A pilot on your own weeks.
It starts with eight sealed weeks of your own history: held by someone on your side, run once, and scored against bars written down before anyone sees a result. A pilot answers one question: would this planning team use Kelynto every Monday?
How a pilot is run. The bars are fixed first, the weeks are run once, and the result is judged in the open. A planning service or software company starts differently, with a sandbox: see the partner path.
What a pilot is.
One team or one region, alongside your current process. Nothing you do today is switched off, and nothing is written back to your planning system.
-
First: the eight sealed weeks
You send 26 weeks or more of weekly sales and plan by store and category, your store list, and in-stock and promotions if you have them. Someone on your side who is not on the planning team keeps the last eight weeks sealed until our set-up is frozen and written down. We then run those eight weeks once and give each one back restated the way your Monday review would read it, with a scorecard and a written finding whichever way it goes.
-
Then, if it earns it: about eight live weeks
A drafted review each week, with one planner editing, confirming or correcting it. The live weeks feed Score and learn, counted only from what your planner did. The first weeks can be run by hand: you send the week's files and we return the review.
-
At each end: a decision
After the sealed weeks, a review with your planning director and the review owner: stop, or go to live weeks. After the live weeks, a readout of the three measures below. Nothing obliges you to buy anything.
The eight sealed weeks, day by day.
In business days from the day your files arrive, as planned for each pilot.
| Day | What happens | Who |
|---|---|---|
| Before day 1 | A mutual NDA is signed, then the pilot agreement and the data processing addendum before the first upload. Your custodian is named. Your materiality bar and the dates of the sealed weeks are agreed and written down. | Both |
| Day 1 | Files loaded, every validation report read with your data contact, and one week from before the sealed period run by hand and read. A 45-minute call with your planner. | Kelynto, your data contact and planner |
| Days 2 and 3 | Corrected files loaded. Settings fixed: fiscal week, time zone, materiality bars, noise guard, category lists, which outside sources are on. The frozen set-up is recorded with a checksum and sent to your custodian. | Kelynto, your admin |
| Day 4 | Your custodian releases the eight sealed weeks and the team's commentary for them. The eight weeks are run once, in order. | Your custodian, then Kelynto |
| Days 5 to 9 | Two of your planners, working separately and without seeing Kelynto's output, judge a sample of the material movements and a random sample of quiet cells. Kelynto builds the scorecard from their judgements and the week restatements. | Your two planners, Kelynto |
| Day 10 | Written result and scorecard delivered. A 60-minute review with your planning director and the review owner. Decision: stop, or go to live weeks. | Both |
- The pace depends on you. Mostly on your planners' time on days 5 to 9 (plan on two to three hours each) and on whether the first files load clean.
- Where it runs is agreed with your IT before day 1. On Kelynto Managed, on Microsoft Azure in Central US, with the data deleted at the end under the pilot agreement's exit terms. Or in your own Azure subscription with the deployment kit, through an enterprise implementation set up with your platform team ahead of day 1.
What you get back.
Three things, whichever way the result goes.
-
Your weeks, restated in three parts
For each of the eight weeks, by store and category, rolled up to region and total: what was recorded that week with an estimated amount (a stockout, a store closure, a promotion); what was on file near your stores with no amount attached (weather, local events, road work, a competitor record), because the evidence does not support an amount; and what is not explained. Readable in the application, and as Word, PDF and PowerPoint files and a CSV of every finding.
-
The backtest scorecard
Every store and category checked against your own totals; each kind of cause scored against what two of your planners judged separately; the misses, the false attributions and the share left unexplained. Beside them, the same measures for simple baselines (no cause, the plan gap alone, a seasonal naive rule, a historical anomaly, availability alone), computed by Kelynto from the same files with the code recorded, so you can see whether Kelynto beat the obvious.
-
The written result
Two pages: what held up, what did not, which segments were weak, what was left out and why, and whether a live pilot is worth your team's Mondays.
The rules of the test.
Written down before the seal opens, so that neither side can move them after a result is seen.
- Your custodian seals the holdout, not Kelynto. Kelynto does not hold the eight weeks' actuals or your commentary for them until the set-up is frozen and recorded.
- The bars are fixed before scoring. Your materiality threshold and bands, the store groups, the excluded weeks and which kinds of cause are eligible are written down and signed before the seal opens. A change that a result prompted needs a new sealed period in the future.
- One run. The eight weeks are run once. There is no second attempt with different settings.
- Every result is reported. What the records do not explain is stated as an amount, a kind of cause with too few judgements is reported as unproven, and every segment is in the result, strong or weak.
- Your planners' judgement is the reference. Two of your planners judge the samples separately, before they see Kelynto's output. Planner agreement says how often a planner accepted what the review recorded.
- Amounts are scored only where a defensible label exists. Where none does, the amount is reported as unknown.
- Your data stays yours. No model is trained on it, and nothing is written back to your planning system. The engine computes every number; the sealed weeks are written by Kelynto's built-in writer, and the AI writer is your admin's choice for the live weeks. Everything is deleted at the end under the exit terms, with a deletion record produced by the software.
The data a pilot needs.
