Stop paying for customers who'd buy anyway.
WeaveSight estimates which customers are likely to be moved by your message, offer or incentive — then gives lifecycle and growth teams a budget-sized action list with an incrementality readout. No direct identifiers required — you re-identify and activate inside your own systems.
Pseudonymised keys only — synthetic figures: expected extra orders per customer, ranked high-to-low and re-targetable to any budget. The bottom two get no offer for different reasons: one would order anyway (~0.00 — protect the margin), the other would order less if nudged (−0.13 — leave alone).
3,200 → 5,100
$15.60 → $9.80
$42k → $68k
Synthetic illustration. Assumptions: contribution margin, incentive cost and fulfilment cost. Not benchmark data.
Most teams can't tell incremental growth from the sales they'd get anyway.
Given a business goal and an audience — everyone, or a specific group like newly-registered customers — every campaign turns on two questions: who should get an action, and what action is worth the cost. Standard reporting shows who received the campaign and what happened after — not what would have happened without it, so attributed sales get mistaken for incremental growth.
Set the budget and pricing — often blind to what's incremental.
A vague ask handed to analytics, downstream of the real decision.
Shows who looks alike — but not how they'd behave without the action, or how to design one that moves them.
Design an action from proxies and past experience — a guess at who responds and what works.
After launch there's no what-if baseline — standard reporting can't show what would have happened without the action. "They saw it and ordered" gets reported as impact, even though most would have ordered anyway. Standard analytics isn't built to answer that what-if question — so marketing sees attributed sales, finance sees spend, and the data team is left explaining why correlation isn't growth.
Pick an audience and a goal — everyone in it responds one of four ways.
Choose who you're acting on and what you want — say, churned customers you'd like to reactivate (just one example). Within that audience, WeaveSight sorts everyone by how they're estimated to respond: who to treat, who to keep warm without spending, and who to leave alone.
Persuadable
Estimated to act because of the action — a minority of customers, but most of the incremental growth. Prioritise them when the expected incremental value beats the cost.
Would-buy-anyway
Likely to act anyway — so don't spend the incentive on them. Keep a light, no-cost touch instead of a discount.
Unlikely to move
Low estimated response under the tested actions — not worth incentive spend this campaign. Save it, test a different lever, or route them into research.
Do-not-disturb
Estimated negative response — higher risk of opt-out, annoyance or churn. The best action may be no action.
Personalised when it pays — sometimes the best answer is one good offer for everyone, or none at all, and we'll say so. Either way, you get the customer-key list per group, ready for your tools.
Any lever, any lifecycle moment.
WeaveSight decides who gets which — under budget, with an incrementality readout — whatever the lever or the moment.
Spend on customers whose behaviour is likely to change.
Most discount, message and incentive budgets are paid to people who'd act anyway — or never would. WeaveSight gives you the recommendation, the budget logic and the evidence behind it — not another dashboard to interpret, and not a one-off analysis to rebuild before every campaign. Under the hood, causal uplift modelling, validation checks and budget optimisation estimate each customer's incremental response and recommend the next best action: robust, rigorous, fast.
A decision readout
Causal validation and confidence bands — so the number is an honest lift estimate, not a correlation dressed up as impact. We run the method; you get the readout and the assumptions behind it.
A per-customer action list
key → recommended action, expected extra orders, confidence and cost — ready for your CRM/campaign tool, sized to a budget. Decisions on your own keys; no names, emails or phone numbers required.
A budget & ROI plan
Spend → expected incremental orders with a confidence band, cost per incremental order, break-even and scenarios for finance.
You get the decision. We carry the rigour.
A clear readout of what's estimated to be incremental, a per-customer action list ready for your tools, and the assumptions and validation behind it — built on a minimised pseudonymised extract, so direct identifiers stay inside your systems.
No direct identifiers required — your customer identity stays with you.
WeaveSight is a decision layer, not a data grab. We need only a pseudonymised extract you control — not names, phone numbers or emails — which you re-identify and activate on your side.
A pseudonymised extract
A hashed customer key, a few binned pre-campaign features, what each customer was sent, and what they did. Your team provides the extract, or grants restricted read access to a prepared view with only these fields. No direct customer identifiers in the modelling dataset.
We return decisions on your keys
Each key gets its recommended action, uplift, group and cost. We don't receive names, emails, phone numbers or direct identity fields.
You re-identify and act
You join our decisions back to your customers on your side and send through your own tools. Your customer identity stays inside your systems.
