Skip to content
sayak.webdesignerWeb · Software · Data · AI
Six projects · real numbers

The number that changed, and how

Client names are withheld where the work touches production, pricing or patient data — which is most of it. Everything else is exactly as it happened, including the projects where the most valuable finding was an error in the process we were replacing.

Cement · Data & Reporting

Four hours of manual EOD reporting to zero

Integrated cement group, Eastern India

The situation

Each plant produced its end-of-day report by hand — a person exporting from the historian, keying figures into a spreadsheet that had evolved over years, and emailing the result mid-morning. Three plants, three formats, and no way to reconstruct where a challenged figure came from.

What we did

Two days on site per plant for tag and report inventory, then AURA configured to read OPC tags read-only, apply their formulae, aggregate by shift and day, and publish their exact Excel template to the LAN drive. Parallel run against the manual process for five weeks.

4 hrs → 0
Daily manual effort per plant
06:30
Report on the drive, before the shift
3
Errors found in the legacy spreadsheet
100%
On-premise, LAN-only, read-only
AURA platformData engineeringCement industry
INSIDE YOUR FIREWALL — LAN ONLY, NO INTERNET PATHSOURCE SYSTEMSDCSPLCSCADAHistorianread-only OPCAURA ENGINEWindows service · unattendedValidate readingsApply your formulaeAggregate shift / day / MTDFlag deviation vs targetStore-and-forward bufferREPORT BUILDERyour exact Excel templateLAN SHARED DRIVEEOD_2026-08-08.xlsxEOD_2026-08-09.xlsxAudittag → formulaAD / ACLpermissions
Distribution · Custom Software

Three people re-typing orders, replaced by twenty minutes of exception handling

Industrial distributor, Howrah

The situation

Orders arrived by email, WhatsApp and phone, and three people re-typed them into Tally — frequently until nine at night. Pricing errors ran to several lakh a year because a nine-tier price list was applied by hand.

What we did

A costed process map ranked every process by hours consumed and money leaked. Order capture and dispatch came first: multi-channel intake, a pricing engine with the real slab and scheme logic, credit checks, and direct Tally integration with reconciliation reporting.

−78%
Manual data entry hours
7 months
Payback on labour cost alone
~0
Pricing errors after go-live
2 days
Month-end close, from eleven
Custom softwareERP developmentTally integration
SingleSource of TruthPostgreSQL · auditedFinanceGL · AP/AR · GSTProcurementPR → PO → GRNInventorybatch · bin · FIFOProductionBOM · routing · WOSalesquote → invoiceQualityIQC · IPQC · CoAMaintenancePM · breakdownHR Linkshift · payroll feed
D2C · Commerce

Same traffic, same ads, conversion from 1.1% to 1.9%

Home and living D2C brand, Kolkata

The situation

A bought Shopify theme carrying eleven apps, a 4.2-second mobile LCP, and attribution that disagreed with the bank account. The team was optimising spend against numbers that were quietly wrong.

What we did

App audit with a written verdict on each, then a headless Next.js storefront on Shopify with an image pipeline, a real search index, and a server-side event layer feeding a warehouse where cohort, LTV and contribution margin could be computed honestly.

+64%
Conversion rate
0.8s
Mobile LCP after rebuild
38%
Purchase events previously lost
2.3x
Repeat rate after identity resolution
eCommerceNext.jsAnalytics
storefront · Next.jsShopify / Medusacatalog · ordersPaymentsRazorpay · StripeSearchTypesense · AlgoliaLogisticsShiprocket · DelhiveryEvent Busevery click → KafkaWarehousecohorts · LTV · RFMRetention AIwin-back journeys
Steel · Plant Systems

A section running 5 points below yield for a year, found in six weeks

Rerolling mill, Durgapur belt

The situation

Yield was reported monthly as a single aggregate around 94%. Nobody could attribute loss to a section, a shift or a cause, so improvement conversations were general rather than specific.

What we did

Heat-level identity captured at existing material-handling points, yield decomposed into melting, casting, scale, crop and finishing, and each component attributed by shift, grade and section. Specific energy added at heat level with time-of-day tariffs applied.

+1.4%
Yield within two quarters
−6%
Specific energy consumption
−22%
Rejection rate
0
Capital investment required
Data engineeringManufacturingControl-room dashboards
Sinter / DRIburden mixBlast / EAFheat no. 4,182LRFchemistry trimCasterbillet / slabReheatsoaking curveRolling MillTMT / HR coilHEAT-LEVEL TRACEABILITY — every coil traced back to ladle chemistry, mould level and roll passYield +1.4%Sp. energy −6%Reject rate −22%Auto mill balance report
EPC · Data Collection

Fourteen months without missing a tender notice

EPC contractor, Kolkata

The situation

Two people checked government and PSU portals every morning and still missed notices — which in this business means missing a quarter. Dozens of portals, no common format, short response windows.

What we did

A monitored collection system across the portal set with strict field validation, per-field fill-rate alerting, document retrieval with text extraction, and a deduplicated structured digest delivered before 07:00 each day.

0
Notices missed in 14 months
07:00
Digest before the working day
~2 hrs
Median time to repair a portal change
2 people
Redeployed to bid preparation
Web scrapingData engineeringAutomation
FrontierURL queue · priorityHEADLESS BROWSER FARM — 240 concurrentProxy MeshIN · residentialSG · datacenterDE · mobileUS · residentialPolitenessrobots · rate · backoffAnti-bot logicfingerprint rotationParseschema mapValidatepydanticStoreS3 + Postgres12M pages / week99.1% parse yield
Distribution · AI

Supplier invoices: 94% straight through, 20 minutes of review a day

Distribution group, Kolkata

The situation

Two people spent most of the working day keying supplier invoices into the accounting system. Volumes were growing and the next step was a third hire.

