranges were what I appreciated most; our previous vendor always presented a single number as if there was no variance.
The readiness check was the right first step for us. We went in thinking our data was in reasonable shape. The assessment identified three specific gaps in how we were labelling shipment events — gaps we would not have found on our own before starting a full forecasting project. The written report was clear enough that I could present it directly to our IT team without needing Telvar on the call.
We used the customer behaviour analysis to answer a question we had been sitting on for about a year: whether our lunch and dinner customer bases were actually different segments with different motivations, or the same people at different times. The answer was more nuanced than either hypothesis — there were three meaningful groups, not two. That changed how we thought about our loyalty programme design.
Fixed-price, defined scope, written output. That is exactly what our finance function wanted — no open-ended retainer, no dashboard to license, no training required to read the results. The walkthrough session was efficient: 45 minutes and our planning lead came away confident she could update the model herself at the next quarterly cycle.
We went in wanting churn prediction. The scoping call was honest: our event log data did not have enough labelled churn events to build a reliable model yet. Instead of taking the money, they redirected us to the readiness check first, which gave us a clear remediation plan. We are now three months into that work and will return for the full engagement once the data is in shape.
We had tried to build a forecasting tool in-house and got stuck on the seasonal calibration. Telvar took our historical order data and came back three weeks later with a model that was noticeably better at capturing our Chinese New Year and Hari Raya demand patterns. The explanation of why it worked — not just that it worked — made it much easier to explain to our board.
Three engagements in detail
Retail importer: overstock after Deepavali
A Singapore retail importer of household goods had accumulated three consecutive years of post-Deepavali overstock, creating storage cost and markdown pressure. Their existing purchasing was based on flat year-on-year uplift applied manually by the buying team.
Demand Forecasting Model — 4 weeks
A demand forecasting model was built using five years of weekly sales data, incorporating Deepavali date shifts, pre-holiday purchasing lead times, and post-event markdown patterns. Three SKU categories were modelled separately due to different demand profiles.
Purchasing confidence improved
The buying team used the model's output for the following Deepavali cycle. The uncertainty ranges allowed them to set a conservative purchase floor and a stretch ceiling. Post-event overstock in that category was significantly reduced without a meaningful sales loss.
B2B services firm: high-value client retention
A Singapore-based professional services business had noticed that clients tended to disengage after 18–24 months but could not identify which early behavioural signals predicted this. They had 3 years of CRM interaction data and billing records.
Customer Behaviour Analysis — 3 weeks
The engagement focused on two questions: which interaction patterns preceded disengagement, and whether clients who received proactive check-ins showed different retention rates. The analysis used cohort segmentation across the CRM data and billing history.
Two actionable early indicators identified
The analysis surfaced two CRM engagement signals that consistently appeared 3–4 months before disengagement. The account management team restructured their check-in schedule around these indicators and began tracking them as standard KPIs in their reporting cycle.
Healthcare equipment distributor: planning data in poor shape
A medical equipment distributor wanted to commission demand forecasting to support import planning. Their leadership assumed their ERP export data was modelling-ready. Telvar suggested starting with the Readiness Check to confirm this before committing to a larger engagement.
Readiness Check — 6 days
The assessment found significant gaps: product category labelling was inconsistent across years, backorder events were not flagged in the data, and two product lines had data gaps of 4–7 months. The report prioritised the five issues by their impact on model reliability.
Data remediation completed in 10 weeks
The IT team used the gap analysis to guide ERP configuration changes and historical data back-filling. The client returned 10 weeks later with a cleaned export and commissioned the forecasting engagement. The assessment cost was credited against the forecasting project fee.
Work completed to date
Ready to add your own experience?
A scoping call is free and takes 20–30 minutes. That is usually enough to establish whether any of our three services fits what you are working on.
Start a ConversationThe right engagement starts with the right question
Tell us what you are trying to understand. We will work out together whether our services can help — and if not, we will say so.
Get in Touch