AI decisioning for retail customer value · pilot programme

One export.
Per-customer offers.
Revenue you can prove.

Tailo's decisioning layer, no integration project. Send a pseudonymized export, get your first next-best-offer recommendation back free — on your real customers. Run one campaign. We measure the lift against a holdout. You pay only once you've seen it work.

Not ready to send data?
Free first recommendation Measured against holdouts Pay only once it works
customers_export.csv · 312,419 rows
customer_idlast_orderorders_12maov
a91f…3c41 days ago6€34
7be2…9d9 days ago14€61
c044…1a163 days ago2€27
TAILO DECISIONING→ 7 DAYS →nbo_output.csv
customer_idnext_best_offerpred_upliftgroup
a91f…3cRestock bundle, no discount+8.4%treat
7be2…9dEarly access drop+6.1%treat
c044…1aWin-back, free delivery+11.7%hold
Works with the export button you already use
Proof

The same decisioning layer.
Measured the same way.

The pilot runs the same engine as Tailo's full CVM deployments. Only the data path changes.

+32%

incremental revenue vs. matched control

−18%

discount waste in 4 months

2.4×

more reliable than attribution reporting

Results from Tailo CVM deployments. Built by a team with experience at retail & ecommerce brands across Europe, the Middle East, Africa and Asia.

Silpo Askona Keells X5 Group Woolworths
How it works

Reads your data. Decides per customer.
Proves the uplift. In one campaign cycle.

You keep sending from your own ESP. Tailo does the modeling and the measurement. Nothing in your stack changes.

Week 1 · Reads

You export

Customers, transactions, campaign history, offer list — hashed IDs only, straight from your ESP. We send a one-page spec and check data quality first.

Your effort: ~2 hours
Week 2 · Decides

Tailo decides

The model reads each customer's signals and picks the one offer with the highest predicted uplift — who to reach, what to send, when, and who to leave alone.

Your effort: none
Weeks 3–5 · Runs

You send

Load the NBO file into your ESP and run the campaign as usual. The holdout gets your business-as-usual treatment — that's the control we measure against.

Your effort: one campaign setup
Week 6 · Proves

Tailo proves it

You send the results back. We measure incremental revenue per contact against the holdout and hand you a P&L-ready report: uplift, confidence, discount cost.

Your effort: ~1 hour
What you get back

Files your team can use the same day.

  • 01

    The NBO file

    One row per customer: recommended offer, predicted uplift, send timing, channel — the top-ranked offer for each individual. Imports into any ESP as a custom property.

  • 02

    The holdout design

    A statistically matched control group, sized for your base, so the result is measured — not an attribution story. We design it; you just exclude it from the new treatment.

  • 03

    The uplift report

    Incremental revenue per contact, total incremental revenue, discount spend avoided, confidence intervals. The number you take to your manager — with the math shown.

nbo_output.csv — column spec customer_id hash, your key
next_best_offer from your offer list
offer_rank_2 fallback offer
pred_uplift % vs. no contact
send_window date range
channel email / sms / push
group treat / holdout
do_not_contact suppression flag
Why not just use a chatbot

"Can't I just paste my data
into ChatGPT or Claude?"

Fair question — they're brilliant at language. But choosing the next-best-offer for a million customers isn't a writing task, and the gap shows fast.

General chatbot

Guesses from patterns

It predicts plausible text. Ask for an offer and you get something that reads right — with no model of your customers' actual behaviour, and no way to tell a good guess from a bad one.

General chatbot

Can't prove anything

No holdout, no control, no measurement. You get recommendations you can't defend to finance — the opposite of what this offer delivers.

Tailo decisioning

Models, decides, measures

Purpose-built uplift models trained on your transactions, a matched holdout designed in, a measured revenue report at the end. A result, not a suggestion.

And the part that should stop you cold: pasting customer data into a public chatbot is a data-protection incident waiting to happen. It leaves your control, may be retained or used for training, and likely breaches your GDPR obligations and your own privacy policy — no DPA, no retention limit, no deletion. Tailo takes hashed IDs only, under the DPA you agree to before transfer, deleted after the pilot. Same AI horsepower, built for the job, safe to put in front of legal.

Data safety

Your customers stay yours.
Their names stay home.

The model never needs to know who anyone is — only what they bought and how they responded. So that's all you send.

Pseudonymized by design

Hashed customer IDs only. No names, emails, or phone numbers ever leave your systems — you map results back to real customers on your side.

DPA before data

You review and agree to Tailo's DPA before the first file moves. GDPR-compliant by default — clear terms for legal review, without extra paperwork.

Secure data transfer

Use your company's approved secure channel, or an encrypted upload link we provision — never email attachments. Access is restricted to approved users, transfers are logged, and files are handled only within the agreed pilot scope.

Deleted after the pilot

When the pilot ends, your data is deleted within 30 days, confirmed in writing. Continue with Tailo, and you decide what's retained.

How it works commercially

See it work first.
Pay only for proof.