We receive no direct identity fields — only an approved pseudonymised extract or a restricted warehouse view (customer key, binned pre-campaign features, campaign exposure, outcome and cost / margin). Most pilots start from a client-prepared extract or CSV; where it helps, we define the SQL. Names, emails, phone numbers and direct identity fields aren't part of the approved extract — they're rejected at intake. Retained — with no direct identifiers — as your knowledge repository for the contract and deleted on contract end (or on request); backups per the agreed DPA schedule · NDA / DPA on request. Read the full data-handling & security note →
What your team gets in the validation pack.
WeaveSight isn't magic, and won't pretend to be. The proprietary workflow stays protected — the reasoning behind every recommendation doesn't.
✓ Assumptions and limitations
✓ Confidence bands where appropriate
✓ Sample-size warnings
✓ Leakage & bias checks
✓ Control-group readiness assessment
✓ Recommendation logic in plain English
✓ Budget & ROI readout
✓ A clear "evidence too weak — don't act" flag
One decision readout for lifecycle, commercial, finance and data science.
Each team reads the same per-customer (uplift, cost, confidence) through its own lens.
Lifecycle / CRM
Who to message, with what offer, under what budget — and who to leave alone.
Commercial / Growth
Which levers create incremental demand — and where the next dollar goes.
Finance
Cost per incremental order, break-even and budget scenarios.
Data Science — the partner, not just a stakeholder
WeaveSight is the rigorous causal capability — uplift modelling and validation, made trustworthy — productised. Your team validates what's truly incremental and aligns the business on it, without diverting a year of roadmap into internal tooling before the business gets a usable decision, or fighting correlation-mistaken-for-impact alone.
Selected future pilots: pricing, product-feature impact, website / page changes, and sales motions.
Start where budget leaks fastest.
We lead with lifecycle and promo Next Best Action — where budget leaks fastest and the payback is clearest. The same engine extends to retention, then pricing and product, once the decision layer is trusted.
| Use case | The lever | Outcome | The decision |
|---|---|---|---|
| Promo / discount live | offer / incentive | orders, GMV | who to treat, under budget |
| Lifecycle live | message / channel / timing | first order, reorder, retention | who to message, who to leave alone |
| Retention / churn next | win-back / save offer | reactivation, retained revenue | who to save, who to let go |
| Pricing roadmap | price point | revenue, margin | how deep, raise or hold |
| Product feature roadmap | feature on/off | adoption, retention | roll out, redesign, who to research |
We don't pitch theory — here's the workflow, end to end.
"Nivora," a fictional online retailer, ran a weekly discount to everyone and believed it drove most of their growth. WeaveSight designed a study where their usual offer was one of the options tested — so we could estimate what it was doing. Synthetic data; represents no real company.
What we found
- Much of the discount appeared non-incremental — a large share went to customers who'd have ordered anyway.
- Deeper isn't better — a heavy discount burned far more margin for about the same order lift as a modest one.
- More orders ≠ more customers — much of the lift was pull-forward; repeat behaviour barely moved.
- Personalising per customer didn't pay here — within noise of one good uniform offer. So we said so.
What Nivora got
- A plain-English readout — what moved customers, where the money's going, what to do next.
- A per-customer action list (each key's group + recommended action), re-runnable every cycle to any budget.
- The honest verdict: stop paying for the sales you'd get for free, and redirect that budget to the customers likely to respond — the same spend, more incremental growth (or banked as margin — your call).
| Synthetic readout · one promo cycle | Blanket promo | WeaveSight-targeted |
|---|---|---|
| Campaign budget | $50,000 | $50,000 |
| Attributed orders | 18,400 | 15,900 |
| Estimated incremental orders | 3,200 | 5,100 |
| Discount spent on would-buy-anyway | $21,000 | $8,500 |
| Cost per incremental order | $15.60 | $9.80 |
| Expected contribution | $42,000 | $68,000 |
Fully synthetic — assumptions: illustrative contribution margin, incentive cost and fulfilment cost. Built to show the shape of the decision, not benchmark data.
Finance note: the readout separates attributed orders from estimated incremental orders, and shows discount cost, cost per incremental order, contribution and break-even. Contribution = gross margin after discount / incentive and fulfilment cost, replaced with your finance-approved assumptions.
And the budget you reclaim isn't just saved — it's fuel. Put it into bolder bets and new channels, deeper engagement with the customers who respond, and the user research to learn what's working and what isn't. Cutting the waste funds the exploration that compounds growth.
