How to forecast demand when you don't have a data science team
A practical way to build a first demand forecast with a spreadsheet or a few lines of Python, and how to tell whether it is any better than guessing.
Data analytics and AI consulting that works for you
InsightsQ is a data analytics consultancy. We build the tracking, pipelines, dashboards and predictive models that let product, marketing and operations teams trust their numbers, see what is coming and act on it.
Our team has worked for Fortune 50 and Fortune 500 companies on
- Predictive models
- Customer analytics
- Data pipelines
- Experimentation
- Reporting automation
We make your data trustworthy first, then useful.
Most analytics projects fail quietly. The tracking is patchy, the tables disagree with each other, and the dashboard on top looks fine until someone asks where a number came from. We work from the bottom up: get the data right, then build the reports and analysis that depend on it.
- Data strategy and planningA clear plan for what to measure, where it should live and who looks after it.Read more
- Data analysis and insightsAnswers to specific business questions, with analysis you can check.Read more
- Predictive modelling and forecastingModels that estimate what is likely to happen next, so you can plan around it.Read more
- AI solutions and automationPractical uses of machine learning and language models that save time or improve decisions.Read more
- Data visualization and reportingDashboards built around the decisions people make each week.Read more
- Product analyticsUnderstand what users do, where they drop off and what keeps them coming back.Read more
- Marketing analyticsSee which channels and campaigns really bring in customers, and what they cost.Read more
- Customer 360One joined-up view of each customer across web, app, orders and support.Read more
- Data engineering and warehousingPipelines and warehouse models that turn messy source data into tables people can trust.Read more
- Analytics tracking and event auditsFind the gaps and mistakes in your tracking before they distort every report.Read more
- Experimentation and A/B testingTest changes properly, so you launch what works and drop what doesn't.Read more
Know what is likely to happen next, and let AI take the repetitive work
Reports tell you what happened. Predictive models help you plan for what is coming, and well-chosen AI tools take the slow, repetitive jobs off your team. We start with a business question, test on your own data, and keep a person in the loop.
Forecasting and prediction
Demand, churn, cancellations and lead quality, with honest error ranges.
Predictive modellingReading text at scale
Reviews, tickets and survey answers sorted into themes you can count.
AI solutionsAutomation with checks
Reports, tagging and document search that a person can verify.
AI solutions
Try a forecast
Click inside the chart to add a point. Click a point to remove it.
- Trend
- Rising, about 4.8 a period
- Forecast at the end
- 79 (likely 61 to 97)
- Points follow the trend
- 79%
A toy example. Real models use many inputs and are tested on data they have not seen, but the idea is the same: more evidence narrows the range, and one odd point can pull the line.
Machine learning at work: what it changed
Projects from our founder’s earlier roles, each started from a business problem and measured against it. Every one has a plain-English case study.
2,700 hrsof manual review removed
Turning thousands of customer comments into themes
A topic model of reviews and questions, with a Power BI view on top, showed the category team what customers kept raising. Satisfaction measures rose 4% as the team acted on it.
Read the case study6%lower product page exit rate
Ranking products by what shoppers are likely to click
A model predicted click-through rate for each product, and the ranking used those scores so the most relevant items appeared first.
Read the case study4%more revenue
Recommending what to buy next
A recommendation system built from product attributes suggested related items and encouraged customers to add to their basket.
Read the case study2.5% to 4%conversion on a key programme
A better model for choosing who to approach
A predictive model replaced the earlier marketing model for a customer advocacy programme, lifting conversion by about 60%.
Read the case study$20,000extra upsell revenue in a month
Finding upsell opportunities in order data
Market basket analysis surfaced products that sell together, and those findings became targeted upsell rules.
Read the case study35%better marketing efficiency
Segmenting buyers and sellers
Buyer and seller cohorts let marketing send personalised messages and offers to groups that behaved differently.
Read the case study
Solid data is what makes the models trustworthy
Groundwork results from our founder’s earlier roles. Clean tracking and tidy pipelines come first, because a model is only as good as what it learns from.
- 90%fewer data discrepancies
Rebuilt messy JSON from an app backend into a clean star schema, which also made reporting about 50% faster.
Consumer app
- 624 hrssaved every year
Replaced a manual reporting routine with automated Python jobs.
E-commerce operations
- 80%faster queries
Worked with data engineers on new pipelines behind slow, heavily used queries.
E-commerce operations
How an engagement runs
Four stages, in plain terms, so you know what happens and when.
A first conversation
Thirty minutes on your questions, your data and what is getting in the way. No slides, and no obligation to continue.
A written proposal
A plain-language plan with scope, timeline and price, so you know what you're buying before it starts. If the scope is still moving, we give a range and say why.
The work
We build in short stretches and review progress with you every week. Your knowledge of the business shapes each step, so surprises come early and cheap.
Handover
Documentation, a walkthrough for your team and a support window, so you can run it without us.
Led by Achyutanand Jha
Achyut started out analysing field hockey matches in college and writing about what the numbers showed. That curiosity became a career in analytics at Amazon, Myntra, Lowe’s and The Luxury Closet.
He has spent more than eight years close to the data in online retail, marketplaces and software, as an analyst, a data scientist and an analytics engineer. InsightsQ is where he brings that experience to other teams.
From the blog
Practical notes on tracking, reporting, experiments, prediction and AI, written for people who have to use the results.
When it makes sense to hire a data analytics consultant (and when it doesn't)
A practical look at when a data analytics consultant earns their fee, when you'd be better off hiring or waiting, and how to judge one before you sign.
Predictive modelling in plain English: what it can and can't do for a business
What predictive models actually do, the handful of business questions they suit, what data they need and why a model is only the middle of the project.
Questions people ask us
What does a data analytics consultant actually do?
We help a business use its data to answer specific questions. That can mean fixing the tracking that feeds your reports, building the pipelines and warehouse tables underneath them, analysing a problem in depth, or setting up dashboards. We work alongside your team and hand over what we build.
Who do you work with?
We are best suited to product, marketing and operations teams in e-commerce, marketplaces, software and consumer apps, especially teams that don't have a full data function of their own. Larger teams hire us for a specific project.
How does an engagement start?
With a 30-minute call. If we both think we can help, we send a written proposal with scope, timeline and price.
Do you work with remote teams?
Yes. Most of our work is done remotely, with calls and shared documents, so where your team sits is rarely an issue. We agree meeting times that work for both sides.
Can you help us use AI?
Yes, in a practical way. We look for specific jobs where machine learning or language models save time or improve decisions, such as forecasting, sorting customer feedback or automating reports, and we test them on your data before anything goes live.
Do we need AI, or will plain analytics do?
Often plain analytics is enough, and cheaper. We tell you honestly which side of the line your problem falls on, and we don't recommend a model where a clear chart would do.
Which tools do you use?
SQL and Python for most analysis and modelling; Airflow and Airbyte for pipelines; BigQuery, Redshift and PostgreSQL for warehousing; Power BI, Tableau, Looker and QuickSight for dashboards; Mixpanel, Google Analytics and Adobe Analytics for tracking. We work with what you have wherever we can.
Can you help if we don't have clean data yet?
That is the usual situation. A short look at your data tells us what can be answered now and what needs fixing first.
Tell us what you are trying to work out
A few lines is enough. We will reply within two working days, or you can book a call straight away.