What we did

A document intelligence workflow with a 150-case evaluation set built before any implementation. Extraction, validation against master data and GSTIN checks, confidence thresholds routing uncertain cases to a human, and every decision logged against the source document.

94%
Straight-through processing
6.5 hrs
Daily manual work removed
20 min
Exception review each morning
0
Additional hires needed
AI automationDocument intelligenceERP integration
Layout parsetables, stamps, handwritingField extractionGSTIN, HSN, qty, rateHuman reviewonly < 92% confidenceStructured output → ERPvendorOrient Cement Ltdinvoice_noOCL/26/00871date2026-08-04taxable₹ 18,42,300gst_18₹ 3,31,614confidence0.987
Why the numbers are specific

We record a baseline before we build

Improvement that was never measured against a starting point is an assertion. Every figure on this page has a before as well as an after.

They ran our new pipelines alongside the old ones for six weeks and found four places where our existing reports had been quietly wrong for years. That alone was worth the project.
ASAnirban SenHead of MIS · Multi-plant manufacturer
We knew our yield was around 94 per cent. What we did not know was that one section on night shift was running at 89, and had been for a year. Finding that paid for the entire system.
WMWorks ManagerRerolling mill · Durgapur belt
We had two people checking tender portals every morning and still missed things. Now we get a structured digest at 7 AM with attachments, and we have not missed a notice in fourteen months.
SBS. BhattacharyaBusiness Development Head · EPC contractor, Kolkata
Reading this honestly

How to read any agency portfolio, including this one

Portfolios are marketing documents and should be read as such, ours included. Every agency shows its best work, describes its role generously, and quotes the most flattering number available. Knowing what to probe turns a portfolio from a sales asset into genuinely useful evidence.

Ask what the supplier actually did on each project. A great deal of showcased work involved a team of five of whom the agency supplied one, and the case study will not distinguish. Ask who else was involved and what they contributed. Ask when it was built, because a site shipped in 2019 tells you little about a team that has turned over twice since.

Probe the numbers hardest. A claim that a redesign increased conversions by a large percentage is meaningless without the baseline, the period, and what else changed — if a new advertising campaign launched the same month, the site may have contributed little. Reputable suppliers can tell you what was measured and how. Ask whether the improvement held twelve months later, because a great many launch-week gains do not.

Then ask for a client who will speak to you, ideally one whose project had a difficulty. Every long-running supplier has projects that went wrong; a supplier who claims otherwise has either not been doing this long or is not being straight with you. What matters is what they did about it, and the client is the only reliable source for that. We will give you those numbers, including the projects where the honest answer is that the client did the hard part and we built what they specified.

Six questions for any portfolio

  • What exactly did the supplier do, and who else was on the team?
  • When was it built, and is that team still in place?
  • What is the baseline behind every percentage quoted?
  • What else changed in the same period that could explain the result?
  • Did the improvement hold a year later?
  • Can you speak to a client whose project hit a difficulty?
What clients say

Rated 4.7 out of 5 across 64 reviews

Verified on Google Business Profile

Read all reviews
Direct

Enquiries now arrive with specifications already attached

Our previous website had a stock photograph and a paragraph about commitment to quality. It brought us nothing in four years. Sayak spent two days on our shop floor before drawing anything, and what they built lists our machine capacities, tolerance ranges and certifications in a way a purchasing officer can actually evaluate. The change was not in the number of enquiries so much as in their quality — people now call having already decided we can do the job, so the conversation starts at commercial terms.
Arindam SenguptaManaging Director, Sengupta Engineering WorksHowrah ·
Email

They talked us out of half of what we asked for

We went in with a long list of features we were certain we needed. They built roughly half of it and explained clearly why the rest would be maintained forever and used by nobody. Six months on they were right about every item. The store is fast on phones, which matters because that is where almost all of our traffic comes from, and our team updates the catalogue ourselves without calling anyone.
Priyanka RoyDirector, Roy ExportsKolkata ·
Direct

Parallel running for a full month meant nobody had to trust it blind

Replacing a system a clinic depends on is frightening, and most vendors we spoke to proposed a weekend cutover. Sayak ran the new system alongside our registers for a full month and only switched once the numbers matched every day. Our front desk staff were part of the design rather than being trained at the end, which is why they actually use it. Report preparation that took a person most of a morning now takes minutes.
Dr. Subhankar DasAdministrator, Das Diagnostic CentreKolkata ·
Next step

Tell us the number you want to change.

Enquiries per month, days to close month-end, hours of manual entry, cost per tonne. If you can name it, we can usually tell you within a call whether it is addressable and roughly what it takes.

Direct line

+91 70033 91355

Mon–Sat · 9:30 AM – 7:30 PM IST · Sealdah, Kolkata

Reply within one working hour NDA signed before any brief Fixed-price option on every scope
Verify us independently

Don’t take our word for it — ask an AI

Every page on this site is written to be read and quoted by AI answer engines, and we publish llms.txt and llms-full.txt so they can read us properly. One click opens the assistant with the question already typed.

The question we will ask for you

What projects has Sayak Web Designer (sayakwebdesigner.in), an IT company in Kolkata, India delivered, and what results did they produce?

Opens in a new tab. We do not see your conversation.

Call now WhatsApp Get quote