Illustrative, synthetic, and shaped like real results — built without touching anyone's data. On your programs it runs on your own pseudonymised extract, used only for you.
Built by a practitioner who turns causal analysis into growth decisions.
WeaveSight is built by Ehsan Karim — a data scientist who has spent over a decade building causal growth-decision, uplift, and experimentation systems for global fintech, marketplace and SaaS companies. WeaveSight productises that work: rigorous causal methods plus seasoned data-science strategy, delivered as honest, budget-aware decisions.
- Built uplift and incrementality workflows for real growth and budget decisions
- Across marketplace, fintech, SaaS and digital-first growth systems
- Designed experimentation and observational causal analysis for commercial teams
- Translated data-science outputs into finance, lifecycle, product and leadership decisions
We work with a small number of companies at a time — founder-led delivery, pilot terms, and a direct hand in shaping the product.
What WeaveSight is, in plain answers.
What is WeaveSight?
A causal growth-decision service that estimates each customer's likely incremental response and recommends the Next Best Action — the right communication or incentive — across onboarding, lifecycle, retention and promotions, under a fixed budget. It validates the estimate with confidence bands and delivers a ready-to-use action list. No direct identifiers required; the proprietary workflow stays protected, while your team gets the assumptions, validation and limitations.
How is this different from normal customer targeting?
Normal targeting groups customers who look alike. WeaveSight targets by who is estimated to change behaviour because of an offer. Only that tells you who to spend budget on — it separates the customers an offer moves from those who'd act anyway or never would.
Which customers should get a discount, voucher, cashback or free delivery?
Only the persuadables — those estimated to be moved by that specific offer. WeaveSight ranks every customer per lever, so you spend each incentive — a voucher, cashback, free delivery, BOGO or loyalty perk — only where it's likely to change behaviour, and suppress the would-buy-anyways.
Who should get a welcome, nurture, re-engagement or win-back message?
WeaveSight estimates who each lifecycle message is likely to move — across welcome / onboarding, nurture, re-engagement, win-back and cross-sell — so you contact the customers likely to respond, leave the rest alone, and size it to budget.
Does it work for product features and website changes too?
Yes — in selected pilots. The same causal decision logic extends to other changes — a price, a feature, a website variant. It shows who wouldn't have converted, who'd convert anyway, and who the change was estimated to move — plus the list of customers to recruit for research.
Do you need direct customer identifiers?
No — not in the modelling dataset. Names, emails, phone numbers, addresses and device IDs aren't part of the approved modelling dataset — they're rejected at intake. You provide a minimised extract with a hashed (pseudonymised) key, a few binned features, what each customer was sent and what they did — and we return decisions on your own keys, which you re-identify on your side. (A re-linkable key can still be personal data under GDPR; you remain the controller.)
What does WeaveSight deliver?
A clear verdict on what's incremental, the four customer groups, a budget/ROI plan with a confidence band, and a per-customer action list (key → recommended action) that re-targets to any budget.
Is WeaveSight a fit for us?
Best fit if you run recurring lifecycle or promo campaigns, have customer-level outcomes, know your offer cost/margin, and can export campaign history (or hold back a small control group). If the data's too thin to be sure, we'll tell you. Not a fit if you can't link campaigns to customer-level outcomes, don't know offer cost/margin, or can't create a small control or comparison group.
Do you replace our CRM or agency?
No. We don't replace your CRM, campaign tools, agency or lifecycle team — we hand them an incrementality-backed customer/action list and a measurement readout they activate in your existing stack.
From first call to your first action list.
Timeline depends on your team's setup, engineering and analytics support, and how long the incentive program runs — typically 4–12 weeks.
Clarify the campaign, audience, goal, budget, margin and action options.
Get the data — a client-prepared pseudonymised extract or CSV, or restricted access to a prepared view. We say early if it's too thin.
Once the program has run, estimate who was likely moved, who was over-incentivised, and where the evidence is weak.
Deliver the next-campaign action list, budget scenario, assumptions and validation notes.
If your historical data has valid variation, we can audit past campaigns. If not, we design the next campaign with a small held-back (control) group or comparison so the readout is credible.
Find out where your lifecycle or promo budget is leaking.
Share one campaign, one audience and one goal. We'll tell you whether WeaveSight can help — and exactly what data you'd need. We work with a small number of companies at a time.
or email [email protected]