You never pay before Tailo produces real output on your own customers. Three steps — each a smaller decision than the last.

Step 1
€0

Standard first recommendation — free

Send a qualifying export. Get a next-best-offer preview for one segment, up to 50,000 customers — real offers on real customers — plus projected uplift across your whole list. No card, no commitment. Wider scope? A fixed fee, quoted up front.

No payment · no risk
Step 2
€3–6k credited & refundable

The full measured pilot

Like what you see? We build the full NBO file, design the holdout, you run one campaign, we hand back the verified revenue report. €3k up to 250k customers, €6k to 2M — credited against year one, refunded if it doesn't beat your baseline.

Pay after the preview
Step 3
from €2k/mo

Keep the engine running

Continue on a monthly plan by customer count (€2–6k/mo), or go live with the full integrated decisioning layer across your CDP, loyalty, and CRM — quoted separately.

Only if it works
The standard first recommendation is genuinely free; only custom-scope work carries a fixed, agreed-up-front fee. The pilot fee is credited and refundable, and the success metric is agreed with you in advance. You see Tailo work on your own data before any money changes hands.
Not ready to send data?

Start wherever you're comfortable.

Sending customer data — even hashed — is a real decision. So we don't make it the first one. Start on the rung that fits, move up when you're ready.

Lower commitmentHigher commitment
01

Walkthrough

A 30-minute call. We show the engine running on anonymized sample data from a similar retailer — you see exactly what an NBO file and uplift report look like, without touching your own data.

No data from you
Book a walkthrough →
02

Free first recommendation

When you're ready, send a pseudonymized export — hashed IDs only, with the DPA agreed before transfer. We return real next-best-offers on a slice of your base, plus projected uplift. This is Step 1 of the pilot, and it's free.

Hashed data · DPA agreed · deleted after
Start the free preview →
Honest qualification

Built for some teams.
Not for everyone.

The pilot only works when the proof can be real. We'd rather tell you now than waste your six weeks.

A strong fit if you…

  • Have 100k+ customers with purchase history
  • Run regular email or SMS campaigns from Klaviyo, Mailchimp, Brevo or similar
  • Have a loyalty program or identifiable repeat buyers
  • Can hold back a control group for one campaign
  • Are tired of blanket discounts eating your margin

Probably not (yet) if you…

  • Have mostly anonymous, one-time buyers
  • Send fewer than one campaign a month

If you need live decisioning across your CDP, loyalty and CRM stack, the full Tailo integration is the better fit — start there instead.

FAQ

The questions your manager will ask.

What exactly do we need to export?
Four files: customers (hashed ID + attributes), transactions (12–24 months), campaign sends and responses, and your current offer list with margins or constraints. We send a one-page spec with example rows — most teams produce it in under two hours from their ESP or database.
How is this different from the full Tailo platform?
Same decisioning engine, different plumbing. The full platform connects live to your CDP, loyalty and CRM stack and decides continuously; the pilot runs on file exchange, one campaign cycle at a time. Most teams start here, see the measured uplift, then move to the integrated setup.
Couldn't we just do this with ChatGPT or Claude ourselves?
Two problems. First, a general chatbot predicts plausible text — it has no uplift model of your customers and no way to measure whether an offer actually worked, so you'd get confident-sounding guesses you can't defend to finance. Second, pasting customer data into a public chatbot means that data leaves your control and likely breaches GDPR and your own privacy policy — no DPA, no retention controls, no deletion. Tailo uses purpose-built uplift models with a holdout designed in, and takes hashed IDs only under the DPA agreed before transfer.
Is this GDPR-compliant?
Yes. You send pseudonymized data only — hashed identifiers, no names, emails or phone numbers. You agree to Tailo's DPA in the flow before any transfer, and we delete everything within 30 days of pilot completion, confirmed in writing.
How is uplift actually measured?
Against a matched holdout. Part of your base receives your business-as-usual campaign; the rest receive Tailo's per-customer offers. The difference in revenue per contact between the groups is the incremental revenue — a controlled comparison that ties back to the P&L, not click attribution.
What if the pilot shows no uplift?
You'll have seen that risk-free. The first recommendation is free, so you only commit after the preview convinces you — and the pilot fee is refunded if the measured result doesn't beat your agreed baseline. The data-quality check up front also exists to catch cases where a pilot can't win: we'll decline rather than run one we expect to fail.
Do we have to integrate anything?
No. The pilot runs entirely on file exchange. If you continue afterwards, you can keep a monthly export cadence or move to a live connection with your stack — that's your choice, on your timeline.
Who sees our data?
Only the Tailo team working on your pilot, under the DPA. Your data is never pooled into other clients' models without your explicit written approval, and never used for anything beyond your pilot.
Start your free preview

Your next campaign could be the proof.

Send a pseudonymized export, get your first next-best-offer recommendation back free — on your real customers. No meeting required, though if you'd rather talk first, that's fine too.

Prefer to talk first? Book a call
Free first recommendation Hashed data, DPA agreed Pay only once it works