Your own weekly numbers, in your own column names and your own calendar. Totals only: no transactions, baskets, shopper, loyalty, payment or employee data, because there is no field for any of them.
| Item | What | How much |
|---|---|---|
| Stores | Store code, region, latitude and longitude | Every store in scope |
| Sales | Store, week, category, actual sales, plan sales (the plan as it was approved, not as later restated) | 26 weeks or more, ending with the eight sealed weeks; 34 or more lets every sealed week stand on a full 26 weeks of history |
| In-stock | Store, week, category, in-stock rate | The same weeks, if you measure it. It is what lets the review record a stock-out |
| Promotions | Week, category, store if store-specific, type, discount | The same weeks, if you have a calendar |
| What only you know | Remodels, closures, competitor openings, with dates and stores | Optional, any shape |
| Your own reviews | The weekly commentary your team actually sent for the eight sealed weeks | Held by your custodian with the sealed weeks |
- CSV or Excel workbooks. Your own column names are fine. Two facts are needed before the first load: the day your fiscal week starts and the time zone most stores are in.
- A file recipe for your planning system. If your plan comes from a named planning system, there is a file recipe for its export: which export carries what, how that system defines the week, how promotions and availability appear, and the traps. Each recipe says what to confirm with your own administrator.
- In-stock is the most valuable addition. It is what makes a stock-out visible in the review: the record shows that availability dropped, store by store, with an estimated amount beside it.
- Store coordinates matter. Local signals are matched to stores by location. Without coordinates, weather, events and road work cannot be matched.
- A bad export cannot overwrite good data. The loader reports what it read row by row and rejects a file outright when more than one row in five has an error.
- A recurring export can be held to its own totals. Send control totals with each weekly file and Kelynto reconciles them to the cent before anything loads. A file that does not reconcile is held for your analyst to decide, and under a source profile so is a file whose columns changed.
- Last week's recap workbook as it is. The older Excel .xls as well as .xlsx and CSV, and a sheet laid out as a grid, with stores down the side and weeks or categories across the top, read as rows before the columns are mapped. The workbook's own totals row is read as control totals.
Success criteria, fixed before anything is scored.
For the sealed weeks, the bars of the scorecard are signed before the seal opens. For the live weeks, three measures from Score and learn, counted from what your planners actually did. They are targets, set before the pilot starts.
- 20%or less of the draft rewritten. Editing a few words means the draft is doing the work.
- 60%or more planner agreement: your planners' labels that confirmed what the review recorded, out of all their labels.
- 4 of 5or higher when planners are asked whether they would use it every week.
Each figure shows its sample size and reads "insufficient data" until there are enough reviews, labels or answers behind it. If your own bar is different, say so at the start so it can go into the pilot agreement.
Who does what.
What a pilot asks of your team, and what Kelynto does.
- A data contact who can export the files and answer questions about them. Kelynto loads, validates and reports on every file, with every problem located by row and column.
- A custodian outside the planning team who holds the eight weeks and the commentary, and releases them only against the frozen set-up record. Kelynto changes nothing after it.
- Two planners who judge the samples separately, two to three hours each on days 5 to 9, and a review owner who sets the materiality bar.
- A sponsor who receives the result.
- For live weeks, a weekly file and a planner's attention each week.
- On Kelynto's side, you work directly with the founder, Marc Estinville.
Cost, paper, and what happens after.
Pricing depends on deployment model, number of locations and pilot scope. The fastest way to get a number for your situation is a 20-minute conversation.
- The fee for a pilot is agreed per pilot, in that conversation.
- Included: everything on this page for the sealed weeks: loading, the frozen set-up, the run, the scorecard, the written result and the two calls. Not included: your own staff time, your cloud costs if the software runs in your subscription, and anything after the decision on day 10.
- The paper. A mutual NDA before anything is shared, then a pilot agreement and a data processing addendum before the first upload, and for the customer-hosted option a statement of how any access by Kelynto is granted. Drafts of each are ready for review by counsel.
- You own your data, the drafts and the labels your planner creates, and you can export them. At the end everything is deleted under the exit terms.
What you work with
- The whole weekly loop, built and rehearsed end to end, with a product tour on a synthetic retailer
- Kelynto Managed on Microsoft Azure in Central US, deployed by an automated pipeline with health checks and automatic rollback
- Every security and data control with its status and evidence in the Trust center
- A founder with more than 17 years in retail supply chain and planning technology, working with your team directly
Five ways to take the next step.
-
Explore
See the product
The product tour shows real screens of the product on a synthetic retailer, in the order a week runs.
A private copy of a synthetic retailer, Harborview Markets. No conversation needed first.
-
Evaluate with your data
Run it on your own weeks
Eight sealed weeks of your own history, run once and scored against bars fixed before anyone sees a result, then about eight live weeks.
Create a workspace and load your own files. Sign in by an emailed link. No password to create.
-
Review the security
Check every control
Kelynto Managed on Microsoft Azure or in your own Azure account, what crosses the boundary, and every control with its status and evidence.
-
Offer it to your clients
Build it into your service or product
For planning services, retail software vendors and consultancies: the partner API, embedded views and your brand, starting with a sandbox of synthetic clients.
-
Talk to us
Twenty minutes with the founder
Your process and your questions. Or write to marc@kelynto.com.